Micro-awakening detection method, device, equipment and readable storage medium

Abstract

The present invention provides a micro-awakening detection method, device, equipment and readable storage medium, the method comprising: obtaining original data, dividing the original data into segments according to a preset time length to obtain a plurality of first data; denoising the first data to obtain denoised first data, extracting shallow features of the denoised first data to obtain shallow features; decomposing the first data using set empirical mode decomposition to obtain a modal component sequence; training a deep neural network model based on the shallow features corresponding to each of the first data and the modal component sequence to obtain a trained deep neural network model; and detecting the user's sleep state using the trained deep neural network model. Based on the trained deep neural network model, effective discrimination of the sleep state can be achieved by relying on only a single electrocardiogram signal, so the present invention has the advantages of low cost and high application value.

Description

Technical Field

[0001] The present invention relates to the field of sleep detection technology, and in particular to a micro-arousal detection method, device, equipment and readable storage medium. Background Art

[0002] Sleep arousal events are closely related to sleep disorders and are also the main symptoms of most patients with sleep disorders during sleep. Too frequent sleep awakenings can lead to fragmented sleep, resulting in a decline in the user's sleep quality, and are directly related to diseases such as insomnia and sleep disorders. In addition, the prevalence of sleep disorders is positively correlated with age. With the aging of the population, my country will face greater pressure in the diagnosis and treatment of sleep disorders in the future. Therefore, it is particularly important to study the detection method of sleep arousal events and scientifically evaluate sleep quality. Summary of the invention

[0003] The object of the present invention is to provide a micro-awakening detection method, device, equipment and readable storage medium to improve the above-mentioned problems.

[0004] In order to achieve the above objectives, the present application provides the following technical solutions:

[0005] On the one hand, an embodiment of the present application provides a method for detecting micro-awakening, the method comprising:

[0006] Acquire original data, the original data including an electrocardiogram signal of at least one subject in a sleeping state within a first time period, and divide the original data into segments according to a preset time length to obtain a plurality of first data;

[0007] Denoising the first data to obtain the denoised first data, extracting shallow features of the denoised first data to obtain shallow features; decomposing the first data using set empirical mode decomposition to obtain a modal component sequence;

[0008] Training a deep neural network model based on the shallow features corresponding to each of the first data and the modal component sequence to obtain a trained deep neural network model;

[0009] The trained deep neural network model is used to detect the user's sleeping state.

[0010] In a second aspect, an embodiment of the present application provides a micro-awakening detection device, which includes an acquisition module, a denoising module, a training module and a recognition module.

[0011] An acquisition module, used to acquire raw data, wherein the raw data includes an electrocardiogram signal of at least one subject in a sleeping state within a first time period, and the raw data is segmented according to a preset time length to obtain a plurality of first data;

[0012] A denoising module, configured to denoise the first data to obtain denoised first data, extract shallow features of the denoised first data to obtain shallow features; and decompose the first data using ensemble empirical mode decomposition to obtain a modal component sequence;

[0013] A training module, used for training a deep neural network model based on the shallow features corresponding to each of the first data and the modal component sequence to obtain a trained deep neural network model;

[0014] The recognition module is used to detect the user's sleeping state using the trained deep neural network model.

[0015] In a third aspect, an embodiment of the present application provides a micro-awakening detection device, the device comprising a memory and a processor. The memory is used to store a computer program; the processor is used to implement the steps of the above-mentioned micro-awakening detection method when executing the computer program.

[0016] In a fourth aspect, an embodiment of the present application provides a readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the steps of the above-mentioned micro-awakening detection method are implemented.

[0017] The beneficial effects of the present invention are:

[0018] 1. In the present invention, after the model is trained, the sleep state of this period of time can be detected based on a real-time collected ECG signal and the trained deep neural network model. That is to say, when performing real-time detection, only a single ECG signal is needed on the basis of the trained deep neural network model to achieve effective discrimination of the sleep state, so it has the advantages of low cost and high application value; at the same time, the local characteristics and global timing characteristics of the signal can be taken into account at the same time during detection, so the sleep state can be effectively detected; in addition, because the ECG signal is easy to collect, the method of the present invention can also be applied to micro-awakening detection in home scenes, which has great applicability.

