CN110742621B - Signal processing method and computer equipment - Google Patents

Abstract

The invention provides a signal processing method and computer equipment, and relates to the technical field of signals. Wherein the method comprises the following steps: acquiring an original blood oxygen signal; preprocessing the original blood oxygen signal to obtain a first blood oxygen signal; determining a first time domain characteristic of the first blood oxygen signal; the first time domain feature is used for characterizing whether the blood oxygen saturation of the first blood oxygen signal is reduced or not; and inputting the first time domain characteristic into a preset respiratory classification model to obtain the target respiratory category to which the original blood oxygen signal belongs. In the embodiment of the present invention, the signal processing device may classify the first blood oxygen signal according to a time domain feature that can represent whether the blood oxygen saturation is decreased, and classify the first blood oxygen signal through a preset respiration classification model, so as to determine a respiration type to which the original blood oxygen signal corresponding to the first blood oxygen signal belongs.

Description

A signal processing method and computer equipment

technical field

The present invention relates to the field of signal technology, in particular to a signal processing method and computer equipment.

Background technique

In recent years, with the continuous development of signal technology, the application field of signal technology has become more and more extensive. For example, in the medical field, medical signal data such as blood oxygen signal and ECG signal can be processed, and the obtained processing results can assist Clinical analysis.

At present, people pay more and more attention to the quality of sleep, so the problem of sleep has gradually begun to be paid attention to. Apnea refers to the complete cessation of oral and nasal airflow for more than 10 s during sleep, and hypopnea refers to the decrease of respiratory airflow amplitude by more than 50% compared with the baseline level, and the blood oxygen saturation is decreased by 4% compared with the basic level. The apnea-hypopnea index (AHI) is the sum of the number of apnea and hypopnea per hour of sleep.

In the related art, whether it is a normal sleep breathing type can only be determined by a simple threshold discrimination method according to the sleep apnea-hypopnea index, but the accuracy of the threshold discrimination method is low.

SUMMARY OF THE INVENTION

The present invention provides a signal processing method and computer equipment to solve the problem that the existing sleep breathing type discrimination adopts a simple threshold discrimination method, so the accuracy rate is low.

In order to solve the above problems, the present invention discloses a signal processing method, including:

Obtain the original blood oxygen signal;

Preprocessing the original blood oxygen signal to obtain a first blood oxygen signal;

determining a first time domain feature of the first blood oxygen signal; the first time domain feature is used to represent whether the blood oxygen saturation of the first blood oxygen signal decreases;

Inputting the first time domain feature into a preset respiratory classification model to obtain a target respiratory category to which the original blood oxygen signal belongs.

Optionally, the determining the first time domain feature of the first blood oxygen signal includes:

performing window processing on the first blood oxygen signal to obtain n windowed blood oxygen signals;

Perform segmentation processing on each of the sub-window blood oxygen signals respectively, and obtain a total of n×m sub-window blood oxygen signals;

determining the blood oxygen saturation corresponding to each of the segmented blood oxygen signals;

The first time domain feature of the first blood oxygen signal is determined according to the blood oxygen saturation levels corresponding to the m segmented blood oxygen signals in each of the windowed blood oxygen signals.

Optionally, the first time domain feature includes the mean value of blood oxygen saturation, the standard deviation of blood oxygen saturation, the local maximum value of blood oxygen saturation, the local minimum value of blood oxygen saturation, the local range of blood oxygen saturation, the average value of blood oxygen saturation. At least one of percent first drop, average time to first drop, percent drop second, and time to second drop.

Optionally, determining the first time domain feature of the first blood oxygen signal according to the blood oxygen saturation levels corresponding to m segmented blood oxygen signals in each of the windowed blood oxygen signals, including:

For m segmented blood oxygen signals in each of the segmented blood oxygen signals, the blood oxygen saturation of the windowed blood oxygen signal is determined according to the blood oxygen saturation levels corresponding to the m segmented blood oxygen signals mean;

According to the blood oxygen saturation corresponding to the m segmented blood oxygen signals and the mean value of the blood oxygen saturation, determine the blood oxygen saturation standard deviation of the windowed blood oxygen signal;

determining the local maximum value of the blood oxygen saturation of the sub-window blood oxygen signal according to the blood oxygen saturation corresponding to the m segmented blood oxygen signals;

determining the local minimum value of the blood oxygen saturation of the sub-window blood oxygen signal according to the blood oxygen saturation corresponding to the m segmented blood oxygen signals;

According to each of the local maximum values of the blood oxygen saturation and each of the local minimum values of the blood oxygen saturation, determine the local range of the blood oxygen saturation of each of the windowed blood oxygen signals;

According to the blood oxygen saturation levels corresponding to the m segmented blood oxygen signals, determine the average first-order drop percentage of the windowed blood oxygen signals;

According to the blood oxygen saturation levels corresponding to the m segmented blood oxygen signals and a first preset constant, determine the average first-order continuous decline time of the windowed blood oxygen signals;

According to the mean value of blood oxygen saturation, determine the secondary drop percentage of the windowed blood oxygen signal;

According to the mean value of blood oxygen saturation and a second preset constant, the second-level continuous falling time of the windowed blood oxygen signal is determined.

