CN114391808A - A sleep staging method and system based on nonlinear interdependence - Google Patents

Abstract

The invention provides a sleep staging method and a system based on nonlinear interdependence, which comprises the steps of obtaining two electroencephalogram signals of a tested person within a certain period of time; extracting various nonlinear interdependencies between the two electroencephalogram signals and cross-correlation coefficients between the two electroencephalogram signals, and combining the correlation coefficients into a sleep characteristic; and based on the sleep characteristics, obtaining the sleep stage of the tested person in the period of time by adopting a classifier. Compared with the traditional linear method, more information can be provided, and the sleep staging accuracy is improved.

Description

A sleep staging method and system based on nonlinear interdependence

technical field

The invention belongs to the technical field of EEG signal processing, and in particular relates to a method and system for sleep staging based on nonlinear interdependence.

Background technique

The statements in this section merely provide background information related to the present invention and do not necessarily constitute prior art.

Detecting and evaluating sleep quality has important clinical significance and practical application value, and sleep state analysis is an important basis for evaluating sleep quality. In the 1930s, German psychiatrist Berger discovered that the electroencephalogram (EEG) activity of people during sleep and wakefulness showed different rhythms. In 1968, the R&K sleep staging standard was proposed, which divided the sleep process into wakefulness (W), rapid eye movement (REM) and non-rapid eye movement (NREM). The non-REM stage is further divided into S1, S2, S3 and S4 stages, therefore, the problem of sleep staging is a six-category problem. Traditional sleep staging relies on the naked eye of experts, which is a time-consuming process with subjective judgment probability. Therefore, the accuracy of traditional sleep analysis methods for sleep staging is not high.

In recent years, there have been more and more studies on sleep staging using classifiers to extract features based on physiological signals such as single-channel EEG, multi-channel EEG, ECG, OMG, EMG, and respiration. Among them, sleep staging based on EEG has the best effect, followed by ECG, and the effect of EMG and OMG is poor. However, the existing EEG-based sleep staging methods rely on a priori assumptions, and the accuracy of sleep staging is not high and the speed is slow.

SUMMARY OF THE INVENTION

In order to solve the technical problems existing in the above background technology, the present invention provides a method and system for sleep staging based on nonlinear interdependence, extracting nonlinear metrics as features and applying them to sleep detection. Compared with traditional linear methods, it can Provide more information and improve the accuracy of sleep staging.

In order to achieve the above object, the present invention adopts the following technical solutions:

A first aspect of the present invention provides a method for sleep staging based on nonlinear interdependence, comprising:

Obtain two EEG signals of the subject within a certain period of time;

Extract various nonlinear interdependencies between two EEG signals and the cross-correlation coefficient between two EEG signals, and combine them into sleep features;

Based on the sleep characteristics, a classifier is used to obtain the sleep stage of the subject during this period.

Further, the specific steps of extracting multiple nonlinear interdependencies between the two EEG signals and the cross-correlation coefficient between the two EEG signals are:

Reconstruct each EEG signal to obtain the reconstructed EEG signal;

Calculate the independent neighborhood distance and coupled neighborhood distance for each time point in the reconstructed EEG signal;

Based on the independent neighborhood distance and the coupled neighborhood distance at each time point in the two EEG signals, various nonlinear interdependencies between the two EEG signals and the cross-correlation coefficient between the two EEG signals are calculated .

Further, the independent neighborhood distance is the mean squared Euclidean distance from a certain time point in one type of EEG signal to the time point in k-neighborhood time points of this time point in the same type of EEG signal.

Further, the coupled neighborhood distance is the mean squared Euclidean distance from a certain time point in one kind of EEG signal to the time point in k neighborhoods of this time point in another kind of EEG signal.

Further, the multiple nonlinear interdependence degrees include a first nonlinear interdependence degree, and the calculation method of the first nonlinear interdependence degree is: calculating an independent neighbor of an EEG signal at each time point. The ratio of the domain distance to the coupled neighborhood distance is calculated, and the mean value of the ratio at all time points is calculated, and the mean value is used as the first nonlinear interdependence.

Further, the multiple nonlinear interdependence degrees include a second nonlinear interdependence degree, and the calculation method of the second nonlinear interdependence degree is: calculating an independent neighbor of an EEG signal at each time point. The ratio of the domain distance to the coupled neighborhood distance is calculated, and the log value of the ratio at each time point is calculated, taking the mean of the log values at all time points as the second nonlinear interdependence.

