CN116434739A - Device and related components for constructing a classification model for identifying different stages of heart failure - Google Patents

Abstract

The invention discloses a device for constructing a classification model for identifying different stages of heart failure and related components. The device includes a sample processing unit, which is used to convert the collected voice analog signal into a voice digital signal, and perform pre-processing on the voice digital signal. processing, and performing feature extraction on the preprocessed speech digital signal to obtain multi-class speech feature samples; the model training unit is used to construct a classification model for identifying heart failure stages, and use the multi-class speech feature samples to classify the The classification model is trained and optimized to obtain the optimal classification model. Using the optimal classification model, the different stages of heart failure can be identified more accurately.

Description

Device and related components for constructing a classification model for identifying different stages of heart failure

Technical Field

The present invention relates to the field of computer technology, and in particular to a device for constructing a classification model for identifying different stages of heart failure based on sound features, a computer-readable storage medium and a computer device.

Background Art

Heart failure is a group of complex clinical syndromes caused by abnormal changes in cardiac structure and/or function due to various reasons. It is a serious and terminal stage of various common cardiovascular diseases. At present, according to the development process of heart failure, it is divided into four stages: heart failure risk stage (stage A), pre-heart failure stage (stage B), heart failure stage (stage C) and end-stage heart failure stage (stage D). The purpose of staging heart failure is early detection, early diagnosis and early intervention, especially for timely identification and treatment of patients at risk of heart failure and pre-heart failure stage. Early intervention is of great significance for delaying ventricular remodeling and progression of heart failure, protecting cardiac function, improving quality of life and reducing re-hospitalization rate.

In the existing technology, the identification of heart failure stages usually relies on medical history, physical examination, laboratory tests, cardiac imaging tests and functional tests. These tests are often performed only when patients seek medical treatment due to symptoms, which is not conducive to early identification. In recent years, the development of hemodynamic or lung water content monitoring through implanted devices, such as CardioMEMS, MultiSENSE, ReDS and other sensor devices, as well as HeartLogic multi-sensor index and alarm algorithm evaluation, can achieve early warning of patients' heart failure decompensation events. However, the equipment of the above methods is expensive and invasive, requiring implanted sensors or pacemakers to be installed. It is only suitable for a small number of patients with severe or refractory heart failure, and is not suitable for screening patients in the risk period of heart failure and pre-heart failure stage. On the basis of strengthening standardized diagnosis and treatment and patient education, developing non-invasive, convenient and universal monitoring and early warning methods, identifying patients with heart failure at different stages, and strengthening home monitoring and early warning are the key to reducing the re-hospitalization rate and mortality in the management of chronic heart failure. The prior art also has technologies for training and learning certain parameters to identify classification models of certain diseases, but the data used by these classification models also rely on a lot of examination data, and the recognition accuracy is not high.

Summary of the invention

The embodiments of the present invention aim to provide a device for constructing a classification model based on sound features that can accurately identify different stages of heart failure, a computer-readable storage medium, and a computer device.

In a first aspect, an embodiment of the present invention provides a device for constructing a classification model for identifying different stages of heart failure based on sound features, comprising:

A sample processing unit, used for converting the collected voice analog signal into a voice digital signal, preprocessing the voice digital signal, and extracting features from the preprocessed voice digital signal to obtain multiple types of voice feature samples;

A model training unit is used to construct a classification model for identifying heart failure stages, and to train and optimize the classification model using the multi-category speech feature samples to obtain an optimal classification model, wherein the classification model for identifying heart failure stage A and heart failure stage B is an AdaBoost classification model based on original variables; the classification model for identifying heart failure stage B and stage C is an AdaBoost classification model based on Lasso dimensionality reduction; and the classification model for identifying heart failure stages A, B, and C is an AdaBoost classification model based on Lasso dimensionality reduction.

In a second aspect, an embodiment of the present invention further provides a computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the following method when executing the computer program: converting the collected voice analog signal into a voice digital signal, preprocessing the voice digital signal, and extracting features from the preprocessed voice digital signal to obtain multiple categories of voice feature samples; constructing a classification model for identifying heart failure stages, and training and optimizing the classification model using the multiple categories of voice feature samples to obtain an optimal classification model, wherein the classification model for identifying heart failure stage A and heart failure stage B is an AdaBoost classification model based on original variables; the classification model for identifying heart failure stage B and stage C is an AdaBoost classification model based on Lasso dimensionality reduction; and the classification model for identifying heart failure stages A, B, and C is an AdaBoost classification model based on Lasso dimensionality reduction.

In a third aspect, an embodiment of the present invention further provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the following method is implemented: converting the collected voice analog signal into a voice digital signal, preprocessing the voice digital signal, and extracting features from the preprocessed voice digital signal to obtain multiple categories of voice feature samples; constructing a classification model for identifying heart failure stages, and training and optimizing the classification model using the multiple categories of voice feature samples to obtain an optimal classification model, wherein the classification model for identifying heart failure stage A and heart failure stage B is an AdaBoost classification model based on original variables; the classification model for identifying heart failure stage B and stage C is an AdaBoost classification model based on Lasso dimensionality reduction; and the classification model for identifying heart failure stages A, B and C is an AdaBoost classification model based on Lasso dimensionality reduction.

The embodiment of the present invention is based on the sound features that can reflect different stages, so as to achieve the purpose of constructing a classification model for identifying the corresponding stages according to different stages, thereby improving the accuracy of model recognition.

BRIEF DESCRIPTION OF THE DRAWINGS

In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying any creative work.

FIG1 is a schematic diagram of the structure of a device for building a classification model for identifying different stages of heart failure based on sound features provided by an embodiment of the present invention;

FIG2 is a diagram showing the effects of different heart failure stages on Jitter provided by an embodiment of the present invention;

FIG3 is a diagram showing the effects of different heart failure stages on Shimmer provided by an embodiment of the present invention;

FIG4 is a diagram showing the effect of different heart failure stages on harmonic difference provided by an embodiment of the present invention;

FIG5 is a diagram showing the effect of different heart failure stages on HNR provided by an embodiment of the present invention;

FIG6 is a diagram showing the effect of different heart failure stages on Alpha Ratio provided by an embodiment of the present invention;

FIG. 7 is a diagram showing the effect of different heart failure stages on voiced/unvoiced duration provided by an embodiment of the present invention;

FIG8 is a diagram showing the effects of Loudness on different heart failure stages provided by an embodiment of the present invention;

FIG9 is a diagram showing the effects of different heart failure stages of the Hammarberg Index provided by an embodiment of the present invention;

FIG10 is a diagram showing the effects of spectral slope on different heart failure stages provided by an embodiment of the present invention;

FIG11 is a cross-correlation coefficient diagram of glottal cycles in different heart failure stages provided by an embodiment of the present invention;

FIG12 is a nonlinear analysis diagram of voice characteristics of different heart failure stages provided by an embodiment of the present invention;

FIG13 is a diagram showing changes in voice acoustic characteristics based on cepstrum for different heart failure stages provided by an embodiment of the present invention;

FIG14 is a sample-level ROC curve diagram of the optimal model (i.e., AdaBoost of the original variables) provided in an embodiment of the present invention;

FIG15 is a sample-level ROC curve diagram of the optimal model (i.e., AdaBoost using Lasso dimensionality reduction) provided in an embodiment of the present invention;

FIG16 is a sample-level ROC curve diagram of the optimal model (i.e., AdaBoost using Lasso dimensionality reduction) provided in an embodiment of the present invention.