Optionally, determining the average first-order drop percentage of the windowed blood oxygen signals according to the blood oxygen saturation levels corresponding to the m segmented blood oxygen signals, including:

According to the blood oxygen saturation levels corresponding to the m segmented blood oxygen signals, the following formula is used to determine the average first-order drop percentage of the windowed blood oxygen signals;

in,

MD1_wi represents the average first-order drop percentage of the i-th windowed blood-oxygen signal among the n segmented blood-oxygen signals; D1_di represents the i-th segmented blood-oxygen signal among the n×m segmented blood-oxygen signals. The first-order drop percentage of the segmented blood oxygen signal; Sd represents the blood oxygen saturation corresponding to the segmented blood oxygen signal;

The determining, according to the blood oxygen saturation levels corresponding to the m segmented blood oxygen signals and the first preset constant, the average first-order continuous decline time of the windowed blood oxygen signals, including:

According to the blood oxygen saturation corresponding to the m segmented blood oxygen signals and the first preset constant, the following formula is used to determine the average first-order continuous decline time of the windowed blood oxygen signals;

in,

MT1_wi represents the average first-order continuous fall time of the i-th sub-window blood oxygen signal among the n sub-window blood oxygen signals; T1_di represents the i-th sub-window blood oxygen signal in the n×m sub-window blood oxygen signals. the first-level continuous falling time of the segmented blood oxygen signal; a represents the first preset constant;

The determining the second-level drop percentage of the windowed blood oxygen signal according to the average blood oxygen saturation includes:

According to the mean value of blood oxygen saturation, the second-level drop percentage of the windowed blood oxygen signal is determined by the following formula;

Among them, D2_wi represents the secondary drop percentage of the i-th windowed blood oxygen signal among the n sub-window blood oxygen signals; Mean_wi represents the i-th sub-window blood oxygen signal in the n sub-window blood oxygen signals. mean blood oxygen saturation of the window blood oxygen signal;

The determining, according to the mean value of blood oxygen saturation and the second preset constant, the second-level continuous fall time of the sub-window blood oxygen signal includes:

According to the blood oxygen saturation mean value and the second preset constant, the second-level continuous fall time of the windowed blood oxygen signal is determined by the following formula;

Wherein, T2_wi represents the average first-order continuous fall time of the i-th windowed blood oxygen signal among the n blood oxygen signals in the sub-windows, and b represents the second preset constant.

Optionally, before obtaining the original blood oxygen signal, the method further includes:

acquiring a plurality of sample blood oxygen signals and a respiratory category corresponding to each of the sample blood oxygen signals;

Preprocessing each of the sample blood oxygen signals to obtain a plurality of second blood oxygen signals;

determining a second time domain feature of each of the second blood oxygen signals; the second time domain features are used to characterize whether the blood oxygen saturation of the second blood oxygen signal decreases;

Build an initial respiratory classification model;

The initial breathing classification model is trained by using the second time domain feature and the breathing category corresponding to each group as training parameters to obtain the preset breathing classification model.

Optionally, the preset breathing classification model includes a stepwise linear discriminant analysis model, a linear discriminant analysis model or a support vector machine model.

Optionally, the preset breathing classification model is a stepwise linear discriminant analysis model, and the initial breathing classification model is the stepwise linear discriminant analysis model to be trained; each of the second time domain features includes at least two sub-times. Domain features;

The said initial breathing classification model is trained using the second time domain feature and the breathing category corresponding to each group as training parameters, and the preset breathing classification model is obtained, including:

Taking each group of corresponding second time-domain features and the breathing category as training parameters, inputting the step-by-step linear discriminant analysis model to be trained;

According to each group of the corresponding second time domain features and the respiration category, a significance test is performed on each of the sub-time domain features in the second time domain features, and a significance weight exceeding a preset threshold is obtained. Sub-time domain features;

The step-by-step linear discriminant analysis model is obtained by training according to the sub-time domain features of which the saliency weight exceeds a preset threshold.

Optionally, the first time-domain feature and the second time-domain feature have the same sub-time-domain feature attribute of which the saliency weight exceeds the preset threshold.

In order to solve the above problems, the present invention also discloses a computer device, comprising a processor, a memory, and a computer program stored on the memory and executable on the processor, the computer program being executed by the processor When implementing the steps of the signal processing method as described above.

Compared with the prior art, the present invention includes the following advantages:

In this embodiment of the present invention, the signal processing device may first obtain the original blood oxygen signal, and then may preprocess the original blood oxygen signal to obtain the first blood oxygen signal, and then the signal processing device may determine the first blood oxygen signal of the first blood oxygen signal. A time-domain feature, the first time-domain feature can represent whether the blood oxygen saturation of the first blood-oxygen signal has decreased, and then the signal processing device uses the first time-domain feature as an input parameter and inputs a preset respiratory classification model to obtain The target breathing class to which the raw blood oxygen signal belongs. In this embodiment of the present invention, the signal processing device may classify the first blood oxygen signal according to a time domain feature that can characterize whether the blood oxygen saturation is decreased, and use a preset respiratory classification model to determine the corresponding blood oxygen signal of the first blood oxygen signal. Compared with the threshold determination method based on the number of apnea and the number of hypopnea, the breathing category to which the original blood oxygen signal belongs, improves the accuracy of judging the breathing type.

Description of drawings

1 shows a flowchart of a signal processing method according to Embodiment 1 of the present invention;

2 shows a flowchart of a signal processing method according to Embodiment 2 of the present invention;

Fig. 3 shows the calculation flow chart of a first-level continuous descent time according to the second embodiment of the present invention;

FIG. 4 shows a flow chart of calculating a second-level continuous falling time according to the second embodiment of the present invention.

Detailed ways

In order to make the above objects, features and advantages of the present invention more clearly understood, the present invention will be described in further detail below with reference to the accompanying drawings and specific embodiments.

Example 1

Referring to FIG. 1, a flowchart of steps of a signal processing method according to Embodiment 1 of the present invention is shown, and the method includes the following steps:

Step 101: Obtain the original blood oxygen signal.

In the embodiment of the present invention, after the blood oxygen signal detection device is connected to the human body, the original blood oxygen signal of the human body can be collected within a period of time, and the collected raw blood oxygen signal data can be imported into the signal processing device, Therefore, the signal processing device can obtain the original blood oxygen signal. Optionally, the blood oxygen signal may specifically be a pulse blood oxygen signal, which is not specifically limited in this embodiment of the present invention, as long as it is a signal capable of measuring the blood oxygen content.

Step 102: Preprocess the original blood oxygen signal to obtain a first blood oxygen signal.