[0019] 2. In the present invention, a comprehensive ECG denoising method based on median filtering and discrete wavelet transform is designed, which can filter common noises in ECG signals and effectively solve the problem that the model classification effect is easily disturbed in high-noise scenarios.

[0020] Other features and advantages of the present invention will be described in the following description, and partly become apparent from the description, or be understood by practicing the embodiments of the present invention. The purpose and other advantages of the present invention can be realized and obtained by the structures particularly pointed out in the written description, claims, and drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments are briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.

[0022] Figure 1 is a schematic flow chart of a micro-awakening detection method described in an embodiment of the present invention;

[0023] Figure 2 is a schematic structural diagram of a micro-awakening detection device according to an embodiment of the present invention;

[0024] Figure 3 is a schematic structural diagram of a micro-awakening detection device described in an embodiment of the present invention;

[0025] Markings in the figure: 701, acquisition module; 702, denoising module; 703, training module; 704, recognition module; 7021, filtering unit; 7022, first computing unit; 7023, denoising unit; 7024, second computing unit; 7025, correction unit; 7026, third computing unit; 70261, first computing subunit; 70262 recognition subunit; 70263, second computing subunit; 7027, decomposition unit; 7028, repetition unit; 7031, input unit; 7032, splicing unit; 7033, dimensionality reduction unit; 800, micro-awakening detection device; 801, processor; 802, memory; 803, multimedia component; 804, I / O interface; 805, communication component. DETAILED DESCRIPTION

[0026] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of the embodiments. The components of the embodiments of the present invention generally described and shown in the drawings herein can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the drawings is not intended to limit the scope of the claimed invention, but merely represents selected embodiments of the present invention. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0027] It should be noted that similar reference numerals or letters represent similar items in the following drawings, so once an item is defined in one drawing, it does not need to be further defined and explained in the subsequent drawings. At the same time, in the description of the present invention, the terms "first", "second", etc. are only used to distinguish the description and cannot be understood as indicating or implying relative importance.

[0028] Example 1

[0029] like Figure 1 As shown, this embodiment provides a micro-awakening detection method, which includes step S1, step S2, step S3 and step S4.

[0030] Step S1, obtaining original data, wherein the original data includes an electrocardiogram signal of at least one subject in a sleeping state within a first time period, and dividing the original data into segments according to a preset time length to obtain a plurality of first data;

[0031] In this step, the number of experimenters can be customized according to the needs of the user, preferably multiple, so as to improve the accuracy of the final recognition; the length of the first time period, the start time point and the end time point of the first time period can also be customized according to the needs of the user; in this embodiment, the length of the first time period can be the electrocardiogram signal from 12 o'clock in the evening to 7 o'clock in the morning, and then the electrocardiogram signal is segmented at intervals of 10 seconds, and multiple first data are obtained after segmentation;

[0032] Step S2, denoising the first data to obtain denoised first data, extracting shallow features of the denoised first data to obtain shallow features; decomposing the first data using set empirical mode decomposition to obtain a modal component sequence;

[0033] Considering that the first data may contain noise, a comprehensive denoising method combining median filtering and wavelet transform is used to remove common noise in the ECG. This method can ensure the accuracy of the data entering the model for training. The specific implementation steps of denoising the first data to obtain the denoised first data include step S21, step S22 and step S23.

[0034] Step S21, using a median filter to process the first data to obtain second data;

[0035] In this step, a median filter with a window size of 200ms is first designed to eliminate the R wave and P wave in the original signal, and then a median filter with a window size of 500ms is used to eliminate the T wave to obtain the second data;

[0036] Step S22, performing a difference process on the first data and the second data to obtain third data;

[0037] Step S23: Perform discrete wavelet transform denoising on the third data to obtain the denoised first data.