Further, the multiple nonlinear interdependence degrees include a third nonlinear interdependence degree, and the calculation method of the third nonlinear interdependence degree is: calculating to obtain an independent EEG signal at each time point. After the difference between the neighborhood distance and the coupled neighborhood distance, the ratio of the difference to the independent neighborhood distance at the same time point is calculated, and the average value of the ratio at all time points is taken as the third nonlinear interdependence.

A second aspect of the present invention provides a non-linear interdependence-based sleep staging system, comprising:

A signal acquisition module, which is configured to: acquire two kinds of EEG signals of the subject within a certain period of time;

a feature extraction module, configured to: extract multiple nonlinear interdependencies between two kinds of EEG signals and a cross-correlation coefficient between two kinds of EEG signals, and combine them into sleep features;

The classification module is configured to: based on the sleep feature, use a classifier to obtain the sleep stage of the subject within the period of time.

A third aspect of the present invention provides a computer-readable storage medium on which a computer program is stored, and when the program is executed by a processor, implements the above-mentioned method for sleep staging based on non-linear interdependence. step.

A fourth aspect of the present invention provides a computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, the processor implementing the above-mentioned one when executing the program Steps in a non-linear interdependence-based approach to sleep staging.

Compared with the prior art, the beneficial effects of the present invention are:

The present invention provides a method for sleep staging based on nonlinear interdependence, which applies nonlinear interdependence to sleep staging. Non-linear interdependence does not require extraction of other features and variables in EEG signals, and only requires Time series, the expression is relatively direct; the nonlinear interdependence measure is the coupling strength between dynamic systems. When the sleep period changes, the EEG signal changes drastically, reflecting strong nonlinear characteristics. At this time, the nonlinear dependence shows strong fluctuations; nonlinear interdependence uses the time series data of EEG signals, which is easy to obtain and monitor, and has better timeliness.

The present invention provides a sleep staging method based on nonlinear interdependence, which adopts a fuzzy logic classifier, which can identify prototypes from observation data and construct 0-order AnYa-type fuzzy rules; the meta-parameters obtained by the method It is directly derived from data and does not depend on a priori assumptions; and it can balance performance and computational efficiency according to the adjustment of computational complexity. At the same time, fuzzy logic classifiers also support different types of distance metrics, so they can be efficiently based on specific problems. to adjust.

Description of drawings

The accompanying drawings forming a part of the present invention are used to provide further understanding of the present invention, and the exemplary embodiments of the present invention and their descriptions are used to explain the present invention, and do not constitute an improper limitation of the present invention.

FIG. 1 is a flowchart of a sleep staging method based on nonlinear interdependence according to Embodiment 1 of the present invention.

Detailed ways

The present invention will be further described below with reference to the accompanying drawings and embodiments.

It should be noted that the following detailed description is exemplary and intended to provide further explanation of the invention. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

It should be noted that the terminology used herein is for the purpose of describing specific embodiments only, and is not intended to limit the exemplary embodiments according to the present invention. As used herein, unless the context clearly dictates otherwise, the singular is intended to include the plural as well, furthermore, it is to be understood that when the terms "comprising" and/or "including" are used in this specification, it indicates that There are features, steps, operations, devices, components and/or combinations thereof.

Example 1

This embodiment provides a method for sleep staging based on nonlinear interdependence, as shown in FIG. 1 , which specifically includes the following steps:

Step 1. Obtain two EEG signals of the subject within a certain period of time: the first EEG signal X′={x′ _n } and the second EEG signal Y′={y′ _n }, where n= 1, …, N.

Specifically: (101) Acquire two kinds of EEG signals of the subject, wherein different EEG signals are obtained by different EEG lead methods; Suppose, there are two EEG lead methods x and y, and the first EEG lead method is used. (102) Segment each EEG signal to obtain the EEG signal in each period of time; Specifically, according to the R&K rule, each EEG signal is divided into a segment of 30 seconds to obtain the first EEG signal X′={x′ _n } and the second EEG signal Y′={ y' _n }, where n=1,...,N.

Or, directly use two EEG lead methods to obtain two EEG signals of the subject within a period of time (30 seconds).

Therefore, each EEG signal X' or Y' corresponds to a sleep stage (W, REM, S1, S2, S3 or S4).