In this embodiment of the present invention, after acquiring the original blood oxygen signal, the signal processing device may perform preprocessing such as filtering on the original blood oxygen signal for noise reduction, so as to reduce noise interference in the original blood oxygen signal, so as to reduce the noise in the original blood oxygen signal. The first blood oxygen signal after noise.

Step 103: Determine a first time domain feature of the first blood oxygen signal; the first time domain feature is used to represent whether the blood oxygen saturation of the first blood oxygen signal decreases.

In this step, the signal processing device may extract a first time domain feature of the first blood oxygen signal, wherein the first time domain feature is some time domain features that can characterize whether the blood oxygen saturation of the first blood oxygen signal has decreased. For example, the first time domain feature may include the mean value of blood oxygen saturation, the standard deviation of blood oxygen saturation, the local maximum value of blood oxygen saturation, the local minimum value of blood oxygen saturation, the local range of blood oxygen saturation, and the like. All the features can directly or indirectly represent whether the blood oxygen saturation of the first blood oxygen signal decreases, and the first time domain feature is not specifically limited in this embodiment of the present invention.

Step 104: Input the first time domain feature into a preset respiratory classification model to obtain the target respiratory category to which the original blood oxygen signal belongs.

In this step, a preset respiration classification model may be constructed in the signal processing device in advance, for determining the respiration class to which the blood oxygen signal belongs according to the time domain feature of the input blood oxygen signal, and outputting the respiration class. In specific applications, the breathing categories may include normal breathing categories and sleep apnea categories. The signal processing device can use the first time-domain feature as an input parameter to input a preset respiratory classification model, so that the preset respiratory classification model can output the target respiratory category corresponding to the first time-domain feature, that is, to which the output original blood oxygen signal belongs. Target breathing class.

In this embodiment of the present invention, the signal processing device may first obtain the original blood oxygen signal, and then may preprocess the original blood oxygen signal to obtain the first blood oxygen signal, and then the signal processing device may determine the first blood oxygen signal of the first blood oxygen signal. A time-domain feature, the first time-domain feature can represent whether the blood oxygen saturation of the first blood-oxygen signal has decreased, and then the signal processing device uses the first time-domain feature as an input parameter and inputs a preset respiratory classification model to obtain The target breathing class to which the raw blood oxygen signal belongs. In this embodiment of the present invention, the signal processing device may classify the first blood oxygen signal according to a time domain feature that can characterize whether the blood oxygen saturation is decreased, and use a preset respiratory classification model to determine the corresponding blood oxygen signal of the first blood oxygen signal. Compared with the threshold determination method based on the number of apnea and the number of hypopnea, the breathing category to which the original blood oxygen signal belongs, improves the accuracy of judging the breathing type.

Embodiment 2

Referring to FIG. 2, a flowchart of steps of a signal processing method according to Embodiment 2 of the present invention is shown, and the method includes the following steps:

Step 201: Obtain a preset breathing classification model through training.

In this embodiment of the present invention, before classifying the respiratory category based on the original blood oxygen signal, the signal processing device may first obtain a preset respiratory classification model through training, which specifically includes the following steps: acquiring a plurality of sample blood oxygen signals and each Respiration category corresponding to the sample blood oxygen signal; preprocess each sample blood oxygen signal to obtain a plurality of second blood oxygen signals; determine the second time domain feature of each second blood oxygen signal; the second time domain feature Used to characterize whether the blood oxygen saturation of the second blood oxygen signal has dropped; build an initial respiratory classification model; use the second time domain features and respiratory categories corresponding to each group as training parameters to train the initial respiratory classification model to obtain a preset Respiratory classification model.

First, a plurality of sample blood oxygen signals can be input to the signal processing device, as well as the corresponding marked respiration category for each sample blood oxygen signal. The way can be the same. Then, the signal processing device may perform preprocessing for noise reduction, such as filtering, on each sample blood oxygen signal, so as to reduce noise interference in each sample blood oxygen signal, so as to obtain a plurality of second blood samples after noise reduction. oxygen signal. After that, the signal processing device can extract the second time domain feature of each second blood oxygen signal, wherein the second time domain feature is some time domain features that can characterize whether the blood oxygen saturation of the second blood oxygen signal is decreased, and The first time domain characteristic properties are the same. Further, the signal processing device can construct an initial respiratory classification model, in which each model parameter is to be trained. Next, the signal processing device can construct a training feature matrix corresponding to the second time-domain feature of each group and the marked breathing category, and use the training feature matrix as a training parameter to input it into the initial breathing classification model for parameter adjustment, thereby realizing Training of the initial breath classification model. After the training is completed, each model parameter in the initial breathing classification model is determined, thereby obtaining a trained preset breathing classification model.

Optionally, in practical applications, the preset respiratory classification model may include a stepwise linear discriminant analysis (Stepwise Linear Discriminant Analysis, SWLDA) model, a linear discriminant analysis (Linear Discriminant Analysis, LDA) model or a support vector machine (Support Vector Machine, SVM) model. Model. Among them, the step-by-step linear discriminant analysis model can combine the two methods of LDA and two-way step-by-step analysis to carry out the significance test of each dimension on the input features, and finally retain only the feature combination that contributes the most to the classification result to establish a classification model, which can greatly The magnitude reduces the number of features to avoid overfitting. In specific applications, the step-by-step linear discriminant analysis model can optimize the model by adjusting the p-value of salient features, excluding the p-value of salient features, and the total number of salient features, these three important parameters, so that the classification results of the model are more accurate. precise.

