Device with sleep evaluation and intervention functions

Abstract

The invention discloses a device with sleep assessment and intervention functions, which is arranged on a pillow or a back cushion, and comprises an information acquisition module for acquiring physiological data, head posture data and sound data of a human body, and sending the acquired data to a sleep assessment module, the sleep assessment module is used for extracting multi-modal sleep features according to the collected data and performing sleep assessment according to the multi-modal sleep features; wherein the source of the physiological data is analyzed by using the head posture data, and a dynamic weight is given, so that a self-adaptive sleep staging algorithm is completed; and the sleep intervention module is used for acquiring a sleep intervention scheme according to the sleep evaluation result, and activating different working modes of a vibration motor and / or a loudspeaker according to the sleep intervention scheme so as to achieve the effect of intervening the sleep of the user. According to the method, the source of the physiological data is analyzed by using the head posture data, and the dynamic weight is given, so that a self-adaptive sleep staging algorithm is completed, and the sleep staging evaluation accuracy of a real scene is improved.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and particularly to a device with sleep assessment and intervention functions. Background Art

[0002] Physiological signals record the microcurrents during human activities, and these changes can reflect different sleep stages and states. Taking electroencephalogram (EEG) signals as an example, during the wake period, mainly alpha waves (8 - 12 Hz) are observed. When an individual relaxes and closes their eyes, the brain waves of this frequency will significantly increase; non-rapid eye movement (NREM) sleep can be divided into N1, N2, and N3 stages. In the N1 stage, theta waves (4 - 7 Hz) are manifested. In the N2 stage, K-complex waves and sleep spindles (12 - 16 Hz) will appear. In the N3 stage (deep sleep or slow-wave sleep), delta waves (<4 Hz) are dominant; rapid eye movement (REM) sleep is similar to the low-amplitude fast waves in the waking state, but at this time most of the muscles of the body are in a relaxed state. The rest of the physiological signals have similar changes during sleep. By the characteristics of physiological signals in different sleep periods, the sleep stage of the user can be judged.

[0003] Non-invasive EEG acquisition devices do not require any surgical operations and use electrodes placed on the scalp surface to record EEG activities. However, most of these data acquisition methods are in the form of hats or headbands, and it is difficult for users to use these acquisition devices during sleep. At the same time, the existing acquisition devices are complex to wear and not easy to carry, and cannot collect the physiological data of users in various scenarios. In addition, when using portable devices to collect EEG data in daily life, problems such as the loss of brain region data or only specific brain region data being collected will occur as the user's posture changes. Most of the existing EEG sleep assessment methods are designed based on the complete situation of brain region data, so the recognition accuracy is not high in actual scenarios.

[0004] Sleep assessment, also known as sleep staging, is the process of automatically identifying and classifying different sleep stages by analyzing electroencephalogram (EEG) signals. Existing methods usually only extract manual features in EEG data, such as spectral power density (PSD), time-domain statistics, non-linear dynamic characteristics, etc., and then use models such as convolutional neural networks, graph neural networks, and Transformers to analyze the sleep situation.

[0005] Non-drug sleep interventions can be divided into psychological interventions and other auxiliary methods. Psychological interventions, guided by a psychologist, help users relax during sleep and fall asleep more easily. Other auxiliary methods improve sleep conditions through light therapy, sound therapy, temperature regulation, vibration stimulation, electromagnetic field exposure, etc. Existing non-drug sleep intervention methods often specify a plan when the user is awake and execute the same plan during sleep. However, a user's sleep can be divided into different periods, and different plans have different effects in different sleep periods. Using the same plan all the time will reduce the effectiveness of sleep intervention. Summary of the Invention

[0006] To solve at least one of the technical problems existing in the prior art to a certain extent, an object of the present invention is to provide a device with sleep assessment and intervention functions.

[0007] The first technical solution adopted by the present invention is:

[0008] A device with sleep assessment and intervention functions, which is set on a pillow or a lumbar pillow, includes:

[0009] An information collection module, configured to collect physiological data, head posture data, and sound data of the human body, and send the collected data to the sleep assessment module.

[0010] A sleep assessment module, configured to extract multi-modal sleep features based on the collected data and perform sleep assessment according to the multi-modal sleep features; wherein, the head posture data is used to analyze the source of the physiological data and assign dynamic weights to complete an adaptive sleep staging algorithm.

