CN106936526A - A kind of contactless sleep stage device and method based on channel condition information - Google Patents

Abstract

The present invention provides a non-contact sleep staging device and method based on channel state information, which relates to the technical field of sleep monitoring, utilizes the inherent characteristics of channel state information such as good time stability and strong sensitivity to actions, and is used in an indoor environment. Ordinary laptops and home routers collect channel state information, apply signal processing and wavelet analysis methods to extract respiratory signals, use hidden Markov models to describe the corresponding relationship between respiratory characteristics and sleep stages, and realize non-contact sleep state perception in daily environments. By using the method proposed by the present invention, the average error of respiration rate identification can be realized to be about 2bpm, and the accuracy rate of sleep stage identification is 73.52%, which meets the needs of daily family sleep monitoring.

Description

A non-contact sleep staging device and method based on channel state information

technical field

The invention relates to the technical field of sleep monitoring, in particular to a non-contact sleep staging device and method based on channel state information.

Background technique

Sleep is an essential physiological activity of the human body. It is an important and complex physiological phenomenon, which occupies about one-third of the time in life. Sleep is the process of the body's self-repair and improvement, and plays an important regulatory role in maintaining physical and mental health.

Sleep staging is to divide the sleep process into different stages according to the different changes of the human body's physiological signals during sleep. In human sleep, there are about 4 to 6 sleep cycles in one night, which are interconnected and repeated, and each sleep stage has its own specific physiological and behavioral characteristics. According to the different characteristics of the EEG, sleep is mainly divided into non-rapid eye movement (Non-rapid eye movement, NREM) and rapid eye movement (Rapid eye movement, REM), and the NREM period is divided into two periods, light sleep and deep sleep. Light sleep is characterized by shallow breathing, body muscles remain relaxed, and no significant eye movements. Deep sleep is characterized by deep, even and regular breathing without noticeable eye movements. The REM period is characterized by slightly faster and irregular breathing, rapid eye movement, and elevated blood pressure, body temperature, and heart rate.

Channel state information (Channel state information, CSI) is the channel attribute of the communication link, which describes the attenuation factor of the signal on each transmission path, that is, the value of each element in the channel gain matrix H, such as signal scattering, environment Attenuation, distance attenuation and other information. Channel state information has the advantages of good time stability and strong sensitivity to motion, and is suitable for use in the field of non-contact sensing.

At present, the equipment for contact sleep state detection mainly includes medical polysomnography (PSG), brain wave detection belt, etc., and their common feature is that users need to wear additional equipment (electrodes, headbands, etc.), which will affect natural sleep to a certain extent. level of intrusion, unsuitable for continuous monitoring. Devices for non-contact sleep state detection, such as sleep monitoring mattresses, are expensive and not suitable for home use; smart wristbands are low in cost but not high in accuracy.

Contents of the invention

Embodiments of the present invention provide a non-contact sleep staging device and method based on channel state information to solve problems existing in the prior art.

A non-contact sleep staging device based on channel state information, said device includes a notebook computer, a wireless router and an external antenna, said notebook computer is equipped with a wireless network card and CSI Tool software, said wireless network card, an external antenna and a wireless router Sequential communication connection, the notebook computer utilizes CSI Tool to collect channel state information and save;

The notebook computer has a filter module, a correlation elimination module, a respiratory signal extraction module and a sleep staging module;

The filtering module is used to filter and denoise the channel state information by using a Hampel filter to obtain denoised channel state information;

The correlation elimination module is used to use the principal component analysis method to eliminate the correlation between different channels in the denoised channel state information, and only retain the most important component of the CSI change to obtain the principal component of the channel state information;

The respiratory signal extraction module is used to use sym8 to perform multi-scale decomposition of the principal components of the channel state information, and use the eighth layer approximation coefficient to reconstruct the signal to obtain the respiratory signal;

The sleep staging module includes a feature extraction submodule, a feature subset selection submodule and a sleep stage identification submodule;

The feature extraction submodule is used to extract respiratory features from the respiratory signal, and the extracted features include time domain features, frequency domain features and nonlinear features;

The feature subset selection submodule is used to use information gain to select the optimal feature subset in the extracted breath features, so as to reduce the feature dimension;

The sleep stage recognition submodule is used to construct a sleep stage recognition model using a hidden Markov model.