[0038] In this step, the third data is first decomposed by discrete wavelet transform, the wavelet coefficients of the frequency band greater than 45HZ are set to zero, and then the ECG signal is reconstructed by inverse wavelet transform to obtain the denoised first data; by combining the comprehensive ECG denoising method of median filtering and discrete wavelet transform, the baseline drift, power frequency interference, myoelectric interference and other noises in the ECG signal can be effectively filtered;

[0039] In this step, the shallow features of the first data after denoising are extracted, and the specific implementation steps of obtaining the shallow features include step S24, step S25 and step S26;

[0040] Step S24, using the Pan-Tompkins algorithm to process the denoised first data to obtain an R wave peak index in the electrocardiogram signal;

[0041] Step S25, correcting the R wave peak index in the electrocardiogram signal to obtain a corrected R wave peak index, wherein the corrected R wave peak index includes a plurality of index values; in this step, the index values ​​may be represented as X1, X2, X3, X4, ...;

[0042] Step S26, calculating the heart rate variability signal, the R wave peak sequence and the cardiogenic respiration signal according to the corrected R wave peak index. The specific implementation steps of this step include step S261, step S262 and step S263;

[0043] Step S261, performing difference calculation on two adjacent index values ​​to obtain a heart rate variability signal;

[0044] Continuing with the above example, this step can be understood as obtaining the heart rate variability signal through X2-X1, X3-X2, X4-X3, etc.;

[0045] Step S262, identifying each peak value of the R wave in the first data after denoising to obtain an R wave peak sequence;

[0046] Step S263: oversampling the R wave peak sequence to obtain the cardiogenic respiration signal.

[0047] Respiration signals play an important role in micro-arousal detection, but the collection of respiratory signals is relatively cumbersome and often easy to be missing. However, by analyzing the ECG signal, a signal similar to the respiratory signal can be obtained, that is, the cardiogenic respiratory signal;

[0048] Sleep-wake events are closely related to heart rate changes and respiratory movements. Extracting shallow features that can directly reflect heart rate and respiratory changes from ECG signals as feature supplements allows the network model to learn more features related to wake events. By fusion of multiple features, the number of effective signals is increased, avoiding the one-sidedness of traditional feature extraction.

[0049] In this step, the first data is decomposed by using the set empirical mode decomposition to obtain the modal component sequence. The specific implementation steps include step S27 and step S28;

[0050] Step S27, decomposition step: adding white noise of random normal distribution to the first data to obtain fourth data, performing empirical mode decomposition on the fourth data to obtain multiple modal components, arranging all the modal components according to frequency to obtain fifth data, wherein the fifth data includes the arranged modal components;

[0051] Step S28, repeat the decomposition step to obtain multiple fifth data, integrate and average all the modal components arranged at the Nth position in the fifth data to obtain sixth data, and form all the sixth data into a modal component sequence, where N is a positive integer.

[0052] This step can be understood as, for example, if three fifth data are obtained through step S27, then in this step, the modal component ranked first in the first fifth data, the modal component ranked first in the second fifth data, and the modal component ranked first in the third fifth data are integrated and averaged to obtain the sixth data. According to this logic, multiple sixth data can be obtained, and finally a modal component sequence is obtained;

[0053] After calculating the shallow features and the modal component sequence corresponding to each of the first data, the deep neural network model can be trained using them; the specific steps are step S3;

[0054] Step S3, training the deep neural network model based on the shallow features corresponding to each of the first data and the modal component sequence to obtain a trained deep neural network model; the specific implementation steps of this step include step S31, step S32 and step S33;

[0055] Step S31: during each training, each modal component in the modal component sequence is input into a one-dimensional residual convolutional network to extract local features of each modal component;

[0056] The modal component sequence obtained through the above steps has high dimension, large data volume, and unclear features related to arousal. If it is directly input into the long short-term memory recurrent neural network for training, the amount of calculation is large and the effect is not good. Therefore, this embodiment adopts a one-dimensional residual convolution network to learn the deep features of each modal component. The one-dimensional convolution can extract local features without losing the temporal features of the signal. Different convolution kernels are used to extract different convolution features, which can increase the number of effective signals.