[0011] A sleep intervention module, configured to obtain a sleep intervention plan according to the sleep assessment result, and activate different working modes of the vibration motor and / or the speaker according to the sleep intervention plan to achieve the effect of intervening in the user's sleep.

[0012] Further, the physiological data includes at least one of electroencephalogram (EEG) signals, electrodermal activity (EDA) signals, electromyogram (EMG) signals, heart rate signals, blood oxygen signals, or body temperature signals.

[0013] The physiological data is collected by a multi-functional flexible sensor array implanted in a flexible fabric layer. The multi-functional flexible sensor includes EEG electrodes and at least one of EMG electrodes, EDA electrodes, blood oxygen sensors, or body temperature sensors.

[0014] Among them, the EEG electrodes, EMG electrodes, and EDA electrodes are designed as flexible structures with needle-like protrusions to ensure good contact between the flexible electrodes and the scalp and reduce the electrode contact impedance.

[0015] Further, the head posture data is acquired by a flexible sensor array implanted in the flexible fabric layer. The flexible sensor array is a sensor array composed of flexible pressure sensors, and by measuring the pressure distribution in the array, the change in the head posture is detected.

[0016] Further, the working mode of the information acquisition module is as follows:

[0017] For the electroencephalogram (EEG) signal in the physiological data, the EEG signal is filtered and then reconstructed using discrete wavelet transform to remove the noise in the EEG signal.

[0018] For the head posture data, it needs to be excited and amplified during acquisition. Excitation is used to generate a measurable output for the sensor, and amplification is used to increase the amplitude of the measurable value to improve the signal-to-noise ratio.

[0019] For the voice data, preprocessing is performed in the way of frame segmentation and windowing to approximate the voice data as a stationary signal, which is convenient for feature extraction of the subsequent model.

[0020] The processed physiological data, head posture data, and voice data are normalized to map the data to the same range to eliminate the dimension difference.

[0021] Further, the expression for reconstructing the EEG signal using discrete wavelet transform is:

[0022] cA j-1 [n] = ∑ k h[n - 2k]cA j [k] + Σ k g[n - 2k]cD j [k]

[0023] In the formula, h[n - 2k] and g[n - 2k] are reconstruction filters; cA j [k] is the approximation coefficient; cD j [k] is the detail coefficient, both of which are obtained by layer-by-layer decomposition of wavelet transform; c is the coefficient of each part; k is the summation index used to traverse possible time points.

[0024] Further, the preprocessing in the way of frame segmentation and windowing includes:

[0025] The continuous voice data is segmented into small segments of a fixed length, and the formula is expressed as:

[0026]

[0027] In the formula, X[n] is the voice data represented by a discrete-time sequence, N is the total length of the signal, L is the frame length of each frame during frame segmentation, S is the frame shift, that is, the interval between two adjacent frames, fi is the i-th frame after frame segmentation;

[0028] After frame segmentation, windowing processing is performed on each frame to prevent mutations at both ends of the frame and spectral leakage problems after Fourier transform.

[0029] Furthermore, the working mode of the sleep assessment module is as follows:

[0030] For head pose data, use the brain region adaptive analysis algorithm to calculate the posture of the human body during sleep to obtain the brain region source of the current EEG signal, and use the projection layer to map EEG features and other features into the same feature space;

[0031] For physiological data, use the self-attention mechanism to extract the shallow temporal features of physiological data to characterize the current sleep state of the human body;

[0032] For sound data, use the snoring detection algorithm to extract snoring features to assist the judgment of the sleep assessment model, and then use the projection layer to map sound features and other features into the same feature space;

[0033] Through the cross-attention mechanism, use the head pose data to adaptively judge the signal acquisition situation, assign dynamic weights to different modalities of physiological data, complete the multi-modal fusion of multiple physiological data and sound data, and extract multi-modal sleep features;

[0034] Obtain the sleep stage label according to the multi-modal sleep features and generate a sleep assessment report.