DETAILED DESCRIPTION

The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

It should be understood that when used in this specification and the appended claims, the terms "include" and "comprises" indicate the presence of described features, integers, steps, operations, elements and/or components, but do not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and/or combinations thereof.

It should also be understood that the terms used in this specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. As used in the specification of the present invention and the appended claims, unless the context clearly indicates otherwise, the singular forms "a", "an" and "the" are intended to include plural forms.

It should be further understood that the term "and/or" used in the present description and the appended claims refers to and includes any and all possible combinations of one or more of the associated listed items.

Please refer to Figure 1 below, which is a schematic diagram of the structure of a device for building a classification model for identifying different stages of heart failure based on sound features provided by an embodiment of the present invention. The device includes a processing unit 101 and a model training unit 201;

The sample processing unit 101 is used to convert the collected voice analog signal into a voice digital signal, preprocess the voice digital signal, and extract features from the preprocessed voice digital signal to obtain multiple types of voice feature samples.

In this unit, sound is an analog signal, which needs to be converted into a digital signal before it can be used for computer processing. The sample processing unit includes: a first conversion unit for sampling, a second conversion unit for quantization, and a third conversion unit for encoding;

The first conversion unit is specifically used to convert the time-continuous voice analog signal into a time-discrete and amplitude-continuous signal sample according to a predetermined sampling period;

The sampling period is the time interval between two adjacent sampling points, and the sampling frequency is the reciprocal of the sampling period. For example, a sampling frequency of 8kHz means that 8000 samples are collected in 1 second. Therefore, the higher the sampling frequency, the higher the degree of sound restoration and the more realistic the sound.

The second conversion unit is specifically used to convert each signal sample with continuous amplitude values (analog quantity) into a discrete value (digital quantity), and use binary to represent it to obtain digital data;

The quantized signal samples are usually represented in binary. The sampling accuracy refers to the number of binary bits occupied by each signal sample, which can reflect the quality of the sound. For example, common CDs use a sampling depth of 16 bits, which can represent 65535 (2^16) different values. DVDs use a sampling depth of 24 bits, and most telephone devices use a sampling depth of 8 bits.

The third conversion unit is specifically used to convert the digital data into a binary code stream to obtain a voice digital signal.

Coding is the process of converting sampled and quantized digital data into a binary code stream for computer storage, processing and transmission. Pulse code modulation can achieve the highest fidelity level. The data transmission rate of sound can be calculated by sampling frequency and accuracy: data transmission rate (bps) = sampling frequency * accuracy * number of channels. At the same time, the data volume of the sound signal can be calculated: data volume (byte) = data transmission rate * duration / 8.

In this embodiment, in order to improve the quality of the voice digital signal and retain more voice information, the voice digital signal needs to be preprocessed before the voice digital signal is analyzed, that is, the sample processing unit also includes: a synchronization unit, an endpoint detection unit, an emphasis unit, a framing and windowing unit.

The synchronization unit is used to synchronize multiple voice digital signals to a uniform sampling rate by using a downsampling method;

Different speech acquisition devices may have different speech amplitudes due to their different sampling frequencies, durations, and locations, so the original speech needs to be downsampled before signal analysis. For example, if the sampling frequency of the original signal is 22050Hz, in order to reduce the computational complexity, improve the signal processing efficiency, and not lose the main components of the speech, according to the Nyquist sampling theorem, this embodiment can downsample the signal to 16000Hz.

The endpoint detection unit is used to perform endpoint detection on the voice digital signal after the unified sampling rate, and distinguish the voice area and the non-voice area;

Endpoint detection, also known as voice activity detection, aims to distinguish between speech and non-speech areas, that is, to accurately detect the start and end points of speech from noisy speech, and remove the silent and noisy parts to find the truly effective speech content. Common methods include the double threshold method based on short-time energy and short-time zero-crossing rate. Since the energy of voiced sounds is higher than that of unvoiced sounds, and the zero-crossing rate of unvoiced sounds is higher than that of silent parts, short-time energy can be used to detect the voiced parts and zero-crossing rate can be used to extract the unvoiced parts, so as to complete the endpoint detection of speech.

The emphasis unit is used to emphasize the high frequency part of the speech area to increase the high frequency resolution of the speech;

Since the average power spectrum of speech signal is affected by glottal excitation and oral-nasal radiation, the power spectrum decreases with the increase of frequency, and the energy of speech is mainly concentrated in the low-frequency part. Therefore, pre-emphasis processing is required to emphasize the high-frequency part of speech, remove the influence of lip radiation, increase the high-frequency resolution of speech, and make the signal spectrum from low frequency to high frequency The whole frequency band can be calculated with the same signal-to-noise ratio, which is convenient for spectrum analysis or vocal tract parameter analysis. Pre-emphasis is generally achieved by using a high-pass digital filter as the transfer function, that is, H(z)＝1-az ^-1 , where a is the pre-emphasis coefficient, ranging from 0.9<a<1.0, and generally a＝0.97.

The framing and windowing unit is used to perform framing and windowing on the emphasized speech area to obtain a plurality of speech signal segments.

Speech signals are non-stationary continuous analog signals with time-varying characteristics, but within a short time range (generally within 10 to 30 ms), their characteristics remain basically unchanged, that is, relatively stable. Therefore, speech signals have short-term stability, which means that any analysis and processing of speech signals must be "short-term analysis". The speech signal is segmented to analyze its characteristic parameters, where each segment is called a "frame" and the frame length is generally 10 to 30 ms. In order to achieve a smooth transition between frames and maintain continuity, overlapping segmentation can be used to divide the frames. The overlapping part of the previous frame and the next frame is called frame shift, and the ratio of frame shift to frame length is generally 0 to 0.5. In this embodiment, the frame length is 25 ms and the frame shift is 10 ms.

Because speech signals have short-term stability, in addition to framing the signal, it is also necessary to perform windowing, the purpose of which is to emphasize the speech waveform near the sampled sample and weaken the rest of the waveform. Commonly used window functions include rectangular window and Hamming window. The rectangular window has a higher spectral resolution, but the adjacent harmonic interference is more serious, the loss of high-frequency components leads to the loss of waveform details, while the Hamming window is the opposite.

In this embodiment, the sample processing unit further includes: an extraction unit and a merging unit;

The extraction unit is used to extract a multi-dimensional first speech feature sample using the openSMILE open source toolkit, and to extract a multi-dimensional second speech feature sample using Python;

The extraction unit is used to merge the multi-dimensional first speech feature samples and the multi-dimensional second speech feature samples to obtain multi-category speech feature samples.