[0057] The one-dimensional residual convolution network in this embodiment has two residual blocks, each of which consists of two one-dimensional convolutional layers and a BN layer on the left side, and one one-dimensional convolutional layer and a BN layer on the right side. The number of filters in the left convolution layer of the first residual block is 16, the convolution kernel size is 1x5, and the step size is 2, and the number of filters in the right convolution layer is 16, the convolution kernel size is 1x1, and the step size is 4. The number of filters in the left convolution layer of the second residual block is 32, the convolution kernel size is 1x3, and the step size is 2, and the number of filters in the right convolution layer is 32, the convolution kernel size is 1x1, and the step size is 4. The BN layer represents batch normalization, which normalizes the one-dimensional convolution output parameters so that the data of each layer can be passed within a valid range. After the two residual blocks, the output data of the residual blocks are reduced in dimension through a maximum pooling layer of size 1x2 while minimizing feature loss;

[0058] Step S32, concatenating the local features of each of the modal components with all of the shallow features, inputting the concatenated features into a long short-term memory recurrent neural network, and using the output of the long short-term memory recurrent neural network as the input of the attention model;

[0059] In this step, in order to improve the model effect, the three shallow features are fused with multiple groups of features extracted by one-dimensional residual convolution through Concatenate. Concatenate is a merge operation that can splice multiple time series signals. Before the Concatenate operation, each shallow feature needs to be resampled to the same length as the output feature of the one-dimensional residual convolution network to facilitate feature fusion operation;

[0060] In this step, after splicing, two layers of long short-term memory recurrent neural networks can be input for learning. The long short-term memory recurrent neural network can learn the changes of multiple signals at the same time and further integrate the temporal relationship between features. The number of neurons in the first layer of long short-term memory recurrent neural network can be set to 64, and the number of neurons in the second layer of long short-term memory recurrent neural network can be set to 128. The activation function is set to tanh. The learning process from coarse-grained to fine-grained can better capture the correlation of long-term dependent features. At the same time, in order to prevent overfitting, a Dropout operation will be introduced after each long short-term memory recurrent neural network to enhance the generalization of the model. The parameters of the two Dropouts are set to 0.5;

[0061] Since it is difficult for the LSTM RNN to highlight important information in a long time series, resulting in a decrease in the network's ability to learn time series information, this embodiment introduces an attention model Attention after the two-layer LSTM RNN. The Attention layer can pay more attention to important features in the sequence, highlight the subsequences with concentrated attention, and improve the network's learning ability;

[0062] Step S33, input the output of the attention model into two layers of fully connected networks in sequence, wherein the output of the attention model is reduced in dimension by using the first layer of fully connected network, and then input into the second layer of fully connected network after dimension reduction, and the binary classification operation is implemented by using the softmax function, and the final classification result is output to complete one training, and the classification result includes the awake state or the non-awakening state.

[0063] The first fully connected network has 32 neurons, which is used to reduce the dimension of the output of the Attention layer. The second fully connected network has 2 neurons, and uses the softmax function to implement binary classification and finally output the classification results.

[0064] During training, the shallow features corresponding to a first data and the modal component sequence can be used as the data included in a sample to construct a sample set. During the training process, in order to reduce the memory space occupied by the training data and speed up the model training, this embodiment is based on the idea of ​​exchanging time for space, and adopts batch training to complete the model training. Each time training, only one batch of data is loaded into the memory for iterative calculation. All sample data are divided into 20 batches, and one batch of data is loaded for training each time. When all batches are traversed, it means that one iteration is completed, and the final number of iterations is set to 200 times. This method greatly reduces memory overhead and prevents the program from being directly stuck when the memory space is insufficient when the data is too large. In each round of training, the cross entropy function is used to define the model loss, and the Adam optimizer is used to update the parameters.

[0065] Step S4: Detect the user's sleep state using the trained deep neural network model.