Step 2, extracting multiple nonlinear interdependencies between the two EEG signals and the cross-correlation coefficient between the two EEG signals, and combining them into a sleep feature.

(201) Reconstructing each EEG signal to obtain the reconstructed EEG signal.

Specifically, according to the two EEG leads x and y, the obtained time series X′={x′ _n } and Y′={y′ _n }, reconstruct the EEG signal, wherein each EEG signal in the reconstructed EEG signal is The delay vectors of the time points are x _n =(x' _n ,...,x' _n-(m-1)τ ) and y _n =(y' _n ,...,y' _n-(m-1)τ ) , n=1,...N, where n is the time point, x _n represents the delay vector of the nth time point in the first EEG signal, m is the embedding dimension, and τ represents the time delay.

Therefore, the reconstructed first EEG signal X={x _n }, and the reconstructed second EEG signal Y={y _n }.

(202) Calculate the independent neighborhood distance of the delay vector of each time point in the reconstructed EEG signal. The independent neighborhood distance is the mean squared Euclidean distance from a certain time point in a kind of EEG signal to the k-neighborhood time point of this time point in the same kind of EEG signal.

Specifically, based on the delay vector x _n of each time point in the reconstructed first EEG signal X, calculate a certain time point x _n in one EEG signal to the time point x in the same EEG signal The mean square Euclidean distance of the k neighborhoods of _n , that is, the independent neighborhood distance of each time point x _n in the first EEG signal is:

Similarly, for the delay vector y _{n at each time point in the reconstructed second EEG signal Y, the mean square Euclidean distance from the delay vector y n at} each time point to its k neighborhood can be calculated to obtain the second The independent neighborhood distance of the delay vector y _n at each time point in the EEG signal is

(203) Calculate the coupled neighborhood distance of the delay vector at each time point in the reconstructed EEG signal. The coupled neighborhood distance is the mean squared Euclidean distance from a certain time point in one kind of EEG signal to the k neighborhood time points of this time point in another kind of EEG signal.

For the delay vector x _n of each time point in the reconstructed first EEG signal X, calculate the delay vector x _n at each time point to another brain under the condition of the reconstructed second EEG signal Y The mean squared Euclidean distance of the k-neighborhood of this time point in the electrical signal, which is defined by replacing the nearest neighbor with the equal-time neighbor of the nearest neighbor of y, that is, each time point in the first EEG signal The coupled-neighborhood distance of the delay vector x _n :

Among them, rn _,j and sn _,j , j=1,...,k represent the time indices of the K nearest neighbors of x _n and y _n , respectively.

Similarly, for the delay vector _yn at each time point in the reconstructed second EEG signal Y, calculate the delay vector _yn at each time point to its value under the condition of the reconstructed second EEG signal X. The mean square Euclidean distance of the k neighborhoods, to obtain the coupled neighborhood distance of the delay vector y _n at each time point in the second EEG signal

(204) Based on the independent neighborhood distance and the coupled neighborhood distance of the delay vector at each time point in the two EEG signals, calculate various nonlinear interdependencies between the two EEG signals and the relationship between the two EEG signals The correlation coefficient between them. Among them, sleep characteristics (first sleep characteristics, second sleep characteristics, third sleep characteristics, fourth sleep characteristics, fifth sleep characteristics, sixth sleep characteristics and seventh sleep characteristics). The multiple nonlinear interdependencies include a first nonlinear interdependency S, a second nonlinear interdependence H, and a third nonlinear interdependence N, and the first nonlinear interdependence S corresponds to the first sleep feature and the second sleep feature, the second nonlinear interdependence H corresponds to the third sleep feature and the fourth sleep feature, the third nonlinear interdependence N corresponds to the fifth sleep feature and the sixth sleep feature, the two brains The cross-correlation coefficient between the electrical signals corresponds to the seventh sleep characteristic.

(a) The first nonlinear interdependence degree S, the calculation method of the first nonlinear interdependence degree is: calculating the ratio of the independent neighborhood distance and the coupled neighborhood distance of an EEG signal at each time point, and The average of the ratios at all time points is calculated and used as the first nonlinear interdependence.