Specifically, when the preset breathing classification model is a step-by-step linear discriminant analysis model, the initial breathing classification model is a step-by-step linear discriminant analysis model to be trained, and each second time domain feature includes at least two sub-time domain features, each The second time domain feature corresponding to the group and the breathing category are training parameters, the initial breathing classification model is trained, and the step of obtaining the preset breathing classification model can be specifically implemented by the following methods, including:

Taking each group of corresponding second time domain features and respiratory categories as training parameters, input the step-by-step linear discriminant analysis model to be trained; The saliency test of each sub-time domain feature is carried out, and the sub-time domain feature whose saliency weight exceeds the preset threshold is obtained; according to the sub-time domain feature whose saliency weight exceeds the preset threshold, the step-by-step linear discriminant analysis model is obtained by training.

Wherein, the signal processing device may use each group of corresponding second time-domain features and respiratory categories as training parameters, and input the step-by-step linear discriminant analysis model to be trained. Wherein, each second time-domain feature may include at least two sub-time-domain features, and the sub-time-domain features may be, for example, the mean value of blood oxygen saturation, the standard deviation of blood oxygen saturation, the local maximum value of blood oxygen saturation, and the blood oxygen saturation Local minimum value, local range of blood oxygen saturation, average first-order drop percentage, average first-order continuous drop time, second-order drop percentage, and second-order continuous drop time, etc. The step-by-step linear discriminant analysis model to be trained can be combined with LDA and two-way step-by-step analysis. Based on the parameter adjustment results of each group of corresponding second time-domain features and respiratory categories for model parameters, determine the contribution of each sub-time-domain feature to the classification result. In the decision-making stage of training, the step-by-step linear discriminant analysis model to be trained can obtain the saliency weight corresponding to each sub-time domain feature, and according to the saliency weight Sort each sub-temporal feature in order from largest to smallest. The sub-temporal features with higher saliency weights have a more significant impact on the accuracy of classification results. Furthermore, the signal processing device can train to obtain a step-by-step linear discriminant analysis model according to the sub-time-domain features whose significance weight exceeds the preset threshold, that is, the sub-time-domain features that have the greatest impact on the accuracy of the classification result.

In this embodiment of the present invention, the first time domain feature and the second time domain feature have the same sub-time domain feature attribute whose saliency weight exceeds a preset threshold. Since the step-by-step linear discriminant analysis model can be obtained by training according to the sub-time-domain features in the second time-domain feature that have a significant impact on the accuracy of the classification result, the original blood oxygen signal is then subjected to subsequent training through the trained preset respiratory classification model. When classifying, only the time domain features that have the same attributes as the sub-time-domain features with significant influence can be extracted from the original blood oxygen signal, and it is not necessary to extract all the sub-time-domain features from the original blood oxygen signal and the second time-domain features. Therefore, the number of time-domain features extracted from the original blood oxygen signal can be reduced, thereby reducing the time-domain feature calculation amount of the original blood oxygen signal.

For example, the second time domain feature may include the following 9 sub-time domain features: mean value of blood oxygen saturation, standard deviation of blood oxygen saturation, local maximum value of blood oxygen saturation, local minimum value of blood oxygen saturation, local blood oxygen saturation value Range, Average 1st Drop, Average 1st Duration, 2nd %, and 2nd Duration. During model training, it can be determined that there are 6 sub-time-domain features in the second time-domain feature that have a significant impact on the accuracy of the classification result, which are the average blood oxygen saturation, the local range of blood oxygen saturation, and the average first-level drop. Percentage, average first-degree duration of descent, second-degree percentage of descent, and second-degree duration of descent. The signal processing device can train to obtain a step-by-step linear discriminant analysis model according to the above-mentioned six sub-time domain features with significant influence. Therefore, when extracting the first time domain feature that needs to be input from the original blood oxygen signal, only the mean blood oxygen saturation, the local range of blood oxygen saturation, the average first-order drop percentage, and the average first-order drop of the original blood oxygen signal can be extracted. The six time domain features of the first-level continuous decline time, the second-level decline percentage, and the second-level continuous decline time can be used as the input parameters of the step-by-step linear discriminant analysis model. The properties of the six sub-time domain features that have a significant impact on the accuracy of the classification results are the same, so the number of time domain features extracted from the original blood oxygen signal is reduced, and there is no need to calculate the standard deviation of blood oxygen saturation, blood The local maximum value of oxygen saturation and the local minimum value of blood oxygen saturation reduce the amount of time-domain feature calculation of the original blood oxygen signal.

After the trained preset breathing classification model is obtained, the signal processing device can obtain a new original blood oxygen signal, and then according to the time-domain characteristics corresponding to the new original blood oxygen signal, through the trained preset breathing classification model. Classification.

Step 202: Obtain the original blood oxygen signal.

For the specific implementation manner of this step, reference may be made to the implementation process of the foregoing step 101, which will not be described in detail in this embodiment.

Step 203: Preprocess the original blood oxygen signal to obtain a first blood oxygen signal.

For a specific implementation manner of this step, reference may be made to the implementation process of the foregoing step 102, which is not described in detail in this embodiment.

Step 204: Perform window processing on the first blood oxygen signal to obtain n windowed blood oxygen signals.

In this embodiment of the present invention, the signal processing device may perform window processing on the preprocessed first blood oxygen signal according to the preset window length, and may obtain n windowed blood oxygen signals, wherein each windowed blood oxygen signal is The duration corresponding to the oxygen signal is the preset window length. For example, the preset window length may be 60s, then the first blood oxygen signal may be subjected to window processing every 60s to obtain n 60s windowed blood oxygen signals.

Step 205 : Perform segmental processing on each sub-window blood oxygen signal respectively, and obtain n×m segmented blood oxygen signals in total.

In this embodiment of the present invention, the signal processing device may perform segment processing on each windowed blood oxygen signal according to a preset segment length, and each windowed blood oxygen signal may be segmented into m segmented blood oxygen signals , a total of n×m segmented blood oxygen signals can be obtained, wherein the time length corresponding to each segmented blood oxygen signal is the preset segment length. For example, the preset segment length can be 10s, then a 60s windowed blood oxygen signal can be segmented every 10s, and each windowed blood oxygen signal can be segmented into six 10s segmented blood oxygen signals, A total of 6n segmented blood oxygen signals.