[0066] In this embodiment, after the model is trained, the sleep state during this period of time can be identified based on a real-time collected ECG signal and the trained deep neural network model. That is to say, when performing real-time identification, only a single ECG signal is needed based on the trained deep neural network model to effectively distinguish the sleep state. Compared with current research based on multimodal signals or EEG signals, ECG signals are easy to collect and do not rely on professional medical equipment or environment, and have higher application value. Therefore, the method in this embodiment is convenient for sleep quality detection applications in home scenarios.

[0067] Example 2

[0068] like Figure 2 As shown, this embodiment provides a micro-awakening detection device, which includes an acquisition module 701, a denoising module 702, a training module 703 and a recognition module 704.

[0069] An acquisition module 701 is used to acquire raw data, wherein the raw data includes an electrocardiogram signal of at least one subject in a sleeping state within a first time period, and the raw data is segmented according to a preset time length to obtain a plurality of first data;

[0070] A denoising module 702 is used to denoise the first data to obtain denoised first data, extract shallow features of the denoised first data to obtain shallow features; decompose the first data using ensemble empirical mode decomposition to obtain a modal component sequence;

[0071] A training module 703 is used to train a deep neural network model based on the shallow features corresponding to each of the first data and the modal component sequence to obtain a trained deep neural network model;

[0072] The identification module 704 is used to detect the user's sleep state using the trained deep neural network model.

[0073] In a specific implementation of the present disclosure, the denoising module 702 further includes a filtering unit 7021 , a first calculating unit 7022 and a denoising unit 7023 .

[0074] A filtering unit 7021, configured to process the first data using a median filter to obtain second data;

[0075] A first calculation unit 7022, configured to perform a difference process on the first data and the second data to obtain third data;

[0076] The denoising unit 7023 is used to perform discrete wavelet transform denoising on the third data to obtain the denoised first data.

[0077] In a specific implementation of the present disclosure, the denoising module 702 further includes a second calculation unit 7024 , a correction unit 7025 and a third calculation unit 7026 .

[0078] The second calculation unit 7024 is used to process the denoised first data using a Pan-Tompkins algorithm to obtain an R wave peak index in the electrocardiogram signal;

[0079] A correction unit 7025 is used to correct the R wave peak index in the electrocardiogram signal to obtain a corrected R wave peak index, wherein the corrected R wave peak index includes a plurality of index values;

[0080] The third calculation unit 7026 is used to calculate the heart rate variability signal, the R wave peak sequence and the cardiogenic respiration signal according to the corrected R wave peak index.

[0081] In a specific implementation of the present disclosure, the third computing unit 7026 further includes a first computing subunit 70261 , an identification subunit 70262 and a second computing subunit 70263 .

[0082] A first calculation subunit 70261 is used for performing difference calculation on two adjacent index values ​​to obtain a heart rate variability signal;

[0083] The identification subunit 70262 is used to identify each peak value of the R wave in the first data after denoising to obtain an R wave peak value sequence;

[0084] The second calculation subunit 70263 is used to perform oversampling processing on the R wave peak value sequence to obtain the cardiogenic respiratory signal.

[0085] In a specific implementation of the present disclosure, the denoising module 702 further includes a decomposition unit 7027 and a repetition unit 7028 .

[0086] A decomposition unit 7027 is used for the decomposition step: adding white noise of random normal distribution to the first data to obtain fourth data, performing empirical mode decomposition on the fourth data to obtain multiple modal components, arranging all the modal components according to frequency to obtain fifth data, wherein the fifth data includes the arranged modal components;

[0087] The repeating unit 7028 is used to repeat the decomposition step to obtain multiple fifth data, integrate and average all the modal components arranged at the Nth position in the fifth data to obtain sixth data, and form all the sixth data into a modal component sequence, wherein N is a positive integer.

[0088] In a specific implementation of the present disclosure, the training module 703 further includes an input unit 7031 , a concatenation unit 7032 and a dimensionality reduction unit 7033 .