那么如果这些系统是强相关的，有

如果他们是独立的，那么

因此，可以定义一个相互依赖的度量(第一睡眠特征)S^(k)(X∣Y)为If the lattice {x _n } of the reconstructed first EEG signal X has an average square radius

Then if these systems are strongly correlated, we have

If they are independent, then

Therefore, an interdependent measure (the first sleep feature) S ^(k) (X∣Y) can be defined as

有because

Have

0<S ^(k) (X∣Y)≤1

If the value of S ^(k) (X∣Y) approaches 0, then the relationship between XY is independent, and when the value of S ^(k) (X∣Y) approaches 1, it means that XY will reach the maximum value at the same time .

Similarly, the second sleep feature S ^(k) (Y∣X) can be obtained.

(b) The second nonlinear interdependence degree H, the calculation method of the second nonlinear interdependence degree is: calculate the ratio of the independent neighborhood distance and the coupled neighborhood distance of an EEG signal at each time point, and The logarithm of the ratio at each time point was calculated, and the mean of the log values at all time points was used as the second nonlinear interdependence.

Define another nonlinear interdependence metric (third sleep feature) H ^(k) (X∣Y) as

H ^(k) (X∣Y) has the value 0 if X and Y are completely independent, while it will be positive if proximity in Y also implies proximity of peer-to-peer time partners in X.

Similarly, the fourth sleep feature H ^(k) (Y∣X) can be obtained.

(c) The third nonlinear interdependence degree N, the calculation method of the third nonlinear interdependence degree is: calculate the difference between the independent neighborhood distance and the coupled neighborhood distance of an EEG signal at each time point Then, the ratio of the difference to the independent neighborhood distance at the same time point is calculated, and the average of the ratios at all time points is taken as the third nonlinear interdependence.

In previous studies of coupled chaotic systems, H is more robust to noise, but the disadvantage is that it is not normalized, therefore, a new measure (the fifth sleep feature) N(X∣Y) is proposed, which is normalized and more robust than S.

Similarly, the sixth sleep feature N ^(k) (Y∣X) can be obtained.

Generally speaking, S(X∣Y), H(X∣Y), N(X∣Y) are not equal to S(Y∣X), H(Y∣X), N(Y∣X). The asymmetry of S, H, N is the main advantage of other nonlinear metrics for mutual information and phase synchronization.

与σ_x表示{x_n}的均值与方差，

与σ_y表示{y_n}的均值与方差，τ表示时滞，且有c_xy＝c_yx。(d) Use S(X∣Y), H(X∣Y), N(X∣Y) and S(Y∣X), H(Y∣X), N(Y∣X) as features to extract, In addition, coupled with the cross-correlation function (the seventh sleep feature) c _xy , the cross-correlation function is the most common measure between two EEG signals, here

and σ _x represents the mean and variance of {x _n },

and σ _y represents the mean and variance of {y _n }, τ represents the time delay, and c _xy = _cyx .

To sum up, the first sleep feature, the second sleep feature, the third sleep feature, the fourth sleep feature, the fifth sleep feature, the sixth sleep feature, and the seventh sleep feature together constitute the sleep feature of the subject within a certain period of time.

Step 3. Based on the sleep characteristics, a classifier is used to obtain the sleep stage of the subject within the period of time.

As an embodiment, the classifier adopts a fuzzy logic classifier.

When the set of sleep characteristics of the subject within a certain period of time is reached, it is represented by a, and its emission intensity is calculated, and the label at the maximum value is selected as the final result:

label = arg max(λ ^M (a))

Among them, M={W, REM, S1, S2, S3, S4}, a is the test sample, P is the final prototype formed by each type, P is divided into six types, and each type of sleep stage produces a prototype set. The sleep stage is divided into 6 categories. For each category, there are six types of prototypes, which are {W, REM, S1, S2, S3, S4}. When a test sample arrives, calculate the difference between it and the six types of prototypes. The attack intensity of a certain type of prototype is not only one, so the intensity in this type is also different, and the maximum emission intensity in this type is taken as the representative of the test sample and the prototype, so there are six kinds of representatives, and then in the The maximum value of the six representatives is taken as its final label, so the maximum value is taken twice. First take the maximum value in a certain category, and then compare the maximum value of the six categories.

Fuzzy logic classifiers use nonparametric EDA quantities to objectively reveal the integrated nature and mutual distribution of data. Specifically, the following three EDA quantities are used:

Assuming that there is a training sample set A={a ₁ ,a ₂ ,..., _ak }, the only data sample set corresponding to the training sample set A is U={u ₁ ,u ₂ ,...,u _Uk }:

① Cumulative proximity, the cumulative proximity of the data sample a _i (that is, the sleep characteristics of a subject in a certain period of time) is expressed as:

Among them, d(a _i , a _j ) represents the distance between a _i , a _j .