Step 206: Determine the blood oxygen saturation corresponding to each segmented blood oxygen signal.

In this step, the signal processing device may determine the blood oxygen saturation corresponding to each segmented blood oxygen signal. Optionally, the blood oxygen partial pressure can be measured based on the electrochemical principle, and then the blood oxygen saturation can be calculated according to the blood oxygen partial pressure. For details, reference may be made to the related art. The embodiment of the present invention does not specifically limit the method for determining the blood oxygen saturation. .

Step 207 : Determine the first time domain feature of the first blood oxygen signal according to the blood oxygen saturation levels corresponding to the m segmented blood oxygen signals in each windowed blood oxygen signal.

Optionally, the first time domain feature may include the mean value of blood oxygen saturation, the standard deviation of blood oxygen saturation, the local maximum value of blood oxygen saturation, the local minimum value of blood oxygen saturation, the local range of blood oxygen saturation, and the average at least one of percent decline in grades, average duration of first grade declines, percent declines in grades two, and durations in second grades.

Correspondingly, this step can be implemented in the following ways, including:

Sub-step 2071: For the m segmented blood oxygen signals in each windowed blood oxygen signal, determine the average blood oxygen saturation of the windowed blood oxygen signals according to the blood oxygen saturation levels corresponding to the m segmented blood oxygen signals.

Taking the 60s windowed blood oxygen signal, the 10s segmented blood oxygen signal and m=6 as examples, in the case of considering the window boundaries, the blood oxygen saturation corresponding to the 6n segmented blood oxygen signals can be expressed as Sw1d1, Sw1d2, Sw1d3, Sw1d4, Sw1d5, Sw1d6, Sw2d1, Sw2d2, Sw2d3, Sw2d4, Sw2d5, Sw2d6, …, Swnd5, Swnd6, can be recorded as a secondary blood oxygen saturation sequence. Without considering the boundaries of the sub-windows, the blood oxygen saturation levels corresponding to the 6n segmented blood oxygen signals can be expressed as Sd1, Sd2, Sd3, ..., Sd6n, which can be recorded as a first-order blood oxygen saturation sequence. Among them, the values of the primary blood oxygen saturation sequence and the secondary blood oxygen saturation sequence are the same, but the representation is different. The representation of the secondary blood oxygen saturation sequence can reflect the sub-window to which the segmented blood oxygen signal belongs. The blood oxygen signal, and the representation of the first-level blood oxygen saturation sequence does not reflect the boundary of the window, and can be named according to requirements in specific applications, which is not limited in this embodiment of the present invention.

Optionally, with m segmented blood oxygen signals in a certain windowed blood oxygen signal for illustration, the signal processing device can be based on the blood oxygen saturation corresponding to the m segmented blood oxygen signals in the windowed blood oxygen signal. , the mean value of the blood oxygen saturation of the sub-window blood oxygen signal is determined by the following formula (1). Among them, in the following formula (1), Mean_wi represents the blood oxygen saturation mean value of the i-th sub-window blood oxygen signal in the n sub-window blood oxygen signals, and Swidj represents the j-th blood oxygen signal in the i-th sub-window blood oxygen signal The blood oxygen saturation corresponding to each segmented blood oxygen signal.

For example, when m=6, the formula (1) can be embodied as the following formula (1-1).

Sub-step 2072: Determine the blood oxygen saturation standard deviation of the windowed blood oxygen signals according to the blood oxygen saturation levels and the mean blood oxygen saturation levels corresponding to the m segmented blood oxygen signals.

Optionally, with m segmented blood oxygen signals in a certain windowed blood oxygen signal for illustration, the signal processing device can be based on the blood oxygen saturation corresponding to the m segmented blood oxygen signals in the windowed blood oxygen signal. , and the mean value of the blood oxygen saturation of the sub-window blood oxygen signal, the standard deviation of the blood oxygen saturation of the sub-window blood oxygen signal is determined by the following formula (2). Wherein, in the following formula (2), Sd_wi represents the blood oxygen saturation standard deviation of the i-th sub-window blood oxygen signal.

For example, when m=6, the formula (2) can be embodied as the following formula (2-1).

Sub-step 2073: Determine the local maximum value of the blood oxygen saturation of the windowed blood oxygen signal according to the blood oxygen saturation corresponding to the m segmented blood oxygen signals.

Optionally, with m segmented blood oxygen signals in a certain windowed blood oxygen signal for illustration, the signal processing device can be based on the blood oxygen saturation corresponding to the m segmented blood oxygen signals in the windowed blood oxygen signal. , the local maximum value of the blood oxygen saturation of the sub-window blood oxygen signal is determined by the following formula (3). Wherein, in the following formula (3), Max_wi represents the local maximum value of the blood oxygen saturation of the i-th sub-window blood oxygen signal.

Max_wi=max{Swidj},(j=1,2,3,...,m,i=1,2,3,...,n) (3)

For example, when m=6, the formula (3) can be embodied as the following formula (3-1).

Max_wi=max{Swid1,Swid2,Swid3,Swid4,Swid5,Swid6},(j=1,2,3,...,m,i=1,2,3,...,n)

(3-1)

Sub-step 2074: Determine the local minimum value of the blood oxygen saturation of the windowed blood oxygen signal according to the blood oxygen saturation corresponding to the m segmented blood oxygen signals.

Optionally, with m segmented blood oxygen signals in a certain windowed blood oxygen signal for illustration, the signal processing device can be based on the blood oxygen saturation corresponding to the m segmented blood oxygen signals in the windowed blood oxygen signal. , the local minimum value of the blood oxygen saturation of the windowed blood oxygen signal is determined by the following formula (4). Wherein, in the following formula (4), Min_wi represents the local minimum value of the blood oxygen saturation of the i-th sub-window blood oxygen signal.