Preferably, Ubuntu operating system is installed in the notebook computer, and the model of the wireless network card is Intel5300.

Preferably, the feature extraction sub-module analyzes the continuous breathing intervals and extracts features, and the time domain features in the extracted features include mean value, variance, maximum value, minimum value, standard deviation, root mean square of difference, standard difference Difference and coefficient of variation; frequency domain features include the total energy of the frequency band and the energy ratio of the low frequency band to the high frequency band; nonlinear features include sample entropy.

The present invention also provides a non-contact sleep staging method based on channel state information, the method comprising:

Use CSI Tool to collect channel status information and save it;

Use the Hampel filter to filter and denoise the channel state information to obtain the denoised channel state information;

Use the principal component analysis method to eliminate the correlation between different channels in the denoised channel state information, and only keep the most important component of the CSI change to obtain the principal component of the channel state information;

Use sym8 to perform multi-scale decomposition of the principal components of channel state information, and use the eighth layer approximation coefficient to reconstruct the signal to obtain the respiratory signal;

Breathing features are extracted from the breathing signal, and the extracted features include time-domain features, frequency-domain features and nonlinear features;

Use information gain to select the optimal feature subset in the extracted breath features to reduce the feature dimension;

Building a sleep stage recognition model using hidden Markov models.

Preferably, when performing respiratory feature extraction on the respiratory signal, it is to analyze the continuous breathing interval and extract features. The time domain features in the extracted features include mean value, variance, maximum value, minimum value, standard deviation, root mean square of difference, Difference standard deviation and coefficient of variation; frequency domain features include the total energy of the frequency band and the energy ratio of the low frequency band to the high frequency band; nonlinear features include sample entropy.

Preferably, using a hidden Markov model to construct a sleep stage identification model specifically includes:

Establish a respiratory feature set G={mean value, variance, maximum value, minimum value, standard deviation, root mean square of difference, standard deviation of difference, coefficient of variation, total energy of frequency band, energy ratio of low frequency band and high frequency band, sample entropy} ;

Establish a state set S={S ₁ , S ₂ , S ₃ , S ₄ }, where S ₁ , S ₂ , S ₃ and S ₄ are sleep stages, that is, S ₁ is the awake period, S ₂ is the rapid eye movement period, S ₃ is light sleep period, S ₄ is deep sleep period;

Obtain the initial state probability distribution π={π ₁ , π ₂ , π ₃ , π ₄ }, where the value is the proportion of each sleep stage provided by the medical authority, π ₁ =0.07, π ₂ =0.23, π ₃ =0.55, π ₄ =0.15;

Determine the observation sequence O={o ₁ , o ₂ ,...o _i ,...,o _t } according to the respiratory feature set G, where o _i is the respiratory feature set G;

Get state sequence q = {q ₁ , q ₂ , ..., q _i , ..., q _t }, where q _i ∈ {S ₁ , S ₂ , S ₃ , S ₄ };

Use the training set data to perform probability statistics on sleep stage transitions, and obtain the state transition probability matrix A={a _ij }, where a _ij =P(q _t+1 =S _j │q _t =S _i ), 1≤i,j ≤4;

Perform probability statistics on the respiratory characteristics corresponding to sleep stages, and obtain the observation probability distribution matrix B={b _j (k)}, b _j (k)=P(o _t =v _k │q _t =S _j ), 1≤ j≤4, v _k is the actual value of the tth feature o _t of the observation sequence 0;

Establish a hidden Markov model λ=(π, A, B);

Using the Viterbi algorithm in the decoding process of the hidden Markov model, according to the observation sequence O and the hidden Markov λ=(π, A, B), the most likely hidden state sequence, that is, the sleep stage sequence, is solved.