It should be noted that the first speech feature sample and the second speech feature sample can be extracted according to actual needs. The multi-category speech feature samples extracted in this embodiment have a total of 100 dimensions. The first speech feature sample uses the eGeMAPS feature set, which is an extended feature set of GeMAPS. The feature set is an 88-dimensional manual feature extracted by the openSMILE open source toolkit, which includes 18 low-level descriptors (LLDs), and adds 5 spectral features (MFCC1-4 and spectral flux) and 2 frequency-related features (i.e., the bandwidth of the second resonance peak and the third resonance peak) on the basis of GeMAPS, including frequency, energy/amplitude-related features and spectral features. In addition, a 12-dimensional second speech feature sample was extracted using python.

In this embodiment, the first speech feature sample includes: pitch feature, frequency perturbation feature, formant feature, amplitude perturbation feature, loudness feature, harmonic-to-noise ratio feature, harmonic difference feature, α ratio feature, Hammarberg coefficient feature, spectral slope feature, Mel cepstral coefficient feature, spectral flow feature, loudness peak ratio feature, continuous sound area and silent area feature, equivalent sound level feature. The detailed information of the first speech feature sample is shown in Table 1.

Table 1

In Table 1, the pitch feature is the fundamental frequency of vocal cord vibration, which indicates the number of times the vocal cord vibrates per second, and the feature description is log F0, calculated on a semitone frequency scale, starting from 27.5 Hz; the frequency perturbation feature is the frequency change between adjacent cycles of the sound wave, and the feature description is the deviation within a single continuous fundamental tone cycle; the resonance peak feature, the feature description is the center frequency and bandwidth of the first, second, and third resonance peaks, and the energy ratio of the first three resonance peaks to the fundamental tone; the amplitude perturbation feature is the feature description: reflecting the amplitude change between adjacent cycles of the sound wave; the loudness feature is the size of the sound; the harmonic-to-noise ratio feature, the feature description is the proportion of periodically repeated harmonic components in the sound wave of the steady-state vowel; the harmonic difference feature, the feature description is the energy ratio of the first fundamental tone harmonic H1 to the second fundamental tone harmonic H2 or the energy ratio of the first fundamental tone harmonic H1 to the third resonance peak H3; the α ratio feature, the feature description is 50-1000H The energy sum of z divided by the energy sum of 1-5kHz; Hammarberg coefficient feature, feature description: that is, the strongest energy peak of 0-2kHz divided by the strongest energy peak of 2-5kHz; spectral slope feature, that is, feature description: the linear regression slope of the logarithmic power spectrum in the range of 0-500Hz and 500-1500Hz; Mel cepstral coefficient feature, that is, feature description: Mel cepstral coefficients 1-4; spectral flow feature, that is, feature description: the spectral difference between two adjacent frames; loudness peak ratio feature, that is, feature description: the number of loudness peaks per second; continuous sound area and silent area feature, that is, feature description: the duration of continuous voiced sound (F0>0) and the duration of unvoiced sound (F0＝0); equivalent sound level feature, that is, the equivalent sound level refers to the average value of the A sound level in a certain period of time according to energy.

In this embodiment, the second speech feature sample includes: glottal noise excitation ratio feature, vocal cord excitation ratio feature, cycle density entropy feature, detrending fluctuation analysis feature, sample entropy feature, and multi-scale entropy feature. The detailed information of the second speech feature sample is shown in Table 2.

Table 2

In Table 2, the glottal noise excitation ratio feature (GNE) and the vocal fold vibration excitation ratio feature (VFER) are used to quantify the energy proportion of the normal speech signal in which each frequency band is simultaneously excited by the glottal pulse and the noise signal in which each frequency band is disorderly excited by the chaotic noise (usually caused by incomplete closure of the glottis). For example, to calculate GNE parameters, for the original speech signal with a sampling frequency of 44.1kHz, first downsample the signal to 10KHz; then use the inverse filtering method to find the opening and closing time points of each glottis, so as to find out its opening and closing time series; then for each time series, use a filter with a bandwidth of 500Hz to filter out the frequency bands of 0-500Hz, 500-1000Hz, 1000-1500Hz, etc. up to 11.5KHz; for each frequency band, use the first five frequency bands of the low frequency band (1Hz-2.5KHz) as the signal, and the remaining high frequency band (2.5KHz-11.5KHz) as the noise, and calculate the SEO and TKEO energy values of the signal and noise respectively; finally, the signal-to-noise ratio SNR and the noise-to-signal ratio NSR are calculated according to the calculated energy values. The calculation of VFER parameters is roughly the same as that of GNE, except that the process of downsampling the signal in the GNE process is removed, and the DYSPA algorithm is used in the second step of GNE instead of the inverse filtering method to obtain the glottal opening and closing sequence. In other words, GNE and VFER first use inverse filtering (for GNE) or DYSPA (for VFER) to detect the glottal pulse within a given time window; then the original sound is divided into two parts: noise above 2.5kHz and energy signal below 2.5kHz; then, combined with the concepts of SEO and TKEO, the energy values of different frequency bands of the signal are calculated to obtain the signal-to-noise ratio measured by the empirical mode decomposition excitation ratio (EMDER-ER). After calculating the values of each time window, the average and standard values of GNE and VFER are calculated. The following parameters can be calculated: GNE_SEO_SNR, GNE_TKEO_SNR, GNE_mean, GNE_std, VFER_SEO_SNR, VFER_TKEO_SNR, VFER_mean, VFER_std.

Cyclic period density entropy (RPDE) is a method used in the fields of dynamical systems, random processes and time series analysis to determine the periodicity or repeatability of a signal. Its value is between 0 and 1. For quasi-periodic signals, its value is 0, while the value of uniform white noise is close to 1. Its calculation method is as follows:

The first step is to project the time series _Xn = [ _xn , _xn+r , _xn+2r , …, _xn+(M-1)r ] into a phase space according to Taken embedding theory; here M is the embedding dimension, T is the embedding delay, and these parameters are obtained through the parameter optimization algorithm. The second step is to draw an M-dimensional region with a radius of each point Xn in the phase space, and then record the time difference of each time the time series arrives at this region and leaves this region, draw this time difference into a histogram, and finally normalize it to obtain the regression density function P(T).

By formula

Where T _max is the maximum delay embedded in the phase space, and the value of the cycle density entropy can be obtained.

Detrended fluctuation analysis (DFA) is a scaling index calculation method used to eliminate the influence of trend terms in time series on fluctuation analysis. It is used to analyze the long-range correlation of speech signals, that is, to determine whether the noise terms in the time series have positive or negative autocorrelation. One of its advantages is that it can effectively filter out trend components of various orders in the sequence and detect long-range correlations containing noise and superimposed with polynomial trend signals. It is suitable for long-range power-law correlation analysis of non-stationary time series.

The implementation method is as follows:

1. First, for the sequence x(t), calculate its cumulative deviation y(t)

为序列x(t)的平均值。Where:

is the mean value of the sequence x(t).

Here, the average value of the time series is first filtered out. Since cyclic or fluctuating components may exist in general time series, a time series may have random components, and filtering out these components of the series will be of great help.

2. Reconstruct the sequence and divide Y(t) into m non-overlapping intervals of equal length with length s, where m = [n/s] (rounded); since the sequence length is not always an integer multiple of the increment s, sometimes a small amount of data information at the end of the sequence is not used. Therefore, the same operation is performed on the reverse order of the sequence to obtain 2N intervals of equal length.

3. For each interval v, use the least squares method to perform a first-order linear fit on the S data contained in each interval.