Min_wi=min{Swidj},(j=1,2,3,…,m,i=1,2,3,…,n) (4)

For example, when m=6, the formula (4) can be embodied as the following formula (4-1).

Min_wi=min{Swid1,Swid2,Swid3,Swid4,Swid5,Swid6},(j=1,2,3,...,m,i=1,2,3,...,n)

(4-1)

Sub-step 2075: According to each local maximum value of blood oxygen saturation and each local minimum value of blood oxygen saturation, determine the local range of blood oxygen saturation of each sub-window blood oxygen signal.

Optionally, with a certain windowed blood oxygen signal for illustration, the signal processing device can use the following formula (5) according to the local maximum value of blood oxygen saturation and the local minimum value of blood oxygen saturation of the windowed blood oxygen signal. The local range of blood oxygen saturation of the sub-window blood oxygen signal is determined. Wherein, in the following formula (5), R_wi represents the local range of blood oxygen saturation of the i-th sub-window blood oxygen signal.

R_wi=Max_wi-Min_wi,(i=1,2,3,...,n) (5)

Sub-step 2076: Determine the average first-order drop percentage of the windowed blood oxygen signals according to the blood oxygen saturation levels corresponding to the m segmented blood oxygen signals.

Optionally, the signal processing device may determine the first-order drop percentage of each segmented blood oxygen signal according to the blood oxygen saturation corresponding to the n×m segmented blood oxygen signals by the following formula (6). Wherein, in the following formula (6), D1_di represents the first-order drop percentage of the i-th segmented blood-oxygen signal among the n×m segmented blood-oxygen signals.

For example, when m=6, the formula (6) can be embodied as the following formula (6-1).

Then, taking the m segmented blood oxygen signals in a certain windowed blood oxygen signal for illustration, the signal processing device can, according to the first-order drop percentage of the m segmented blood oxygen signals in the windowed blood oxygen signal, pass the following The above formula (7) determines the average first-order drop percentage of the sub-window blood oxygen signal. Wherein, in the following formula (7), MD1_wi represents the average first-order drop percentage of the blood oxygen signal of the i-th sub-window in the blood oxygen signals of the n sub-windows.

For example, when m=6, the formula (7) can be embodied as the following formula (7-1).

Sub-step 2077: According to the blood oxygen saturation levels corresponding to the m segmented blood oxygen signals and the first preset constant, determine the average first-order continuous fall time of the windowed blood oxygen signals.

Optionally, the signal processing device can determine the first-order drop of each segmented blood oxygen signal by the following formula (8) according to the blood oxygen saturation corresponding to the n×m segmented blood oxygen signals and the first preset constant. percentage. Among them, in the following formula (8), T1_di represents the first-level continuous fall time of the i-th segmented blood oxygen signal in the n×m segmented blood oxygen signals, a represents the first preset constant, the first preset The constant is equal to the preset segment length. The calculation process represented by formula (8) can refer to FIG. 3 .

For example, when m=6 and the preset segment length=10s, the formula (8) can be embodied as the following formula (8-1).

Then, taking the m segmented blood oxygen signals in a certain windowed blood oxygen signal for illustration, the signal processing device can, according to the first-order drop percentage of the m segmented blood oxygen signals in the windowed blood oxygen signal, pass the following The above formula (9) determines the average first-order continuous falling time of the sub-window blood oxygen signal. Wherein, in the following formula (9), MT1_wi represents the average first-order continuous fall time of the blood oxygen signal of the i-th sub-window in the blood oxygen signals of the n sub-windows.

For example, when m=6, the formula (9) can be embodied as the following formula (9-1).

Sub-step 2078: Determine the secondary drop percentage of the windowed blood oxygen signal according to the mean blood oxygen saturation.

Optionally, the signal processing device may determine the secondary drop percentage of each sub-window blood oxygen signal by the following formula (10) according to the blood oxygen saturation mean value corresponding to each sub-window blood oxygen signal. Wherein, in the following formula (10), D2_wi represents the secondary drop percentage of the blood oxygen signal of the i-th sub-window in the blood oxygen signals of the n sub-windows.

Sub-step 2079: Determine the secondary continuous fall time of the windowed blood oxygen signal according to the mean blood oxygen saturation and the second preset constant.

Optionally, the signal processing device can determine the secondary continuous fall time of each windowed blood oxygen signal by the following formula (11) according to the blood oxygen saturation mean value of each windowed blood oxygen signal and the second preset constant: . Among them, in the following formula (11), T2_wi represents the average first-order continuous fall time of the ith windowed blood oxygen signal in the n sub-window blood oxygen signals, b represents the second preset constant, the second preset constant equal to the default window length. The calculation process represented by formula (11) can refer to FIG. 4 .

For example, when the preset window length=60s, the formula (8) can be embodied as the following formula (8-1).

It should be noted that, in practical applications, the preset breathing classification model can determine the time-domain feature dimensions that have a significant impact on the classification results from the time-domain features of each dimension during training. The time domain feature dimension determined by the preset breathing classification model that has a significant impact on the classification result can be extracted by extracting the first time domain feature of the same dimension. For example, the time-domain feature dimensions determined by the preset respiratory classification model that have a significant impact on the classification result are the average first-level drop percentage, the average first-level continuous drop time, the second-level drop percentage, and the second-level continuous drop time. The first blood oxygen signal can extract the time domain features of the average first-level drop percentage, the average first-level continuous drop time, the second-level drop percentage, and the second-level continuous drop time. Correspondingly, the first time-domain feature of which dimension needs to be extracted , the corresponding feature determination step only needs to be performed, and for the dimension that does not need to be extracted, the feature determination step of the corresponding dimension does not need to be performed.

In addition, in this embodiment of the present invention, only the time domain features of the first blood oxygen signal can be extracted, without the need to extract the frequency domain, spectral, nonlinear and other features, so the classification feature extraction and operation can be complicated. low degree.