[0035] Furthermore, the working mode of the sleep intervention module is as follows:

[0036] According to the sleep assessment result, use the dynamic matching algorithm to obtain the scheme with the highest similarity between the current human sleep state characteristics and the scheme characteristics in the preset scheme library, and control the working mode of the vibration motor and / or speaker according to the obtained scheme;

[0037] Among them, the dynamic matching algorithm is the Wasserstein distance algorithm, Hellinger distance algorithm, JS divergence algorithm, KL divergence algorithm or total variation distance algorithm.

[0038] Furthermore, for different schemes, the control method of the vibration motor is as follows:

[0039] During the light sleep stage (N1), use extremely low-frequency vibration below 60Hz to provide a comfortable background feeling for the human body and help enter a deeper sleep state faster;

[0040] During the light sleep stage (N2), vibrations of 60 - 150 Hz are used. This level of vibration can maintain a certain sense of comfort without easily disrupting the established stable sleep pattern.

[0041] During the deep sleep stage (N3), vibrations of 90 - 300 Hz are employed. Such vibration intensity is sufficient to provide additional security and support without causing arousal.

[0042] During the rapid eye movement sleep stage (REM), brain activity increases, muscle tension weakens, and it is easy to wake up. Therefore, vibration intervention is suspended during this period.

[0043] In addition, if severe sleep apnea is detected in the human body, the vibration motor will emit strong vibrations to wake up the user and prevent serious accidents.

[0044] Furthermore, for different scenarios, the control methods of the speaker are as follows:

[0045] During the light sleep stage (N1), the brain gradually transitions from the waking state to the light sleep state. Background sounds with an evenly distributed energy spectrum such as white noise or pink noise are played during this stage, which helps to smoothly enter a deeper sleep.

[0046] During the light sleep stage (N2), a person's consciousness further weakens, and spindle waves and K-complex waves begin to appear. Low-frequency vibrations or gentle natural sounds are used during this stage, which helps to stabilize sleep.

[0047] During the deep sleep stage (N3), the brain generates a large number of delta waves. Extremely low-frequency sounds are used during this period to provide additional security for the human body.

[0048] During the rapid eye movement sleep stage (REM), brain activity increases, muscle tension weakens, and it is easy to wake up. Low-frequency binaural beats (about 2 Hz to 7 Hz) are played during this period to induce theta brain waves in the brain, thereby optimizing sleep quality.

[0049] The beneficial effects of the present invention are as follows: The present invention uses head posture data to analyze the source of physiological data and assigns dynamic weights to complete an adaptive sleep staging algorithm, improving the accuracy of sleep staging assessment in real scenarios. In addition, the present invention also uses sound data as an auxiliary modality for the sleep assessment module to make the recognition results more accurate. Description of the Drawings

[0050] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following provides an introduction to the accompanying drawings of the relevant technical solutions in the embodiments of the present invention or the prior art. It should be understood that the accompanying drawings in the following introduction are only for conveniently and clearly presenting some embodiments of the technical solutions of the present invention. For those skilled in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0051] Figure 1 is a schematic diagram of the device with sleep assessment and intervention functions in the embodiment of the present invention;

[0052] Figure 2 is the overall flowchart of the device with sleep assessment and intervention functions in the embodiment of the present invention;

[0053] Figure 3 is the hardware architecture diagram of the device with sleep assessment and intervention functions in the embodiment of the present invention;

[0054] Figure 4 is the working schematic diagram of the information collection module in the embodiment of the present invention;

[0055] Figure 5 is the working schematic diagram of the sleep assessment module in the embodiment of the present invention;

[0056] Figure 6 is the working schematic diagram of the sleep intervention module in the embodiment of the present invention. Detailed Embodiment

[0057] The following details the embodiments of the present invention. The examples of the embodiments are shown in the accompanying drawings, where the same or similar reference numerals represent the same or similar elements or elements with the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as a limitation of the present invention. For the step numbers in the following embodiments, they are only set for the convenience of explanation and illustration, and no limitation is placed on the order between the steps. The execution order of each step in the embodiments can be adaptively adjusted according to the understanding of those skilled in the art.

[0058] In the description of the present invention, it should be understood that for the orientation description, such as the orientation or positional relationship indicated by up, down, front, back, left, right, etc., is based on the orientation or positional relationship shown in the accompanying drawings. It is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the present invention.