4. Calculate the mean square error after filtering out the trend in each interval (here the sequential and reverse order are calculated separately):

5. Take the mean and square root of all intervals of equal length to calculate the DFA fluctuation function:

6. If the runoff time series {x(t)} is long-range power-law related, then F(s) and s satisfy the following power-law relationship:

ln(F(s))～hln(s)

Taking the logarithm of both sides of the above equation, we get:

F(s)～s ^h

In the scatter plot in double logarithmic coordinates (ln(s), ln(F(s))), the data points are fitted using the least squares method, and the slope of the straight line is the Hurst exponent.

The relationship between Hurst index and correlation

(1) When 0.5<h<1, it means that the time series has long-range correlation and presents a state of increasing trend. That is, if there is an increasing (decreasing) trend in a certain time period, there will also be an increasing (decreasing) trend in the next time period. The closer h is to 1, the stronger the correlation.

(2) When h = 0.5, it means that the time series is uncorrelated and is an independent random process, that is, the current state will not affect the future state.

(3) When 0<h<0.5, it means that the runoff time series has only negative correlation and presents an anti-persistent state, that is, if the time series has an increasing (decreasing) trend in a certain time period, it will have a decreasing (increasing) trend in the next time period.

和香农熵，可以计算出每个IMF的SEO、TKEO和香农熵。在计算信噪比的时候，将前四个IMF作为噪声信号，由公式

(u为每个IMF的SEO、TKEO和香农熵的值,D为分解得到的IMF的个数)可以得到有关信噪比的参数。在计算噪信比时候，先对每个IMF取对数，将前两个IMF作为噪声信号，然后再计算每个IMF的SEO、TKEO和香农熵的值，由公式

(u为每个IMF的SEO、TKEO和香农熵的值，D为分解得到的IMF的个数)可以得到有关噪信比的参数。Empirical Decomposition Mode Ratio (EMD-ER): For the original speech signal with a sampling frequency of 44.1kHz, it can be decomposed into a finite number of intrinsic mode functions (IMFs). The decomposed IMF components contain local characteristic signals of the original signal at different time scales. The initially decomposed intrinsic mode function is a high-frequency noise signal, and the subsequently decomposed intrinsic mode function is an actual useful signal. According to the energy operator formula

And Shannon entropy, we can calculate the SEO, TKEO and Shannon entropy of each IMF. When calculating the signal-to-noise ratio, the first four IMFs are used as noise signals, and the formula

(u is the SEO, TKEO and Shannon entropy value of each IMF, D is the number of IMFs obtained by decomposition) can be used to obtain the parameters related to the signal-to-noise ratio. When calculating the signal-to-noise ratio, first take the logarithm of each IMF, take the first two IMFs as noise signals, and then calculate the SEO, TKEO and Shannon entropy values of each IMF. The formula is

(u is the value of SEO, TKEO and Shannon entropy of each IMF, and D is the number of IMFs obtained by decomposition) parameters related to the noise-to-signal ratio can be obtained.

The sample entropy feature (SampEn) is an improved method for measuring the complexity of time series based on approximate entropy;

Multi-scale entropy feature (MSEn) expands the sample entropy to multiple time scales and calculates the complexity of signals at different time scales.

Glottis quotient (GQ) measures the stability of vocal cord vibration by calculating the mean and standard deviation of speech signals. First, the DYSPA algorithm is used to find the opening and closing points of the glottis, and a speech signal is divided into several glottis opening segments and glottis closing segments. Then, the mean and standard deviation of the glottis opening segments and glottis closing segments are calculated respectively.

MFCC is the cepstral parameter extracted from the Mel scale frequency domain. The Mel scale describes the nonlinear characteristics of the human ear frequency. The 39-dimensional MFCC includes a logarithmic energy and 12 cepstral parameters, difference operation and difference difference operation from left to right.

The model training unit 201 is used to construct a classification model for identifying heart failure stages, and use the multi-category speech feature samples to train and optimize the classification model to obtain the optimal classification model, wherein the classification model for identifying heart failure stage A and heart failure stage B is an AdaBoost classification model based on the original variables; the classification model for identifying heart failure stage B and stage C is an AdaBoost classification model based on Lasso dimensionality reduction; the classification model for identifying heart failure stage A, stage B and stage C is an AdaBoost classification model based on Lasso dimensionality reduction.

In one embodiment, it is necessary to combine the binary classification model and the dimensionality reduction method to optimize the classifier algorithm. In this embodiment, six classification models and two dimensionality reduction methods are selected, namely, support vector machine classification model (SVM), decision tree classification model (DT), adaptive boost classification model (Ada Boost), least absolute shrinkage and selection operator classification model (LASSO), ridge regression classification model (Ridge regression), elastic network classification model (Elastic Net) and principal component analysis (PCA) dimensionality reduction, LASSO dimensionality reduction.

The final model is shown in Table 3, which includes three models: ① classification model based on original variables; ② classification model based on principal components after PCA dimensionality reduction; ③ classification model based on LASSO feature selection and feature variables.

Table 3

The device of the embodiment of the present invention further includes: an evaluation unit, configured to perform model evaluation on the classification model according to a holdout method.

When building a model, we should try to avoid "overfitting", that is, the classifier regards some characteristics of the training sample itself as the general properties that all potential samples will have, which leads to a decrease in generalization performance, which is manifested as the final model has good results on the training set and poor results on the test set. We should also avoid "underfitting", that is, the general properties of the training samples have not been learned well, and the performance on both the training set and the test set is poor. "Overfitting" cannot be avoided, but can only be alleviated, while "underfitting" can be overcome by increasing the number of features, increasing the complexity of the model, and reducing the regularization coefficient. Ultimately, we hope that the model has a good fit for the training data set (i.e., training error), and at the same time, it can have a good fit result (i.e., generalization ability) for the unknown data set (i.e., test set).

Model building usually requires dividing sample data into a training set and a test set, and the two are mutually exclusive sets. Model evaluation requires a test set to test the learner's ability to discriminate new samples, and the "test error" on the test set is approximated as the generalization error. Commonly used methods include holdout method, cross-validation method and bootstrap method.

Holdout method: Directly divide the data set into two mutually exclusive sets, one set as the training set and the other as the test set. The division of the training set and the test set should maintain the consistency of the data distribution as much as possible to avoid the impact of additional deviations introduced by the data division process on the final result. Since the estimation results obtained by using the holdout method once are often unreliable, it is generally necessary to use several random divisions and repeat the experimental evaluation and take the average value as the evaluation result of the holdout method. Generally, 2/3 to 4/5 of the samples are used for training, and the rest are used for testing. The hold-out method is adopted in this embodiment, which is one of the methods of the hold-out method.

The device of the embodiment of the present invention also includes: a performance measurement unit, which is used to perform performance measurement on the classification model according to predetermined indicators, wherein the predetermined indicators include error rate and accuracy, precision rate and recall rate, F1 value, specificity, sensitivity, ROC curve and AUC, and unweighted average recall rate.

To evaluate the generalization performance of the classifier, we need an evaluation standard to measure the generalization ability of the model, that is, performance measurement. For binary classification problems, a "confusion matrix" can be formed based on the combination of the true category and the classifier's predicted category, which includes four cases: true positive (True Positive, TP), true negative (True Negative, TN), false positive (False Positive, FP), and false negative (False Negative, FN). The confusion matrix is shown in Table 4.