Step 208: Input the first time domain feature into a preset respiratory classification model to obtain the target respiratory category to which the original blood oxygen signal belongs.

In this embodiment of the present invention, the signal processing device may construct the first time domain feature as a first feature matrix, and use the first feature matrix as an input parameter to input the trained preset respiratory classification model, thereby preset respiratory classification The model can output the target breathing category corresponding to the first time domain feature, that is, the target breathing category to which the output original blood oxygen signal belongs, so as to determine whether the original blood oxygen signal belongs to the normal breathing category or the sleep apnea category.

In addition, in practical applications, optionally, different preset respiration classification models can be trained for different blood oxygen signal owners, so that when classifying sleep breathing types, the preset corresponding to the current blood oxygen signal owner is used. A respiratory classification model is set for classification, so that the classification accuracy can be improved. Of course, optionally, the same preset respiration classification model can also be used for the blood oxygen signals of different blood oxygen signal owners, that is, the preset respiration classification model can be used universally. In this way, the model training time can be saved. .

In the embodiment of the present invention, the signal processing device can first obtain a preset respiratory classification model through training, and then can obtain the original blood oxygen signal, and preprocess the original blood oxygen signal to obtain the first blood oxygen signal, and then perform signal processing. The device can perform window processing and segmentation processing on the first blood oxygen signal to obtain n windowed blood oxygen signals. Each windowed blood oxygen signal includes m segmented blood oxygen signals. The blood oxygen saturation corresponding to the m segmented blood oxygen signals in the oxygen signal determines the first time domain feature of the first blood oxygen signal, and the first time domain feature can represent whether the blood oxygen saturation of the first blood oxygen signal is Then, the signal processing device uses the first time domain feature as an input parameter, and inputs the preset respiratory classification model to obtain the target respiratory category to which the original blood oxygen signal belongs. In the embodiment of the present invention, the signal processing device may perform time-domain feature extraction on the first blood oxygen signal by window and segment, so as to obtain multi-dimensional and numerous time-domain features, and the blood oxygen can be characterized according to the first blood oxygen signal. The time-domain feature of whether the saturation level is decreased is classified by the preset breathing classification model, so as to determine the breathing category to which the original blood oxygen signal corresponding to the first blood oxygen signal belongs, which is compared with the threshold based on the number of apnea and the number of hypopnea. The method improves the accuracy of judging the type of breathing.

An embodiment of the present invention also provides a computer device, including a processor, a memory, and a computer program stored on the memory and running on the processor, the computer program being executed by the processor to achieve the above The steps of the signal processing method.

For the foregoing method embodiments, for the sake of simple description, they are all expressed as a series of action combinations, but those skilled in the art should know that the present invention is not limited by the described action sequence, because according to the present invention, Certain steps may be performed in other orders or simultaneously. Secondly, those skilled in the art should also know that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily required by the present invention.

The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments, and the same and similar parts between the various embodiments may be referred to each other.

Finally, it should also be noted that in this document, relational terms such as first and second are used only to distinguish one entity or operation from another, and do not necessarily require or imply these entities or that there is any such actual relationship or sequence between operations. Furthermore, the terms "comprising", "comprising" or any other variation thereof are intended to encompass a non-exclusive inclusion such that a process, method, article of manufacture or device comprising a list of elements includes not only those elements, but also includes not explicitly listed or other elements inherent to such a process, method, commodity or apparatus. Without further limitation, an element qualified by the phrase "comprising a..." does not preclude the presence of additional identical elements in the process, method, article of manufacture, or device that includes the element.

A signal processing method and computer equipment provided by the present invention have been introduced in detail above. The principles and implementations of the present invention are described with specific examples in this paper. The descriptions of the above embodiments are only used to help understand the present invention. At the same time, for those skilled in the art, according to the idea of the present invention, there will be changes in the specific implementation and application scope. In summary, the content of this specification should not be construed as Limitations of the present invention.

Claims (10)

1. A signal processing device, wherein the signal processing device is configured to perform the following steps: Obtain the original blood oxygen signal; Preprocessing the original blood oxygen signal to obtain a first blood oxygen signal; determining a first time domain feature of the first blood oxygen signal; the first time domain feature is used to represent whether the blood oxygen saturation of the first blood oxygen signal decreases; Inputting the first time domain feature into a preset respiration classification model to obtain the target respiration category to which the original blood oxygen signal belongs; the preset respiration classification model is based on the second time domain feature of the second blood oxygen signal and the label The training feature matrix constructed by the respiration category is used as a training parameter, and the initial respiration classification model is input and obtained by training the initial respiration classification model; wherein, the second time domain feature is the same as the first time domain feature attribute; Wherein, the second blood oxygen signal is obtained by preprocessing the obtained sample blood oxygen signal. 2 . The signal processing device according to claim 1 , wherein the determining the first time domain feature of the first blood oxygen signal comprises: 2 . performing window processing on the first blood oxygen signal to obtain n windowed blood oxygen signals; Perform segmentation processing on each of the sub-window blood oxygen signals respectively, and obtain a total of n×m sub-window blood oxygen signals; determining the blood oxygen saturation corresponding to each of the segmented blood oxygen signals; The first time domain feature of the first blood oxygen signal is determined according to the blood oxygen saturation levels corresponding to the m segmented blood oxygen signals in each of the windowed blood oxygen signals. 3 . The signal processing device according to claim 2 , wherein the first time domain feature comprises the mean value of blood oxygen saturation, the standard deviation of blood oxygen saturation, the local maximum value of blood oxygen saturation, the At least one of a local minimum of saturation, a local range of blood oxygen saturation, an average first-order drop percentage, an average first-order sustained drop time, a second-order drop percentage, and a second-order sustained drop time. 4 . The signal processing device according to claim 3 , wherein the determination of the said The first time domain feature of the first blood oxygen signal includes: For each of the m segmented blood oxygen signals in the segmented blood oxygen signals, the blood oxygen saturation of the windowed blood oxygen signal is determined according to the blood oxygen saturation levels corresponding to the m segmented blood oxygen signals mean; According to the blood oxygen saturation corresponding to the m segmented blood oxygen signals and the mean value of the blood oxygen saturation, determine the blood oxygen saturation standard deviation of the windowed blood oxygen signal; determining the local maximum value of the blood oxygen saturation of the sub-window blood oxygen signal according to the blood oxygen saturation corresponding to the m segmented blood oxygen signals; determining the local minimum value of the blood oxygen saturation of the sub-window blood oxygen signal according to the blood oxygen saturation corresponding to the m segmented blood oxygen signals; According to each of the local maximum values of the blood oxygen saturation and each of the local minimum values of the blood oxygen saturation, determine the local range of the blood oxygen saturation of each of the windowed blood oxygen signals; According to the blood oxygen saturation levels corresponding to the m segmented blood oxygen signals, determine the average first-order drop percentage of the windowed blood oxygen signals; According to the blood oxygen saturation levels corresponding to the m segmented blood oxygen signals and a first preset constant, determine the average first-order continuous decline time of the windowed blood oxygen signals; According to the mean value of blood oxygen saturation, determine the secondary drop percentage of the windowed blood oxygen signal; According to the mean value of blood oxygen saturation and a second preset constant, the second-level continuous falling time of the windowed blood oxygen signal is determined. 5 . The signal processing device according to claim 4 , wherein the average level of the windowed blood oxygen signals is determined according to the blood oxygen saturation levels corresponding to the m segmented blood oxygen signals. 6 . Percent decline, including: According to the blood oxygen saturation levels corresponding to the m segmented blood oxygen signals, the following formula is used to determine the average first-order drop percentage of the windowed blood oxygen signals;