[0059] In the description of the present invention, the meaning of "several" is one or more, the meaning of "multiple" is more than two, understandings such as "greater than", "less than", "exceeding", etc. do not include the corresponding number, and understandings such as "above", "below", "within", etc. include the corresponding number. If there is a description of "first" and "second", it is only for the purpose of distinguishing technical features, and cannot be understood as indicating or implying relative importance or implicitly indicating the quantity of the indicated technical features or implicitly indicating the sequence relationship of the indicated technical features.

[0060] In the description of the present invention, unless otherwise clearly defined, words such as "set", "installed", "connected", etc. should be understood in a broad sense, and those skilled in the art can reasonably determine the specific meanings of the above words in the present invention in combination with the specific content of the technical solution.

[0061] Term Explanation:

[0062] PCBA: It is the abbreviation of the English Printed Circuit Board Assembly, printed circuit board.

[0063] As Figure 2 shown, this embodiment provides a device with sleep assessment and intervention functions, which is set on a pillow or a lumbar pillow, and includes:

[0064] An information acquisition module, which is used to acquire the physiological data, head posture data and sound data of the human body, and send the acquired data to the sleep assessment module.

[0065] A sleep assessment module, which is used to extract multi-modal sleep features according to the acquired data and perform sleep assessment according to the multi-modal sleep features; among them, the head posture data is used to analyze the source of the physiological data and assign dynamic weights to complete the adaptive sleep staging algorithm.

[0066] A sleep intervention module, which is used to obtain a sleep intervention plan according to the sleep assessment result, and activate different working modes of the vibration motor and / or the speaker according to the sleep intervention plan to achieve the effect of intervening in the user's sleep.

[0067] The working process of the device provided in this embodiment is as follows: The user attaches the device with sleep assessment and intervention functions to a pillow or a lumbar pillow, and during sleep, places the head on top of the device; the information acquisition module of the device will collect the user's posture signals, various physiological signals and sound signals in real time, and send these data to the sleep assessment module after preprocessing; the sleep assessment module realizes the multi-modal sleep staging algorithm through the posture data, sound data and various physiological data, in which the head posture data is used to analyze the source of the physiological data and assign dynamic weights to complete the adaptive sleep staging algorithm and obtain an accurate sleep assessment report. The sleep intervention module adjusts the stimulation intensity of sound waves and vibrations in real time according to the sleep assessment report to improve the user's sleep quality.

[0068] In one embodiment, referring to Figure 1 , the main body of the device with sleep assessment and intervention functions is composed of two fabric layers 3 to ensure sleep comfort. A multi-functional sensor 2, a pressure sensor 4, and a printed circuit board 1 are embedded in the fabric layer 3, and a microphone, a speaker, and a vibration motor are also provided in the printed circuit board 1. The specific composition is as follows:

[0069] 1) Flexible material

[0070] The main body material can be selected from flexible materials such as fabric and leather. The material selection ensures comfort and aesthetics and conforms to the daily usage habits of users.

[0071] 2) Head posture detection sensor array

[0072] Flexible sensors capable of detecting head posture changes are used and implanted into the flexible fabric layer. Multiple sensor units form an array. Specifically, a sensor array composed of flexible pressure sensors can be selected, and by measuring the pressure distribution in the array, the changes in head posture can be detected in real time.

[0073] 3) Multi-functional flexible sensor array

[0074] The multi-functional sensor includes vital sign detection units such as electroencephalogram electrodes, electromyogram electrodes, galvanic skin electrodes, blood oxygen sensors, and body temperature sensors. The multi-functional sensors are implanted into the flexible fabric layer to form an array. The electroencephalogram electrodes, electromyogram electrodes, and galvanic skin electrodes can be designed as flexible structures with needle-like protrusions to ensure good contact between the flexible electrodes and the scalp, reduce the electrode contact impedance, and ensure comfort and safety at the same time. The blood oxygen sensor is used to measure blood oxygen, heart rate, blood pressure, and respiratory rate. The temperature sensor is used to detect changes in the human body temperature.

[0075] 4) Microphone

[0076] Its main function is to collect the user's snoring, breathing sounds, and environmental noises and interact with the user's voice. The number of microphones is one or more, and they are implanted into the flexible fabric layer.