A non-contact sleep staging device and method based on channel state information provided by the present invention utilizes the inherent characteristics of channel state information such as good time stability and strong sensitivity to actions, by using ordinary notebook computers and home routers in indoor environments Collect channel state information, apply signal processing and wavelet analysis methods to extract respiratory signals, use hidden Markov model to describe the corresponding relationship between respiratory characteristics and sleep stages, and realize non-contact sleep state perception in daily environments. By using the method proposed by the present invention, the average error of respiration rate identification can be realized to be about 2bpm, and the accuracy rate of sleep stage identification is 73.52%, which meets the needs of daily family sleep monitoring.

Description of drawings

In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings that need to be used in the description of the embodiments or the prior art. Obviously, the accompanying drawings in the following description are only These are some embodiments of the present invention. Those skilled in the art can also obtain other drawings based on these drawings without creative work.

FIG. 1 is a functional block diagram of a non-contact sleep staging device based on channel state information provided by an embodiment of the present invention;

Fig. 2 is a flow chart of the staging method of the non-contact sleep staging device in Fig. 1 .

detailed description

The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only some, not all, embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by persons of ordinary skill in the art without making creative efforts belong to the protection scope of the present invention.

The present invention provides a non-contact sleep staging device based on channel state information. The device includes a notebook computer 100, a wireless router 110 and an external antenna 120. The notebook computer 100 is equipped with a wireless network card and CSI Tool software. The wireless network card, The external antenna 120 and the wireless router 110 are sequentially connected in communication. In this embodiment, the Ubuntu operating system is installed in the notebook computer 100, the model of the wireless network card is Intel 5300, and the wireless router 110 is a TP-Link wireless router. The wireless router 110 serves as the signal transmitting source AP, the notebook computer 100 receives the wireless signal through the wireless network card, and the external antenna 120 serves as the signal receiver MP. Each pair of AP and MP forms a link, and the notebook computer 100 uses the CSI Tool to collect the wireless signal at a frequency of 100Hz. The channel state information is stored in the notebook computer 100.

The notebook computer 100 has a filtering module 101 , a correlation elimination module 102 , a respiratory signal extraction module 103 and a sleep staging module 104 .

The channel state information is easily affected by electromagnetic interference to generate mutation signals, and the mutation signals have the characteristics of sparse distribution. The filtering module 101 is used to filter and denoise the channel state information stored in the notebook computer 100 by using a Hampel filter, and obtain the denoised channel state information.

The channel state information on each subcarrier and each pair of AP and MP has correlation, and the correlation removal module 102 is used to use the principal component analysis method to remove the correlation between different channels in the denoised channel state information, and only retain the CSI The most important component of the change is to obtain the principal component of the channel state information, thereby significantly reducing the amount of calculation.

The breathing signal extraction module 103 is used to use sym8 to perform multi-scale decomposition on the principal components of the channel state information, and use the eighth layer approximation coefficients to reconstruct the signal to obtain the breathing signal.

The sleep staging module 104 includes a feature extraction submodule 1040 , a feature subset selection submodule 1041 and a sleep stage identification submodule 1042 .

The feature extraction sub-module 1040 is used to extract respiratory features from the respiratory signal, and the extracted features include time domain features, frequency domain features and nonlinear features. In this embodiment, the feature extraction sub-module 1040 analyzes the continuous respiration interval (RR interval) and extracts features. The time domain features in the extracted features include mean (Mean), variance (Var), maximum value (Max), minimum value (Min), standard deviation (SDNN), root mean square difference (RMSSD), difference standard deviation ( SDSD) and coefficient of variation (CV); frequency domain features include total frequency band energy (TF) and low frequency band to high frequency band energy ratio (LF/HF); nonlinear features include sample entropy.

The feature subset selection sub-module 1041 is used to select an optimal feature subset from the extracted respiratory features by using information gain, so as to reduce the feature dimension.