in,

MD1_wi represents the average first-order drop percentage of the i-th windowed blood-oxygen signal among the n segmented blood-oxygen signals; D1_di represents the i-th segmented blood-oxygen signal among the n×m segmented blood-oxygen signals. The first-order drop percentage of the segmented blood oxygen signal; Sd represents the blood oxygen saturation corresponding to the segmented blood oxygen signal; The determining, according to the blood oxygen saturation levels corresponding to the m segmented blood oxygen signals and the first preset constant, the average first-order continuous decline time of the windowed blood oxygen signals, including: According to the blood oxygen saturation corresponding to the m segmented blood oxygen signals and the first preset constant, the following formula is used to determine the average first-order continuous decline time of the windowed blood oxygen signals;

in,

MT1_wi represents the average first-order continuous fall time of the i-th sub-window blood oxygen signal among the n sub-window blood oxygen signals; T1_di represents the i-th sub-window blood oxygen signal in the n×m sub-window blood oxygen signals. the first-level continuous falling time of the segmented blood oxygen signal; a represents the first preset constant; The determining the second-level drop percentage of the windowed blood oxygen signal according to the average blood oxygen saturation includes: According to the mean value of blood oxygen saturation, the second-level drop percentage of the windowed blood oxygen signal is determined by the following formula;

Among them, D2_wi represents the secondary drop percentage of the i-th windowed blood oxygen signal among the n sub-window blood oxygen signals; Mean_wi represents the i-th sub-window blood oxygen signal in the n sub-window blood oxygen signals. mean blood oxygen saturation of the window blood oxygen signal; The determining, according to the mean value of blood oxygen saturation and the second preset constant, the second-level continuous falling time of the windowed blood oxygen signal includes: According to the blood oxygen saturation mean value and the second preset constant, the second-level continuous fall time of the windowed blood oxygen signal is determined by the following formula;

Wherein, T2_wi represents the average first-order continuous fall time of the i-th windowed blood oxygen signal among the n blood oxygen signals in the sub-windows, and b represents the second preset constant. 6 . The signal processing device according to claim 1 , wherein before acquiring the original blood oxygen signal, the method further comprises: 6 . acquiring a plurality of sample blood oxygen signals and a respiratory category corresponding to each of the sample blood oxygen signals; Preprocessing each of the sample blood oxygen signals to obtain a plurality of second blood oxygen signals; determining a second time domain feature of each of the second blood oxygen signals; the second time domain features are used to characterize whether the blood oxygen saturation of the second blood oxygen signal decreases; Build an initial respiratory classification model; The initial breathing classification model is trained by using the second time domain feature and the breathing category corresponding to each group as training parameters to obtain the preset breathing classification model. 7 . The signal processing device according to claim 1 , wherein the preset breathing classification model comprises a stepwise linear discriminant analysis model, a linear discriminant analysis model or a support vector machine model. 8 . 8 . The signal processing device according to claim 6 , wherein the preset breathing classification model is a stepwise linear discriminant analysis model, and the initial breathing classification model is the stepwise linear discriminant analysis model to be trained. 9 . ; Each of the second time domain features includes at least two sub-time domain features; The said initial breathing classification model is trained using the second time domain feature and the breathing category corresponding to each group as training parameters, and the preset breathing classification model is obtained, including: Taking each group of corresponding second time-domain features and the breathing category as training parameters, inputting the step-by-step linear discriminant analysis model to be trained; According to each group of the corresponding second time domain features and the respiration category, a significance test is performed on each of the sub-time domain features in the second time domain features, and a significance weight exceeding a preset threshold is obtained. Sub-time domain features; The step-by-step linear discriminant analysis model is obtained by training according to the sub-time domain features of which the saliency weight exceeds a preset threshold. 9 . The signal processing device according to claim 8 , wherein in the first time domain feature and the second time domain feature, the saliency weight exceeds the sub-time domain of the preset threshold. 10 . Feature properties are the same. 10. A computer device, characterized in that it comprises a processor, a memory, and a computer program stored on the memory and executable on the processor, and the computer program is executed by the processor to achieve the right A signal processing device according to any one of claims 1 to 9 is configured to perform steps.

Images