[0077] 5) Speaker

[0078] During the sleep intervention stage, it plays sounds in a specific frequency band to assist the user in entering a deeper sleep and improving the user's sleep quality, or interacts with the user's voice. The number of speakers is one or more, and they are implanted into the flexible fabric layer.

[0079] 6) Vibration motor

[0080] During the sleep intervention stage, micro-vibrations are used to assist the user in entering a deeper sleep and improving the user's sleep quality. Or when the user has severe sleep apnea, they can be awakened by vibration to avoid accidents. One or more are implanted into the flexible fabric layer.

[0081] Further as an optional implementation method, refer to Figure 3 In the data processing flow of the device with sleep assessment and intervention functions, it can be mainly divided into an input module, an output module, and a communication management module. The input module is composed of an attitude detection sensor array, a multi-functional flexible sensor array, and a microphone, and is mainly responsible for collecting various data of the user. The output module is composed of a speaker and a vibration motor, and is the execution component when generating a sleep intervention plan. The communication management module is composed of power management and wireless communication, and is mainly responsible for the power-on operation of the device and the communication between modules.

[0082] The following combines the drawings and specific implementation methods to explain each module in the device in detail.

[0083] (1) Information collection module

[0084] Refer to Figure 4 The information collection module uses a multi-functional sensor array, an attitude sensor array, and a microphone to receive various physiological data, attitude data, and sound data respectively, providing core data for the subsequent sleep assessment algorithm.

[0085] 1) For various physiological data, taking electroencephalogram (EEG) data as an example, first use a Butterworth band-pass filter to limit the frequency range within a certain high and low frequency range to remove noise and unnecessary signals. Since the frequency range of EEG signals is 0.5 - 70 Hz, a 0.5 Hz low-pass filter and a 70 Hz high-pass filter are adopted. The transfer function formula is as follows:

[0086]

[0087] Where H(s) is the transfer function of the system, which defines the response of the system to different frequency inputs in the Laplace transform domain. G0 is the gain, usually a constant, representing the maximum gain of the filter (usually 1, meaning no amplification or attenuation). ω0 is the center angular frequency. s is the complex frequency variable, used for Laplace transform in continuous-time systems. n is the order of the filter, which determines the selectivity and steepness of the filter. Then use discrete wavelet transform to reconstruct the EEG data to remove noise in the EEG signal. The formula is as follows:

[0088] cA j-1 [n] = ∑ k h[n - 2k]cA j [k] + ∑ k g[n - 2k]cDj [k]

[0089] Among them, h[n - 2k] and g[n - 2k] are reconstruction filters, and A j [k] is the approximation coefficient, and D j [k] is the detail coefficient, both of which are obtained by layer-by-layer decomposition of wavelet transform.

[0090] For other physiological data, this part of the data includes electrodermal activity, electromyogram, heart rate, body temperature and other data. Electrodermal activity and electromyogram data are physiological micro-current data and are easily affected by other factors. Therefore, a filtering operation similar to that of EEG data is used for denoising. Heart rate and body temperature data are single-valued variables, so no processing is done on them. Other types of physiological data added subsequently also follow this preprocessing rule.

[0091] 2) For attitude data, excitation and amplification are required during acquisition. Excitation is used to generate a measurable output for the sensor, and amplification is used to increase the amplitude of the measurable value to improve the signal-to-noise ratio. Specifically, a constant voltage source V EXC is used as the excitation, and the output voltage can be expressed as:

[0092] V OUT = kV EXC ΔR / R

[0093] Then, the obtained attitude signal is amplified using an instrumentation amplifier:

[0094] V OUT = G(V IN+ - V IN- )

[0095] Where V IN+ and V IN- correspond to the positive and negative output terminals of the bridge respectively.

[0096] 3) For speech data, preprocessing is performed in the way of frame segmentation and windowing to approximate the speech data as a stationary signal, which is convenient for feature extraction of subsequent models. Frame segmentation means dividing continuous speech data into small segments of fixed length, which can be expressed by the formula:

[0097]

[0098] Where X[n] is the speech data represented by a discrete time series, N is the total length of the signal, L is the frame length of each frame during frame segmentation, S is the frame shift, i.e., the interval between two adjacent frames, and f i is the i-th frame after frame segmentation. After frame segmentation, windowing is performed on each frame to prevent sudden changes at both ends of the frame and spectral leakage problems after Fourier transform. Here, the Hamming window is used as follows:

[0099]

[0100] The new frame f after applying the window function i ′[n] can be expressed as the element-wise product of the original frame and the window function:

[0101] f i ′[n]=fi i [n]×ω[n], 0 ≤ n < L

[0102] After processing, various physiological data, posture data, and voice data are obtained. To ensure more stability and efficiency when using subsequent algorithms, these data are normalized to map the data into the same range to eliminate the dimensional difference. The specific formula is as follows:

[0103]

[0104] where X is the original data, X min and X max are the minimum and maximum values of this data respectively. Through normalization, the data can be mapped into the interval [0, 1].