Table 4

The predetermined indicators are:

Accuracy (ACC): Also known as precision, it refers to the ratio of all correctly classified samples to the total number of samples. It is applicable to both binary and multi-classification tasks and can be used to determine the overall accuracy. However, in the case of unbalanced samples, the accuracy will be invalid and cannot be used as an indicator to measure the results. Other indicators are needed to supplement it.

ACC＝(TP+TN)/(TP+FN+FP+TN)

Error rate (ERR): refers to the ratio of all misclassified samples to the total number of samples, ERR = 1-ACC.

ERR＝(FN+FP)/(TP+FN+FP+TN)

Precision (P): Also known as accuracy, it mainly refers to the prediction results and refers to the proportion of all samples predicted to be positive to the actual positive samples.

P＝TP/(TP+FP)

Recall rate (R): also known as recall rate or sensitivity (SEN), mainly for the original sample, refers to the proportion of samples predicted to be positive among the samples that are actually positive.

R＝TP/(TP+FN)

Specificity: (SPE): refers to the proportion of samples predicted to be negative among those that are actually negative.

SPE＝TN/(FP+TN)

F1 Score: It is the harmonic mean of precision and recall, ranging from 0 to 1. It is often used in statistics to measure the accuracy of binary classification (or multi-task binary classification) models.

F1＝2*P*R/(P+R)

ROC curve and AUC: The full name of ROC curve is "receiver operating characteristic curve". The ordinate of the ROC curve is the true positive rate (i.e. sensitivity), and the abscissa is the false positive rate (1-sensitivity). The coordinate points are obtained and connected at different thresholds. The closer the ROC curve is to the diagonal line, the lower the accuracy of the model. Assuming that the ROC curve of classifier A can "enclose" classifier B, it can be said that classifier A has better classification performance, but when the ROC curves of the two classifiers intersect, it is difficult to judge the performance of the two. At this time, the area under the ROC curve can be used for measurement, that is, AUC (Area Under ROC Curve). Since the ROC curve is generally above the straight line y=x, the value is generally 0.5~1. The larger the AUC value (area), the better the classifier performance.

Unweighted Average Recall (UAR): If the classifier labels are unevenly distributed, traditional evaluation indicators (such as ACC, P, R, F1, etc.) will lead to overly optimistic results for the class with a large number of samples. UAR can be used as a performance metric to avoid overfitting of the proposed classifier method to a certain category.

Specific embodiments

Patient data: From April 2021 to December 2022, a total of 101 patients were included. According to the stage of heart failure, they were divided into group A (stage A, n=35), group B (stage B, n=26), and group C (stage C, n=40). At the same time, 29 volunteers without heart failure were included in group N (n=29). There were no significant differences in gender, body mass index (BMI), systolic blood pressure, hemoglobin value, creatinine, low-density lipoprotein cholesterol (LDL-C), history of coronary heart disease, history of hypertension, history of diabetes, history of smoking, history of drinking, and history of dyslipidemia among the three groups with different stages of heart failure. The differences in age, creatinine, troponin, N-type brain natriuretic peptide precursor, left ventricular ejection fraction, and left ventricular internal diameter among the three groups with different stages of heart failure were statistically significant (P<0.05). The specific clinical data are shown in Table 5:

Table 5

Phase A (n=35) Phase B (n=26) Phase C (n=40) P Gender (%, M) 28(68%) 28(97%) 32(80%) 0.014 age 46±12 51±13 57±12 <0.001 BMI 25.7±3.5 26.9±3.8 25.3±4.8 0.246 Systolic blood pressure (mmHg) 139±20 141±24 121±18 <0.001 Hemoglobin 145±15 146±11 141±28 0.518 Troponin (ng/ml) 0.005(0.003) 0.011(0.013) 0.031(0.045) <0.001 NT-ProBNP (pg/ml) 31.0(45.4) 59.0(168.4) 1391.0(1754.0) <0.001 LDL-c (mmol/L) 2.90±1.09 2.65±1.12 2.54±1.10 0.326 Left ventricular ejection fraction (%) 67±5 66±5 41±10 <0.001 Left ventricular internal diameter (diastolic mm) 45±3 47±4 61±10 <0.001 History of coronary heart disease (%) 15(37%) 14(48%) 23(57%) 0.168 History of hypertension (%) 24(59%) 24(83%) 19(48%) 0.011 History of diabetes (%) 9(22%) 8(28%) 11(28%) 0.809 Smoking history (%) 7(17%) 11(38%) 16(40%) 0.053 History of dyslipidemia (%) 13(32%) 12(41%) 13(33%) 0.664

In this embodiment, a total of 130 cases that meet the inclusion criteria are included, with a total of 4055 voice samples and a total effective duration of 2.216 hours. Among them, 63 cases (%) were aged 30-50 years old, and 67 cases (%) were aged >50. The heart failure grouping is as follows: Group A 35 cases (%), including 1085 voice samples, effective duration 0.574 hours; Group B 26 cases (%), including 849 voice samples, effective duration 0.462 hours; Group C 40 cases (%), including 1231 voice samples, effective duration 0.715 hours; Group N 29 cases (%), including 890 voice samples, effective duration 0.465 hours; Control group 18 cases (%), including 890 voice samples, effective duration 0.465 hours.

The speech features in this embodiment use the eGeMAPS feature set, which is an extended feature set of GeMAPS. This feature set is an 88-dimensional manual feature extracted by the openSMILE open source toolkit, which includes 18 low-level descriptors (LLDs), and adds 5 spectral features (MFCC1-4 and spectral flux) and 2 frequency-related features (the bandwidth of the second resonance peak and the third resonance peak) on the basis of GeMAPS, including frequency, energy/amplitude-related features and spectral features. In addition, another 12-dimensional features are extracted using python: GNE_SEO_SNR, GNE_TKEO_SNR, GNE_mean, GNE_std, VFER_SEO_SNR, VFER_TKEO_SNR, VFER_mean, VFER_std, cycle density entropy RPDE, detrending fluctuation analysis DFA, sample entropy SampEn, multi-scale entropy MSEn. Therefore, the feature parameters used in this study have a total of 100 dimensions. By analyzing the 100-dimensional speech features of patients with different heart failure stages, it was found that the speech features reflecting voice roughness and breath were significantly affected by different heart failure stages.

As for the roughness of the voice, it reflects the ability to control the glottis, vocal cords and the degree of hoarseness. The main indicators include Jitter, Shimmer, Harmonic difference, HNR, Alpha Ratio, etc.

Jitter represents the deviation within a single continuous fundamental period, reflecting the sound quality characteristics of the voice rhythm. The statistics are mean and standard deviation. There are two features in total, and there are significant differences in different heart failure stages. The experimental results are shown in Figure 2.

Shimmer represents the difference between the peak amplitudes of adjacent fundamental cycles and also reflects the sound quality characteristics of the voice rhythm. The statistics are mean and standard deviation. There are two features in total, and both types of heart failure have significant effects. The experimental results are shown in Figure 3.