[0089] An input unit 7031 is used to input each modal component in the modal component sequence into a one-dimensional residual convolutional network during each training to extract local features of each modal component;

[0090] A concatenation unit 7032 is used to concatenate the local features of each of the modal components with all of the shallow features, and input the concatenated features into a long short-term memory recurrent neural network, and use the output of the long short-term memory recurrent neural network as an input of an attention model;

[0091] The dimension reduction unit 7033 is used to input the output of the attention model into two layers of fully connected networks in sequence, wherein the output of the attention model is reduced in dimension by using the first layer of fully connected network, and then input into the second layer of fully connected network after dimension reduction, and the binary classification operation is implemented by using the softmax function, and the final classification result is output to complete one training, and the classification result includes the awakening state or the non-awakening state.

[0092] It should be noted that, regarding the device in the above embodiment, the specific manner in which each module performs the operation has been described in detail in the embodiment of the method, and will not be elaborated here.

[0093] Example 3

[0094] Corresponding to the above method embodiments, the embodiments of the present disclosure further provide a micro-awakening detection device. The micro-awakening detection device described below and the micro-awakening detection method described above can refer to each other.

[0095] Figure 3 8 is a block diagram of a micro-awakening detection device 800 according to an exemplary embodiment. Figure 3 As shown, the micro-awakening detection device 800 may include: a processor 801 and a memory 802. The micro-awakening detection device 800 may also include one or more of a multimedia component 803, an I / O interface 804, and a communication component 805.

[0096] The processor 801 is used to control the overall operation of the micro-awakening detection device 800 to complete all or part of the steps in the above-mentioned micro-awakening detection method. The memory 802 is used to store various types of data to support the operation of the micro-awakening detection device 800. For example, these data may include instructions for any application or method operating on the micro-awakening detection device 800, as well as application-related data, such as contact data, sent and received messages, pictures, audio, video, etc. The memory 802 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, disk or optical disk. The multimedia component 803 may include a screen and an audio component. The screen may be, for example, a touch screen, and the audio component is used to output and / or input audio signals. For example, the audio component may include a microphone for receiving external audio signals. The received audio signal may be further stored in the memory 802 or sent via the communication component 805. The audio component also includes at least one speaker for outputting audio signals. The I / O interface 804 provides an interface between the processor 801 and other interface modules, and the above-mentioned other interface modules can be keyboards, mice, buttons, etc. These buttons can be virtual buttons or physical buttons. The communication component 805 is used for wired or wireless communication between the micro-awakening detection device 800 and other devices. Wireless communication, such as Wi-Fi, Bluetooth, Near Field Communication (NFC), 2G, 3G or 4G, or a combination of one or more of them, so the corresponding communication component 805 can include: Wi-Fi module, Bluetooth module, NFC module.

[0097] In an exemplary embodiment, the micro-awakening detection device 800 can be implemented by one or more application specific integrated circuits (ASIC), digital signal processors (DSP), digital signal processing devices (DSPD), programmable logic devices (PLD), field programmable gate arrays (FPGA), controllers, microcontrollers, microprocessors or other electronic components to perform the above-mentioned micro-awakening detection method.

[0098] In another exemplary embodiment, a computer-readable storage medium including program instructions is also provided, and when the program instructions are executed by a processor, the steps of the above-mentioned micro-awakening detection method are implemented. For example, the computer-readable storage medium can be the above-mentioned memory 802 including program instructions, and the above-mentioned program instructions can be executed by the processor 801 of the micro-awakening detection device 800 to complete the above-mentioned micro-awakening detection method.

[0099] Example 4

[0100] Corresponding to the above method embodiment, the present disclosure embodiment further provides a readable storage medium. The readable storage medium described below and the micro-awakening detection method described above can refer to each other.

[0101] A readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the micro-awakening detection method of the above method embodiment are implemented.

[0102] The readable storage medium may specifically be a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, or other readable storage medium that can store program codes.