[0105] (2) Sleep assessment module

[0106] See Figure 5 , the input of the sleep assessment module is the posture data, various physiological data, and sound data obtained from the information collection module. Then, a multi-modal sleep staging model is used to obtain the multi-modal sleep characteristics of the user. Next, a classifier is used to classify them. Finally, a visual user sleep assessment report is output.

[0107] Specifically, in the multi-modal sleep staging model, for the posture data, a brain region adaptive analysis algorithm is used to calculate the posture of the user during sleep, accurately obtain the brain region source of the current EEG data, and then a projection layer is used to map it to the same feature space as other features; for various physiological data, a self-attention mechanism is used to extract the shallow temporal features of these physiological data to characterize the current sleep state of the user; for the sound data, a snoring detection algorithm is used to extract snoring features to assist the judgment of the sleep assessment model, and then a projection layer is used to map it to the same feature space as other features. In the cross-attention layer, the posture data is used to adaptively judge the signal acquisition situation and assign dynamic weights to different modalities of physiological data, and multi-modal fusion of various physiological data and sound data is completed in this layer. A multi-layer Transformer is stacked as the temporal feature layer to extract multi-modal sleep characteristics, and then CNN, LSTM, or SVM is used as the classifier to obtain the sleep staging label. Finally, visualization is performed to obtain the sleep assessment report.

[0108] (3) Sleep intervention module

[0109] See Figure 6 , the sleep intervention module mainly uses the dynamic matching algorithm to obtain the most suitable sleep intervention plan according to the sleep assessment report output by the sleep assessment module, and then activates different working modes of the vibration motor and the speaker according to the plan, and finally achieves the effect of intervening in the user's sleep.

[0110] The plan library consists of preset plans and personalized plans. The preset plans are composed of the frequencies of the vibration motor preset by experts according to the sleep characteristics in different periods and the audio played by the speaker; the personalized plans are composed of the frequencies of the vibration motor and the audio played by the speaker customized by the user according to their own needs. The dynamic matching algorithm can use Wasserstein distance, Hellinger distance, JS divergence, etc., to screen the plan with the highest similarity between the current user's sleep state characteristics and the plan characteristics in the plan library, and this algorithm can screen more suitable plans in real time according to the change of the sleep state. Finally, different working modes of the vibration motor and the speaker are activated according to different plans.

[0111] Slight and rhythmic vibrations can sometimes cause drowsiness. For example, you may sleep better on a train or in a car sometimes. At the same time, the somatosensory music therapy system points out that vibrations in the range of 60 - 600 Hz can stimulate the response of Pacinian corpuscles in the human body, thereby alleviating symptoms such as tension and fatigue, which is beneficial to relaxation and falling asleep. Therefore, this patent uses vibration to intervene in the user's sleep state. Specifically, in the light sleep stage (N1), extremely low-frequency vibrations below 60 Hz are used to provide a comfortable background feeling for the body and help enter a deeper sleep state faster; in the moderate sleep stage (N2), slightly higher-frequency and gentle-rhythm vibrations around 60 - 150 Hz are introduced. Such vibrations can maintain a certain degree of comfort and will not easily interrupt the established stable sleep pattern; in the deep sleep stage (N3), deep and slow vibrations between 90 - 300 Hz are adopted. Such vibration intensity is sufficient to provide additional security and support, and at the same time will not cause awakening. During the rapid eye movement sleep stage (REM), brain activity increases and muscle tension weakens, making it easy to wake up. Therefore, vibration intervention is suspended during this period. In addition, if it is detected that the user has severe sleep apnea, the vibration motor will emit a strong vibration to wake up the user and prevent serious accidents.