The sleep stage identification sub-module 1042 is used to construct a sleep stage identification model using a hidden Markov model. Specifically, the sleep stage recognition submodule 1042 constructs a sleep stage recognition model by the following method:

Establish a respiratory feature set G={mean value, variance, maximum value, minimum value, standard deviation, root mean square of difference, standard deviation of difference, coefficient of variation, total energy of frequency band, energy ratio of low frequency band and high frequency band, sample entropy} ;

Establish a state set S={S ₁ , S ₂ , S ₃ , S ₄ }, where S ₁ , S ₂ , S ₃ and S ₄ are sleep stages, that is, S ₁ is the awake period, S ₂ is the rapid eye movement period, S ₃ is a light sleep period, and S ₄ is a deep sleep period.

Obtain the initial state probability distribution π={π ₁ , π ₂ , π ₃ , π ₄ }, where the value is the proportion of each sleep stage provided by the medical authority, π ₁ =0.07, π ₂ =0.23, π ₃ =0.55, π ₄ =0.15.

Determine the observation sequence O={o ₁ , o ₂ ,...o _i ,...,o _t } according to the respiratory feature set G, where o _i is the respiratory feature set G;

Get state sequence q = {q ₁ , q ₂ , ..., q _i , ..., q _t }, where q _i ∈ {S ₁ , S ₂ , S ₃ , S ₄ };

Use the training set data to perform probability statistics on sleep stage transitions, and obtain the state transition probability matrix A={a _ij }, where a _ij =P(q _t+1 =S _j │q _t =S _i ), 1≤i,j ≤4;

Perform probability statistics on the respiratory characteristics corresponding to sleep stages, and obtain the observation probability distribution matrix B={b _j (k)}, b _j (k)=P(o _t =v _k │q _t =S _j ), 1≤ j≤4, v _k is the actual value of the tth feature o _t of the observation sequence 0;

Establish a hidden Markov model λ=(π, A, B);

Using the Viterbi algorithm in the decoding process of the hidden Markov model, according to the observation sequence O and the hidden Markov λ=(π, A, B), the most likely hidden state sequence, that is, the sleep stage sequence, is solved.

Based on the same inventive concept, the present invention also provides a non-contact sleep staging method based on channel state information. For the implementation of the method, refer to the implementation of the above-mentioned device, and repeated descriptions will not be repeated.

Step 200, using the CSI Tool to collect channel state information at a frequency of 100 Hz and saving it to the notebook computer 100;

Step 210, using a Hampel filter to filter and denoise the channel state information to obtain denoised channel state information;

Step 220, use the principal component analysis method to eliminate the correlation between different channels in the denoised channel state information, keep only the most important components of CSI changes, and obtain the principal components of the channel state information;

Step 230, using sym8 to perform multi-scale decomposition on the principal components of the channel state information, and using the eighth layer approximation coefficients to reconstruct the signal to obtain the respiratory signal;

Step 240, perform breathing feature extraction on the respiratory signal, and the extracted features include time-domain features, frequency-domain features and nonlinear features; specifically, analyze continuous breathing intervals, extract features, and the time-domain features in the extracted features include mean (Mean), variance (Var), maximum value (Max), minimum value (Min), standard deviation (SDNN), root mean square of difference (RMSSD), standard deviation of difference (SDSD) and coefficient of variation (CV); The frequency domain features include the total energy of the frequency band (TF) and the energy ratio of the low frequency band to the high frequency band (LF/HF); the nonlinear features include sample entropy.

Step 250, using information gain to select the optimal feature subset in the extracted breath features, so as to reduce the feature dimension;

Step 260, using the hidden Markov model to build a sleep stage recognition model. Specifically, this step includes:

Establish a respiratory feature set G={mean value, variance, maximum value, minimum value, standard deviation, root mean square of difference, standard deviation of difference, coefficient of variation, total energy of frequency band, energy ratio of low frequency band and high frequency band, sample entropy} ;

Establish a state set S={S ₁ , S ₂ , S ₃ , S ₄ }, where S ₁ , S ₂ , S ₃ and S ₄ are sleep stages, that is, S ₁ is the awake period, S ₂ is the rapid eye movement period, S ₃ is light sleep period, S ₄ is deep sleep period;