[0103] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for detecting micro-awakening, characterized in that: include: Acquire original data, the original data including an electrocardiogram signal of at least one subject in a sleeping state within a first time period, and divide the original data into segments according to a preset time length to obtain a plurality of first data; Denoising the first data to obtain denoised first data, and extracting shallow features of the denoised first data to obtain shallow features; Decomposing the first data by using ensemble empirical mode decomposition to obtain a modal component sequence; Training a deep neural network model based on the shallow features corresponding to each of the first data and the modal component sequence to obtain a trained deep neural network model; Detecting the user's sleep state using the trained deep neural network model; The shallow features of the first data after denoising are extracted to obtain the shallow features, including: Processing the denoised first data using a Pan-Tompkins algorithm to obtain an R wave peak index in the electrocardiogram signal; Correcting the R wave peak index in the electrocardiogram signal to obtain a corrected R wave peak index, wherein the corrected R wave peak index includes a plurality of index values; The heart rate variability signal, the R wave peak sequence and the cardiogenic respiration signal are calculated based on the corrected R wave peak index.

2. The micro-awakening detection method according to claim 1, characterized in that: Denoising the first data to obtain the denoised first data, including: Processing the first data by using a median filter to obtain second data; Performing a difference process on the first data and the second data to obtain third data; The third data is subjected to discrete wavelet transform denoising processing to obtain the denoised first data.

3. The micro-awakening detection method according to claim 2, characterized in that: The heart rate variability signal, the R wave peak sequence and the cardiogenic respiration signal are calculated according to the corrected R wave peak index, including: The difference between two adjacent index values ​​is calculated to obtain a heart rate variability signal; Identify each peak value of the R wave in the first data after denoising to obtain an R wave peak sequence; The R wave peak value sequence is oversampled to obtain the cardiogenic respiratory signal.

4. A micro-awakening detection device, characterized in that: include: An acquisition module, used to acquire raw data, wherein the raw data includes an electrocardiogram signal of at least one subject in a sleeping state within a first time period, and the raw data is segmented according to a preset time length to obtain a plurality of first data; A denoising module, configured to denoise the first data to obtain denoised first data, and extract shallow features of the denoised first data to obtain shallow features; Decomposing the first data by using ensemble empirical mode decomposition to obtain a modal component sequence; A training module, used for training a deep neural network model based on the shallow features corresponding to each of the first data and the modal component sequence to obtain a trained deep neural network model; A recognition module, used to detect the user's sleep state using the trained deep neural network model; Wherein, the denoising module comprises: A second calculation unit is used to process the denoised first data using a Pan-Tompkins algorithm to obtain an R wave peak index in the electrocardiogram signal; a correction unit, configured to correct the R wave peak index in the electrocardiogram signal to obtain a corrected R wave peak index, wherein the corrected R wave peak index includes a plurality of index values; The third calculation unit is used to calculate the heart rate variability signal, the R wave peak sequence and the cardiogenic respiration signal according to the corrected R wave peak index.

5. The micro-awakening detection device according to claim 4, characterized in that: Denoising module, including: A filtering unit, configured to process the first data using a median filter to obtain second data; A first calculation unit, used for performing a difference process between the first data and the second data to obtain third data; The denoising unit is used to perform discrete wavelet transform denoising on the third data to obtain the denoised first data.

6. The micro-awakening detection device according to claim 4, characterized in that: The third computing unit includes: A first calculation subunit is used to perform difference calculation on two adjacent index values ​​to obtain a heart rate variability signal; an identification subunit, used to identify each peak value of the R wave in the first data after denoising, and obtain an R wave peak sequence; The second calculation subunit is used to perform oversampling processing on the R wave peak value sequence to obtain the cardiogenic respiratory signal.

7. A micro-awakening detection device, characterized in that: include: Memory for storing computer programs; A processor, configured to implement the steps of the micro-awakening detection method according to any one of claims 1 to 3 when executing the computer program.

8. A readable storage medium, characterized in that: The readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the micro-awakening detection method according to any one of claims 1 to 3 are implemented.

Images