②The single-modal density of the data sample a _i :

③ Multimodal density of the unique data sample _ui :

Among them, _ui is the unique data sample, and fi is the frequency of the unique data sample ui. For example, for the training sample set A={1,1,2,3,4,4}, then its only data sample set is U={1,2,3,4}, and the corresponding frequencies fi are 1/3 respectively , 1/6, 1/6, 1/3.

The recursive form of computation of nonparametric EDA quantities plays an important role in the processing of the data. The classifier can use three classification distances for classification, namely: Mahalanobis distance, Euclidean distance, and cosine similarity. Under these three distances, fast computation can be achieved by saving key meta-parameters in memory, which also ensures the high efficiency of this method.

As an implementation manner, the fuzzy logic classifier is trained using two training methods: offline training and online training:

(1) Offline training

对应的每一个唯一数据样本w_i的多模态密度

并将

进行排列进列表{w}。之后选出列表{w}中的局部极大值，之后将局部极大值存入{p}₀。以S1类的训练数据为例。First, calculate the training sample set

The corresponding multimodal density of each unique data sample w _i

and will

Arrange into list {w}. Then the local maxima in the list {w} are selected, and then the local maxima are stored in {p} ₀ . Take the training data of class S1 as an example.

THEN( _wi ∈{p} ₀ )

After that, each element in {p} ₀ absorbs the nearest other data samples to form a data cloud. Wp is the prototype of the data cloud formed after the p-th element in {p} ₀ adsorbs the nearest data sample, that is, the prototype of the p-th data cloud:

由粒度(L)决定。一般来说，L越大分类越细。Then find the center of the p-th data cloud, and form a neighborhood with the centers of other data clouds according to the average radius. Specifically, if the distance between the center of a data cloud and the center of the p-th data cloud is less than the average radius, the data Cloud is the neighborhood of the p-th data cloud. average radius

Determined by particle size (L). Generally speaking, the larger the L, the finer the classification.

If the multimodal density of the pth data cloud center is greater than the multimodality of all data cloud centers in the neighborhood of the pth data cloud, then the prototype of the pth data cloud is confirmed as a prototype, and the The prototype of the p-th data cloud is assigned to {p} ^S1 .

(2) Online training

之后计算该新的训练数据单模态密度。如果其单模态密度大于已有原型的最大的单模态密度或小于已有原型的最小单模态密度，那么它将成为一个新的原型；When a new training data arrives, first update the average radius

This new training data unimodal density is then calculated. If its single-modal density is greater than the maximum single-modal density of the existing prototype or less than the minimum single-modal density of the existing prototype, then it will become a new prototype;

进行比较。如果大于平均距离，那么它也将成为一个新的原型。If it does not match, then calculate its distance from the nearest prototype and compare it with the average radius

Compare. If it is greater than the average distance, then it will also become a new prototype.

When it becomes a new prototype, the parameters of the classifier are updated. If it cannot be called a new archetype, then it will be relegated to the most recent archetype, and the parameters will be updated.

The steps of prototype formation are repeated six times, and six types of prototypes are obtained, which are represented as {p} ^REM , {p} ^S1 , {p} ^s2 and so on.

Step 4. After obtaining the sleep stage of the subject in multiple consecutive time periods, perform post-processing: for the six stages of sleep stages, take the classification of each stage to interpret, such as for the wake-up stage, divide the results into wake-up stages. Period and non-awake period, make a judgment and get the result. The six stages of each object were judged separately, and the average was taken as the final result. Specifically, post-processing is performed on the sleep stage in h time periods. It is assumed that from the first time period to the h1th time period, there are more than a threshold time period in a certain sleep stage, and the first time period is in a certain sleep stage. and the h1 th time period is also in the sleep stage, the time period from the first time period to the h1 th time period that is not in the sleep stage is modified to be in the sleep stage. For example, post-processing is performed on the sleep stages in 400 time periods, of which the first to 100th periods are S1, and the others are other stages such as S2, S3 or REM. When the judgment is made and 400 pieces of results are obtained, the result is judged for S1, and it becomes two categories - S1 category and non-S1 category. Paragraphs 1 to 100 should be judged as S1, and 101 to 400 should be judged as non-S1. If the 20th segment is judged as S1, then it is correct; if the 200th segment is judged as S1, it is wrong; if the 300th segment is judged as non-S1, then it is correct. Interpret these 400 segments and get the result of S1.