Harmonic difference means H1-H2: the energy ratio of the first fundamental harmonic H1 to the second fundamental harmonic H2. The statistics are mean and standard deviation. There are 2 features in total, and the heart failure type has a significant impact. H1-A3: +The energy ratio of the first fundamental harmonic H1 to the third resonance peak A3. The statistics are mean and standard deviation. There are 2 features in total, and the heart failure type has a significant impact. The experimental results are shown in Figure 4.

HNR stands for harmonic noise ratio, which is the proportion of periodically repeated harmonic components in the sound waves of steady-state vowels. The statistics are mean and standard deviation; there are two features in total, and the type of heart failure has a significant impact. The experimental results are shown in Figure 5.

Alpha Ratio represents the power sum of 50-1000Hz divided by the power sum of 1-5kHz. The statistics are the mean and standard deviation of the voiced area and the mean of the unvoiced area. There are three features in total, and the heart failure type has a significant impact. The experimental results are shown in Figure 6.

As for the vocal breathiness, it reflects the pauses and intensity of the voice, including indicators such as Loudness, voiced/unvoiced duration, Hammarberg Index, Spectral Slope, etc. There are significant differences in the vocal breathiness in different heart failure stages.

Voiced/unvoiced duration indicates the duration of continuous voiced sounds (F0>0), and the statistics are the average length and standard deviation, with a total of 2 features. Unvoiced sound (F0＝0) duration, the statistics are the average length and standard deviation, with a total of 2 features. The number of voiced sound areas per second, with a total of 1 feature. Heart failure type has a significant impact on all 5 features. The experimental results are shown in Figure 7.

Loudness means that the statistics are the mean, standard deviation, 20/50/80 percentiles, 20-80 percentile range, and the mean and standard deviation of the slope of the rising/falling speech signal; there are 10 features in total, and different heart failure stages have significant effects. The experimental results are shown in Figure 8.

The Hammarberg Index represents the strongest energy peak between 0 and 2 kHz divided by the strongest energy peak between 2 and 5 kHz.

The statistics are the mean and standard deviation of the voiced area and the mean of the unvoiced area, a total of 3 features, and the heart failure type has a significant impact. The experimental results are shown in Figure 9.

Spectral Slope represents the linear regression slope (attenuation rate) of the logarithmic power spectrum in the range of 0-500Hz and 500-1500Hz, and the slope of the oblique spectrum envelope. The larger the slope, the greater the attenuation of the signal outside the frequency division slope. The statistics are the mean and standard deviation of the voiced area and the mean of the unvoiced area. There are 6 features in total, and the type of heart failure has a significant impact on 5 features. The experimental results are shown in Figure 10.

The cross-correlation coefficient of each glottal cycle or the ratio of the energy above 2.5KHz to that below 2.5KHz. There are 4 features in total, and the type of heart failure has a significant impact. The experimental results are shown in Figure 11.

For nonlinear analysis, nonlinear parameters are more suitable for describing the intrinsic characteristics of acoustic signals with poor periodicity, and most heart failure voices often show poor periodicity. Entropy reflects the disordered characteristics of voice information distributed in the frequency domain, and sample entropy can reflect the complexity of heart sound signals in the time domain. This study first extracted nonlinear parameters such as recurrence period density entropy (RPDE), detrended fluctuation analysis (DFA), sample entropy (Sample Entropy) and scale entropy to describe the characteristics of periodic, non-periodic and chaotic voice signals.

Cycle-by-cycle density entropy RPDE, detrending fluctuation analysis DFA, sample entropy SampEn, and multi-scale entropy MSEn all reflect the roughness of the sound. There are 4 features in total, and the heart failure type has a significant impact. The experimental results are shown in Figure 12.

For acoustic feature parameters based on cepstrum, Mel cepstrum coefficients are cepstrum parameters extracted from the Mel scale frequency domain. The Mel scale describes the nonlinear characteristics of human ear frequency. Acoustic feature parameters based on cepstrum analysis can effectively avoid the inaccurate analysis caused by the irregularity of the fundamental frequency.

Mel cepstral coefficients 1-4. Statistics are the mean and standard deviation of the whole and voiced segments; there are 16 features in total, and the type of heart failure has a significant impact on 15 features. The experimental results are shown in Figure 13.

It should be noted that the horizontal axis in Figures 2-13 represents the grouping in the heart failure staging study, and the vertical axis represents the corresponding speech feature parameters. For example, in Figure 2, two speech feature parameters, jitterLocal_sma3nz_ameam and jitterLocal_sma3nz_stddevNorm, were extracted for Jitter, and the number of corresponding speech feature parameters extracted is consistent with that in Table 1.

In this embodiment, based on the original 100-dimensional eGeMAPs features of the voice of patients with heart failure at different stages, we further used a binary classification experiment to reduce the dimension through the PCA method and the LASSO method, and compared 6 different classifiers, including support vector machine (SVM), decision tree (DT), adaptive boosting (Adaptive Boosting, Ada Boost), least absolute shrinkage and selection operator (Least Absolute Shrinkage and Selection Operator, LASSO), ridge regression (Ridgeregression), elastic network (Elastic Net) and other methods, to observe the performance of different classifiers in identifying patients with heart failure at different stages and find the optimal classification model.

1. A, B binary classification experiment (leave one out method)

The heart failure stage A and stage B population was used as the dependent variable for binary classification recognition modeling. Based on the classification model with the original variables substituted, the Ada Boost classifier model was the best in the original 100-dimensional speech feature classification model, with an accuracy of 0.869, a precision of 0.846, a recall of 0.846, and an F1 score of 0.846. After PCA dimensionality reduction, the Ada Boost classification model was the best, and the accuracy of the model training using the training set was 0.738; after LASSO dimensionality reduction, the Ada Boost classification model was the best, and the accuracy of the model training using the training set was 0.770, which was lower than the binary classification recognition model based on the original features, which may be related to the feature loss after dimensionality reduction. The optimal model, the original AdaBoost, has a sample-level ROC curve as shown in Figure 14 (AUC = 0.793). The results show that speech features can identify people in stage A and stage B of heart failure, suggesting that speech features can be used to preliminarily screen patients in the risk period of heart failure and patients in the early stage of heart failure with target organ damage. The binary classification accuracy (mean, standard deviation) based on the original 100-dimensional eGeMAPs features is shown in Table 6.

Table 6

A binary classification model was built with the heart failure stage A and stage B population as the dependent variable. The classification model was based on the original variable substitution.

The AdaBoost confusion matrix is shown in Table 7 and Table 8:

Table 7

Table 8

Based on PCA dimensionality reduction, the results of the classification model using principal components are shown in Table 9:

Table 9

The AdaBoost confusion matrix is shown in Table 10 and Table 11:

Table 10

Table 11

The results of the classification model based on LASSO dimensionality reduction and substituting the feature variables into the classification model are shown in Table 12:

Table 12

The AdaBoost confusion matrix is shown in Table 13 and Table 14:

Table 13

Table 14

After dimensionality reduction by LASSO, the classification model of Ada Boost is the best. The accuracy of model training using the training set is 0.770, which is lower than that of the binary classification recognition model based on the original features. It is considered that this may be related to the feature loss after dimensionality reduction. The importance of the original 100-dimensional AdaBoost features (where the importance of 2 dimensions is 0) is shown in Table 15.