Obtain the initial state probability distribution π={π ₁ , π ₂ , π ₃ , π ₄ }, where the value is the proportion of each sleep stage provided by the medical authority, π ₁ =0.07, π ₂ =0.23, π ₃ =0.55, π ₄ =0.15;

Determine the observation sequence O={o ₁ , o ₂ ,...o _i ,...,o _t } according to the respiratory feature set G, where o _i is the respiratory feature set G;

Get state sequence q = {q ₁ , q ₂ , ..., q _i , ..., q _t }, where q _i ∈ {S ₁ , S ₂ , S ₃ , S ₄ };

Use the training set data to perform probability statistics on sleep stage transitions, and obtain the state transition probability matrix A={a _ij }, where a _ij =P(q _t+1 =S _j │q _t =S _i ), 1≤i,j ≤4;

Perform probability statistics on the respiratory characteristics corresponding to sleep stages, and obtain the observation probability distribution matrix B={b _j (k)}, b _j (k)=P(o _t =v _k │q _t =S _j ), 1≤ j≤4, v _k is the actual value of the tth feature o _t of the observation sequence 0;

Establish a hidden Markov model λ=(π, A, B);

Using the Viterbi algorithm in the decoding process of the hidden Markov model, according to the observation sequence O and the hidden Markov λ=(π, A, B), the most likely hidden state sequence, that is, the sleep stage sequence, is solved.

Those skilled in the art should understand that the embodiments of the present invention may be provided as methods, systems, or computer program products. Accordingly, the present invention can take the form of an entirely hardware embodiment, an entirely 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 (including but not limited to disk storage, CD-ROM, optical storage, etc.) having computer-usable program code embodied therein.

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 should be understood that each procedure and/or block in the flowchart and/or block diagram, and a combination of procedures and/or blocks in the flowchart and/or block diagram can be realized by computer program instructions. These computer program instructions may be provided to a general purpose computer, special purpose computer, embedded processor, or processor of other programmable data processing equipment to produce a machine such that the instructions executed by the processor of the computer or other programmable data processing equipment produce a An apparatus for realizing the functions specified in one or more procedures of the flowchart and/or one or more blocks of the 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 operate in a specific manner, such that the instructions stored in the computer-readable memory produce an article of manufacture comprising instruction means, the instructions The device realizes the function specified in one or more procedures of the flowchart and/or one or more blocks of the block diagram.

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

While preferred embodiments of the invention have been described, additional changes and modifications to these embodiments can be made by those skilled in the art once the basic inventive concept is appreciated. Therefore, it is intended that the appended claims be construed to cover the preferred embodiment as well as all changes and modifications which fall within the scope of the invention.

Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and equivalent technologies thereof, the present invention also intends to include these modifications and variations.

Claims (6)