The method of the present invention was validated on a sleep database provided by St Vincent's University Hospital, Dublin. The database contains a variety of physiological information records such as EEG signals, ECG signals, etc. Among them, C3-A2 and C4-A1 are two EEG signals, so these two leads are used as samples to extract the nonlinear interdependence between them. In this experiment, time delay τ=2, embedding dimension m=10, neighborhood K=10, and Taylor correction T=50 are chosen. After the selection and experimental testing of three classification distances, granularity sizes and training set online and offline training ratios, it is obtained that the correct rate is the highest when cosine similarity is selected for the classification distance, the granularity is 12, and the training set is all used for offline training.

From the experimental results in Table 1, it can be seen that in the test of five subjects, the average accuracy rate reached 81%, which is better than other methods.

Table 1 Experimental results

Based on nonlinear dynamics, the different leads that extract EEG signals during sleep can be regarded as several nonlinear dynamic systems; the various connections and differences between multi-channel EEG signal leads can be regarded as several dynamics The coupling relationship between systems; due to the continuous change of EEG signals between different sleep stages, the coupling relationship between different dynamic systems is also constantly changing; therefore, the continuous changes during sleep are comprehensively reflected in the relationship between two EEG leads The change of the coupling relationship, the present invention adopts the method of nonlinear dynamics to study the sleep period. The present invention performs sleep staging by extracting features from multi-channel EEG signals. Since many features of EEG signals cannot be generated by linear models, the present invention extracts nonlinear metrics as features and applies it to sleep detection. Linear methods provide more information.

The present invention adopts a fuzzy logic classifier, which does not rely on any a priori assumptions, but only relies on the internal relationship between data to identify and classify, and only saves key meta-parameters in the memory. The problem of slow calculation speed of other methods is solved; the present invention also has three classification distances for flexible selection in actual situations: Mahalanobis distance, Euclidean distance, and cosine similarity; Overcoming the shortcomings of low accuracy of sleep staging in the past, the speed is faster ; And the computational complexity can be flexibly adjusted: it can not only obtain a higher detection rate, but also avoid over-fitting caused by over-learning of the training set; by flexibly adjusting the parameters, excellent results can be obtained in different situations.

Embodiment 2

This embodiment provides a sleep staging system based on nonlinear interdependence, which specifically includes the following modules:

A signal acquisition module, which is configured to: acquire two kinds of EEG signals of the subject within a certain period of time;

a feature extraction module, configured to: extract multiple nonlinear interdependencies between two kinds of EEG signals and a cross-correlation coefficient between two kinds of EEG signals, and combine them into sleep features;

The classification module is configured to: based on the sleep feature, use a classifier to obtain the sleep stage of the subject within the period of time.

It should be noted here that each module in this embodiment corresponds to each step in Embodiment 1 one by one, and the specific implementation process thereof is the same, which is not repeated here.

Embodiment 3

This embodiment provides a computer-readable storage medium on which a computer program is stored, and when the program is executed by a processor, implements the steps in the non-linear interdependence-based sleep staging method described in the first embodiment above. step.

Embodiment 4

This embodiment provides a computer device, including a memory, a processor, and a computer program stored in the memory and running on the processor, where the processor implements the one described in the first embodiment when the processor executes the program. Steps in a non-linear interdependence-based approach to sleep staging.

As will be appreciated by one skilled in the art, embodiments of the present invention may be provided as a method, system, or computer program product. Accordingly, the invention may take the form of a hardware embodiment, a software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product embodied on one or more computer-usable storage media having computer-usable program code embodied therein, including but not limited to disk storage, optical storage, and the like.

The present invention is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each flow and/or block in the flowchart illustrations and/or block diagrams, and combinations of flows and/or blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to the processor of a general purpose computer, special purpose computer, embedded processor or other programmable data processing device to produce a machine such that the instructions executed by the processor of the computer or other programmable data processing device produce Means for implementing the functions specified in a flow or flow of a flowchart and/or a block or blocks of a block diagram.