Table 15

A, B binary classification experiment (leave one out method) results summary

The performance evaluation indicators of the original classification model in the test set are shown in Table 16:

Table 16

The performance evaluation indicators of "PCA dimensionality reduction + classification model" in the test set are shown in Table 17:

Table 17

The performance evaluation indicators of "LASSO feature selection + classification model" in the test set are shown in Table 18:

Table 18

The sample-level ROC curve (AUC=0.793) of the optimal model (original AdaBoost) is shown in FIG14 , where the horizontal and vertical axes represent the false positive rate and the true positive rate, respectively, and the Receiver Operating Characteristic is the receiver operating curve.

2. B, C binary classification results (leave one out method)

The heart failure stage B and C population was used as the dependent variable for binary classification modeling. Based on the classification model with the original variables substituted, the Ada Boost classifier model was the best in the original 100-dimensional speech feature classification model, with an accuracy of 0.788, a precision of 0.771, a recall of 0.925, and an F1 score of 0.841. Through PCA dimensionality reduction, the classification models of Ada Boost and SVM were similar. The accuracy of model training using the training set was 0.773; the accuracy of the Elastic classification model training was 0.742, both of which were lower than the binary classification recognition model based on the original features. After LASSO dimensionality reduction, the Ada Boost classification model was the best. The accuracy of model training using the training set was 0.803, which was higher than the accuracy of the binary classification recognition model based on the original features. As shown in Figure 15, the optimal model, AdaBoost with Lasso dimensionality reduction, had a sample-level ROC curve area under the curve of 0.819. The results showed that speech features can identify people in stage B and C of heart failure, suggesting that speech features can be used to identify patients in the early stages of heart failure with target organ damage and patients who have experienced symptomatic heart failure.

The binary classification accuracy (mean, standard deviation) based on the original 100-dimensional eGeMAPs features is shown in Table 19:

Table 19

The AdaBoost confusion matrix is shown in Table 20 and Table 21:

Table 20

Table 21

Based on PCA dimensionality reduction, the classification model using principal components is shown in Table 22:

Table 22

The AdaBoost confusion matrix is shown in Table 23 and Table 24:

Table 23

Table 24

The SVM confusion matrix is shown in Table 25 and Table 26:

Table 25

Table 26

Based on LASSO dimensionality reduction, the classification model in which the feature variables are substituted is shown in Table 27:

Table 27

The AdaBoost confusion matrix is shown in Table 28 and Table 29:

Table 28

Table 29

The features with LASSO regularization coefficients that are not 0 are 66 in total, and 34 of them are 0, as shown in Table 30:

Table 30

B. C Binary Classification Experiment (Leave One Out) Results Summary

The performance evaluation indicators of the original classification model in the test set are shown in Table 31:

Table 31

The performance evaluation indicators of "PCA dimensionality reduction + classification model" in the test set are shown in Table 32:

Table 32

The performance evaluation indicators of "LASSO feature selection + classification model" in the test set are shown in Table 33:

Table 33

The sample-level ROC curve (AUC=0.793) of the optimal model (original AdaBoost) is shown in FIG15 , where the horizontal and vertical axes represent the false positive rate and the true positive rate, respectively, and the Receiver Operating Characteristic is the receiver operating curve.

3. AB, C binary classification experiment (leave one out method)

The patients with heart failure A and B (AB) and C stage were used as dependent variables for binary classification recognition modeling. Based on the classification model with the original variables substituted, the Ada Boost classifier model was the best in the original 100-dimensional speech feature classification model, with an accuracy of 0.802, a precision of 0.857, a recall of 0.600, and an F1 score of 0.706. After PCA dimensionality reduction, the SVM classification model was the best, but the accuracy of the model training using the training set dropped to 0.723; after LASSO dimensionality reduction, the Ada Boost classification model was the best, with an accuracy of 0.812 using the training set, which was higher than the accuracy of the binary classification recognition model based on the original features. As shown in Figure 16, the best model, AdaBoost with Lasso dimensionality reduction, had a sample-level ROC curve area under the curve of 0.731. The results show that speech features can distinguish patients with heart failure risk (stage A) and pre-heart failure (stage B) with target organ damage from patients with stage C who have experienced symptomatic heart failure, that is, speech features can help identify patients who have experienced symptomatic heart failure. The binary classification accuracy (mean, standard deviation) based on the original 100-dimensional eGeMAPs features is shown in Table 34.

Table 34

The AdaBoost confusion matrix is shown in Table 35 and Table 36:

Table 35

Table 36

Based on PCA dimensionality reduction, the classification model using principal components is shown in Table 37:

Table 37

The SVM confusion matrix is shown in Table 38 and Table 39:

Table 38

Table 39

The results of dimensionality reduction based on LASSO are shown in Table 40:

Table 40

The AdaBoost confusion matrix is shown in Table 41 and Table 42:

Table 41

Table 42

The features with LASSO regularization coefficients that are not 0 are 56 in total, and 44 of them are 0, as shown in Table 43:

Table 43

AB and C classification results (leave one out)

The performance evaluation indicators of the original classification model in the test set are shown in Table 44:

Table 44

The performance evaluation indicators of the "PCA dimensionality reduction + classification model" in the test set are shown in Table 45:

Table 45

The performance evaluation indicators of "LASSO feature selection + classification model" in the test set are shown in Table 46:

Table 46

The sample-level ROC curve (AUC=0.793) of the optimal model (original AdaBoost) is shown in FIG16 , where the horizontal and vertical axes represent the false positive rate and the true positive rate, respectively, and the Receiver Operating Characteristic is the receiver operating curve.

The above experiments have proved that (1) the voice characteristics of patients with different heart failure stages are different. Based on the eGeMAPS feature set and a total of 100-dimensional features extracted by python, the main indicators reflecting the roughness of the voice of patients with different heart failure stages, such as Jitter, Shimmer, Harmonic difference, HNR, Alpha Ratio, etc., are different between different stages of heart failure. (2) The main indicators reflecting the breathiness of the voice, including Loudness, voiced/unvoiced duration, Hammarberg Index, Spectral Slope, etc., also have significant differences between different heart failure stages. Among these sound indicators, the basic characteristics of the voice, including frequency, energy/amplitude related characteristics, nonlinearity and other characteristics, are different in different heart failure stages. (3) The contribution of speech features in different stages is not exactly the same. (4) Based on the original 100-dimensional eGeMAPs features of the voices of patients with heart failure at different stages, a classification model was constructed using a binary classification method. The performance of the original variables, PCA dimensionality reduction, LASSO dimensionality reduction, and different classifiers in identifying patients with heart failure at different stages was compared, and the classification model was optimized. The optimal model for identifying patients with heart failure at stage A and stage B was the AdaBoost classification method based on the original variables, with an ROC curve AUC of 0.793; the optimal model for identifying patients with heart failure at stage B and stage C was the AdaBoost classification method based on Lasso dimensionality reduction, with an ROC curve AUC of 0.819; the optimal model for distinguishing between stage AB and stage C was the AdaBoost classification method based on Lasso dimensionality reduction, with an ROC curve AUC of 0.731.