1. A non-contact sleep staging device based on channel state information, characterized in that, the device includes a notebook computer, a wireless router and an external antenna, and the notebook computer is equipped with a wireless network card and CSI Tool software, and the wireless network card , the external antenna and the wireless router are sequentially communicated and connected, and the notebook computer utilizes the CSI Tool to collect channel state information and save it; The notebook computer has a filter module, a correlation elimination module, a respiratory signal extraction module and a sleep staging module; The filtering module is used to filter and denoise the channel state information by using a Hampel filter to obtain denoised channel state information; The correlation elimination module is used to use the principal component analysis method to eliminate the correlation between different channels in the denoised channel state information, and only retain the most important component of the CSI change to obtain the principal component of the channel state information; The respiratory signal extraction module is used to use sym8 to perform multi-scale decomposition of the principal components of the channel state information, and use the eighth layer approximation coefficient to reconstruct the signal to obtain the respiratory signal; The sleep staging module includes a feature extraction submodule, a feature subset selection submodule and a sleep stage identification submodule; The feature extraction submodule is used to extract respiratory features from the respiratory signal, and the extracted features include time domain features, frequency domain features and nonlinear features; The feature subset selection submodule is used to use information gain to select the optimal feature subset in the extracted breath features, so as to reduce the feature dimension; The sleep stage recognition submodule is used to construct a sleep stage recognition model using a hidden Markov model. 2. The device according to claim 1, wherein an Ubuntu operating system is installed in the notebook computer, and the model of the wireless network card is Intel 5300. 3. device as claimed in claim 1, is characterized in that, described feature extraction submodule analyzes continuous respiration interval, extracts feature, and time-domain feature comprises mean value, variance, maximum value, minimum value, Standard deviation, root mean square of difference, standard deviation of difference, and coefficient of variation; frequency domain features include the total energy of the frequency band and the energy ratio of the low frequency band to the high frequency band; nonlinear features include sample entropy. 4. A non-contact sleep staging method based on channel state information, characterized in that the method comprises: Use CSI Tool to collect channel status information and save it; Use the Hampel filter to filter and denoise the channel state information to obtain the denoised channel state information; Use the principal component analysis method to eliminate the correlation between different channels in the denoised channel state information, and only keep the most important component of the CSI change to obtain the principal component of the channel state information; Use sym8 to perform multi-scale decomposition of the principal components of channel state information, and use the eighth layer approximation coefficient to reconstruct the signal to obtain the respiratory signal; Breathing features are extracted from the breathing signal, and the extracted features include time-domain features, frequency-domain features and nonlinear features; Use information gain to select the optimal feature subset in the extracted breath features to reduce the feature dimension; Building a sleep stage recognition model using hidden Markov models. 5. method as claimed in claim 4, it is characterized in that, when respiratory signal is carried out breathing feature extraction, for continuous breathing interval is analyzed, feature extraction, time-domain feature comprises mean value, variance, maximum value, Minimum value, standard deviation, root mean square of difference, standard deviation of difference, and coefficient of variation; frequency domain features include the total energy of the frequency band and the energy ratio of the low frequency band to the high frequency band; nonlinear features include sample entropy. 6. The method according to claim 5, wherein using a hidden Markov model to construct a sleep stage identification model specifically comprises: Establish a respiratory feature set G={mean value, variance, maximum value, minimum value, standard deviation, root mean square of difference, standard deviation of difference, coefficient of variation, total energy of frequency band, energy ratio of low frequency band and high frequency band, sample entropy} ; Establish a state set S={S ₁ , S ₂ , S ₃ , S ₄ }, where S ₁ , S ₂ , S ₃ and S ₄ are sleep stages, that is, S ₁ is the awake period, S ₂ is the rapid eye movement period, S ₃ is light sleep period, S ₄ is deep sleep period; Obtain the initial state probability distribution π={π ₁ , π ₂ , π ₃ , π ₄ }, where the value is the proportion of each sleep stage provided by the medical authority, π ₁ =0.07, π ₂ =0.23, π ₃ =0.55, π ₄ =0.15; Determine the observation sequence O={o ₁ , o ₂ ,...o _i ,...,o _t } according to the respiratory feature set G, where o _i is the respiratory feature set G; Get state sequence q = {q ₁ , q ₂ , ..., q _i , ..., q _t }, where q _i ∈ {S ₁ , S ₂ , S ₃ , S ₄ }; Use the training set data to perform probability statistics on sleep stage transitions, and obtain the state transition probability matrix A={a _ij }, where a _ij =P(q _t+1 =S _j │q _t =S _i ), 1≤i,j ≤4; Perform probability statistics on the respiratory characteristics corresponding to sleep stages, and obtain the observation probability distribution matrix B={b _j (k)}, b _j (k)=P(o _t =v _k │q _t =S _j ), 1≤ j≤4, v _k is the actual value of the tth feature o _t of the observation sequence 0; Establish a hidden Markov model λ=(π, A, B); Using the Viterbi algorithm in the decoding process of the hidden Markov model, according to the observation sequence O and the hidden Markov λ=(π, A, B), the most likely hidden state sequence, that is, the sleep stage sequence, is solved.