These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory result in an article of manufacture comprising instruction means, the instructions The apparatus implements the functions specified in the flow or flow of the flowcharts and/or the block or blocks of the block diagrams.

These computer program instructions can also be loaded on a computer or other programmable data processing device to cause a series of operational steps to be performed on the computer or other programmable device to produce a computer-implemented process such that The instructions provide steps for implementing the functions specified in the flow or blocks of the flowcharts and/or the block or blocks of the block diagrams.

Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be implemented by instructing relevant hardware through a computer program, and the program can be stored in a computer-readable storage medium. During execution, the processes of the embodiments of the above-mentioned methods may be included. The storage medium may be a magnetic disk, an optical disk, a read-only memory (Read-Only Memory, ROM), or a random access memory (Random Access Memory, RAM) or the like.

The above descriptions are only preferred embodiments of the present invention, and are not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and changes. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims (10)

1. A sleep staging method based on nonlinear interdependencies, comprising:

acquiring two electroencephalogram signals of a tested person within a certain period of time;

extracting various nonlinear interdependencies between the two electroencephalogram signals and cross-correlation coefficients between the two electroencephalogram signals, and combining the correlation coefficients into a sleep characteristic;

and based on the sleep characteristics, obtaining the sleep stage of the tested person in the period of time by adopting a classifier.

2. The sleep stage classification method based on nonlinear interdependence as claimed in claim 1, wherein the specific steps of extracting various nonlinear interdependences between two electroencephalogram signals and cross-correlation coefficients between two electroencephalogram signals are as follows:

reconstructing each electroencephalogram signal to obtain reconstructed electroencephalogram signals;

calculating the independent neighborhood distance and the coupling neighborhood distance of each time point in the reconstructed electroencephalogram signal;

based on the independent neighborhood distance and the coupling neighborhood distance of each time point in the two electroencephalogram signals, various nonlinear interdependencies between the two electroencephalogram signals and cross-correlation coefficients between the two electroencephalogram signals are calculated.

3. The method of claim 2, wherein the independent neighborhood distance is the mean squared euclidean distance from a time point in one electroencephalogram signal to a time point in k neighborhood of the time point in the same electroencephalogram signal.

4. The method of claim 2, wherein the coupling neighborhood distance is the mean squared euclidean distance from a time point in one brain electrical signal to a time point k neighborhood of the time point in another brain electrical signal.

5. The sleep staging method based on nonlinear interdependencies as recited in claim 2, wherein the plurality of nonlinear interdependencies include a first nonlinear interdependency calculated by: and calculating the ratio of the independent neighborhood distance to the coupling neighborhood distance of the electroencephalogram signal at each time point, calculating the mean value of the ratio at all the time points, and taking the mean value as a first nonlinear interdependence degree.

6. The sleep staging method based on nonlinear interdependencies as recited in claim 2, wherein the plurality of nonlinear interdependencies includes a second nonlinear interdependency calculated by: and calculating the ratio of the independent neighborhood distance to the coupling neighborhood distance of the electroencephalogram signal at each time point, calculating the logarithm value of the ratio at each time point, and taking the mean value of the logarithm values at all the time points as a second nonlinear interdependence.

7. The sleep staging method based on nonlinear interdependencies as recited in claim 2, wherein the plurality of nonlinear interdependencies includes a third nonlinear interdependency calculated by: calculating to obtain a difference value between the independent neighborhood distance and the coupling neighborhood distance of the electroencephalogram signal at each time point, calculating a ratio of the difference value to the independent neighborhood distance at the same time point, and taking the mean value of the ratios at all time points as a third nonlinear interdependence.

8. A sleep staging system based on non-linear interdependencies, comprising:

a signal acquisition module configured to: acquiring two electroencephalogram signals of a tested person within a certain period of time;

a feature extraction module configured to: extracting various nonlinear interdependencies between the two electroencephalogram signals and cross-correlation coefficients between the two electroencephalogram signals, and combining the correlation coefficients into a sleep characteristic;

a classification module configured to: and based on the sleep characteristics, obtaining the sleep stage of the tested person in the period of time by adopting a classifier.

9. A computer-readable storage medium, on which a computer program is stored which, when being executed by a processor, carries out the steps of a non-linear interdependence based sleep staging method as claimed in any one of the claims 1-7.

10. A computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor when executing the program implements the steps in a non-linear interdependence based sleep staging method as claimed in any one of claims 1-7.

Images