The present invention also provides a computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the following method when executing the computer program: converting the collected voice analog signal into a voice digital signal, preprocessing the voice digital signal, and extracting features from the preprocessed voice digital signal to obtain multiple categories of voice feature samples; constructing a classification model for identifying heart failure stages, and training and optimizing the classification model using the multiple categories of voice feature samples to obtain an optimal classification model, wherein the classification model for identifying heart failure stage A and heart failure stage B is an AdaBoost classification model based on original variables; the classification model for identifying heart failure stage B and stage C is an AdaBoost classification model based on Lasso dimensionality reduction; and the classification model for identifying heart failure stages A, B, and C is an AdaBoost classification model based on Lasso dimensionality reduction.

The present invention also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the following method is implemented: converting the collected voice analog signal into a voice digital signal, preprocessing the voice digital signal, and extracting features from the preprocessed voice digital signal to obtain multiple categories of voice feature samples; constructing a classification model for identifying heart failure stages, training and optimizing the classification model using the multiple categories of voice feature samples to obtain an optimal classification model, wherein the classification model for identifying heart failure stage A and heart failure stage B is an AdaBoost classification model based on original variables; the classification model for identifying heart failure stage B and stage C is an AdaBoost classification model based on Lasso dimensionality reduction; and the classification model for identifying heart failure stages A, B and C is an AdaBoost classification model based on Lasso dimensionality reduction.

The various embodiments in the specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same and similar parts between the various embodiments can be referred to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part description. It should be pointed out that for ordinary technicians in this technical field, without departing from the principle of the present invention, several improvements and modifications can be made to the present invention, and these improvements and modifications also fall within the scope of protection of the claims of the present invention.

It should also be noted that, in this specification, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "comprises", "comprising" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the statement "comprising a ..." does not exclude the presence of other identical elements in the process, method, article or device including the element.

Claims (10)

1. An apparatus for constructing a classification model based on acoustic features that identifies different stages of heart failure, comprising:

the sample processing unit is used for converting the collected voice analog signals into voice digital signals, preprocessing the voice digital signals, and extracting the characteristics of the preprocessed voice digital signals to obtain multiple types of voice characteristic samples;

the model training unit is used for constructing a classification model for identifying heart failure stage, training and optimizing the classification model by utilizing the multi-class voice feature samples to obtain an optimal classification model, wherein the classification model for identifying heart failure stage A and heart failure stage B is an AdaBoost classification model based on an original variable; the classification model used for identifying the B phase and the C phase of heart failure is an AdaBoost classification model based on Lasso dimension reduction; the classification model used for identifying heart failure stage A and B and stage C is an AdaBoost classification model based on Lasso dimension reduction.

2. The apparatus for constructing a classification model for identifying different stages of heart failure based on acoustic features according to claim 1, wherein the sample processing unit comprises:

a first conversion unit for converting a time-continuous speech analog signal into time-discrete, amplitude-continuous signal samples at a predetermined sampling period;

The second conversion unit is used for converting each signal sample with continuous value in amplitude into discrete value and representing the discrete value by binary system to obtain digital data;

and the third conversion unit is used for converting the digital data into a binary code stream to obtain a voice digital signal.

3. The apparatus for constructing a classification model for identifying different stages of heart failure based on acoustic features of claim 1, wherein the sample processing unit further comprises:

a synchronizing unit for synchronizing the plurality of voice digital signals to a uniform sampling rate by adopting a downsampling method;

the end point detection unit is used for carrying out end point detection on the voice digital signal after the unified sampling rate and distinguishing a voice area and a non-voice area;

the emphasis unit is used for emphasizing the high-frequency part of the voice area and increasing the high-frequency resolution of the voice;

and the framing and windowing unit is used for framing and windowing the emphasized voice region to obtain a plurality of voice signal segments.

4. The apparatus for constructing a classification model for identifying different stages of heart failure based on acoustic features of claim 1, wherein the sample processing unit further comprises:

An extraction unit for extracting a multi-dimensional first speech feature sample using an openSMILE open source toolkit and extracting a multi-dimensional second speech feature sample using python;

and the merging unit is used for merging the multi-dimensional first voice characteristic sample and the multi-dimensional second voice characteristic sample to obtain multi-class voice characteristic samples.

5. The apparatus for constructing a classification model identifying different stages of heart failure based on acoustic features of claim 4, wherein the first speech feature sample comprises: pitch characteristics, frequency perturbation characteristics, formant characteristics, amplitude perturbation characteristics, loudness characteristics, harmonic to noise characteristics, harmonic difference characteristics, alpha ratio characteristics, hammarberg coefficient characteristics, spectral slope characteristics, mel-frequency cepstrum characteristics, spectral flow characteristics, ratio characteristics of loudness peaks, continuous and unvoiced regions characteristics, equivalent sound level characteristics.

6. The apparatus for constructing a classification model based on acoustic features that identifies different stages of heart failure according to claim 4, wherein the second speech feature sample comprises: glottal noise excitation ratio characteristics, vocal cord excitation ratio characteristics, cyclic period density entropy characteristics, trend fluctuation elimination analysis characteristics, sample entropy characteristics and multi-scale entropy characteristics.

7. The apparatus for constructing a classification model for identifying different stages of heart failure based on acoustic features of claim 1, further comprising:

and the evaluation unit is used for carrying out model evaluation on the classification model according to a set aside method.

8. The apparatus for constructing a classification model for identifying different stages of heart failure based on acoustic features of claim 1, further comprising:

and the performance measurement unit is used for performing performance measurement on the classification model according to a preset index, wherein the preset index comprises error rate and accuracy rate, precision rate and recall rate, F1 value, specificity, sensitivity, ROC curve and AUC, and unweighted average recall rate.

9. A computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, the processor implementing the following method when executing the computer program: converting the collected voice analog signals into voice digital signals, preprocessing the voice digital signals, and extracting features of the preprocessed voice digital signals to obtain multiple types of voice feature samples; constructing a classification model for identifying heart failure stage, and training and optimizing the classification model by utilizing the multi-class voice feature samples to obtain an optimal classification model, wherein the classification model for identifying heart failure stage A and heart failure stage B is an AdaBoost classification model based on an original variable; the classification model used for identifying the B phase and the C phase of heart failure is an AdaBoost classification model based on Lasso dimension reduction; the classification model used for identifying heart failure stage A and B and stage C is an AdaBoost classification model based on Lasso dimension reduction.

10. A computer readable storage medium, wherein a computer program is stored on the computer readable storage medium, the computer program when executed by a processor implementing the method of: converting the collected voice analog signals into voice digital signals, preprocessing the voice digital signals, and extracting features of the preprocessed voice digital signals to obtain multiple types of voice feature samples; constructing a classification model for identifying heart failure stage, and training and optimizing the classification model by utilizing the multi-class voice feature samples to obtain an optimal classification model, wherein the classification model for identifying heart failure stage A and heart failure stage B is an AdaBoost classification model based on an original variable; the classification model used for identifying the B phase and the C phase of heart failure is an AdaBoost classification model based on Lasso dimension reduction; the classification model used for identifying heart failure stage A and B and stage C is an AdaBoost classification model based on Lasso dimension reduction.

Images