[0112] Sound can help regulate the body's biological rhythm. Hearing appropriate audio during sleep can cause synchronous resonance of tissue cells, relieve the body's tension and anxiety, and thus achieve the effect of helping sleep. Therefore, this patent also uses sound as a way of sleep intervention. Specifically, during the light sleep stage (N1), the brain gradually transitions from a waking state to a light sleep state. At this time, playing background sounds such as white noise or pink noise with an evenly distributed energy spectrum helps to smoothly enter a deeper sleep; during the moderate sleep stage (N2), a person's consciousness further weakens, and spindle waves and K complex waves begin to appear. Using low-frequency vibrations or gentle natural sounds such as gurgling water and forest birdsong during this stage helps to stabilize sleep; during the deep sleep stage (N3), the brain produces a large number of delta waves. At this time, extremely low-frequency sounds are used to provide the body with additional security. During the rapid eye movement sleep stage (REM), brain activity increases and muscle tension weakens, making it easy to wake up. Playing low-frequency binaural beats (about 2 Hz to 7 Hz) during this stage induces the brain to produce theta brain waves, thereby optimizing sleep quality.

[0113] In summary, compared with the prior art, the device of the present invention has the following advantages and beneficial effects:

[0114] 1) Most of the existing non-invasive electroencephalogram (EEG) data acquisition methods use the form of hats or headbands, and it is difficult for users to use these acquisition devices during sleep. At the same time, the existing acquisition devices are complex to wear and not convenient to carry, and cannot collect users' physiological data in multiple scenarios. The present invention proposes a device with the function of collecting EEG data and other physiological signals, on which a flexible electrode array with a specific distribution is used. Users can place it on the pillow or on the cushion when traveling, and can complete the acquisition of EEG data and other physiological signals without feeling.

[0115] 2) When the existing technical solutions use portable devices to collect EEG data in daily life, problems such as loss of brain area data or only collecting specific brain area data will occur as the user's posture changes. Most of the existing EEG sleep assessment methods are designed based on the complete situation of brain area data. Therefore, the recognition accuracy is not high in actual scenarios. The present invention proposes a sleep staging algorithm based on brain area adaptive estimation, which can judge the source of the currently collected brain area data through the user's posture, and thus dynamically activate different parts of the model to complete the sleep staging algorithm of the adaptive brain area, improving the sleep staging assessment accuracy in real scenarios.

[0116] 3) Existing sleep staging algorithms often use only electroencephalogram (EEG) data, but other physiological data can also provide auxiliary information for sleep staging to improve the accuracy of recognition. The present invention proposes a multi-modal sleep staging model that uses EEG data as the main data supplemented by physiological data such as voice, electrodermal activity, electromyogram, heart rate, and body temperature, etc., to jointly analyze the current sleep stage of the user and improve the accuracy of the sleep staging algorithm. The sleep assessment module of the present invention can utilize multi-modal physiological signals mainly based on EEG to improve the credibility of the sleep staging algorithm. At the same time, the source of physiological data is analyzed using posture data and dynamic weights are assigned to complete an adaptive sleep staging algorithm, improving the accuracy of sleep staging assessment in real scenarios. In addition, the present invention also uses snoring data as an auxiliary modality for the sleep assessment module to make the results more accurate.

[0117] 4) Existing non-drug sleep intervention methods often specify a plan when the user is awake and keep executing the same plan during sleep. However, the user's sleep can be divided into different periods, and different plans have different effects in different sleep periods. Using the same plan all the time will reduce the effect of sleep intervention. The present invention proposes a sleep dynamic intervention plan based on sleep staging, which dynamically selects different intervention plans according to the sleep assessment results to improve the user's sleep quality. The invention can adjust the sleep intervention plan in real time according to the sleep assessment report through a dynamic matching algorithm, thereby achieving a more effective sleep intervention effect. The method of using vibration and audio to work together is adopted to intervene in the user's sleep, making the intervention effect more effective.

[0118] In the description of this specification, the descriptions with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples", etc. mean that the specific features, structures, materials, or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described can be combined in any one or more embodiments or examples in a suitable manner. In addition, without contradiction, those skilled in the art can combine and combine the different embodiments or examples described in this specification and the features of different embodiments or examples.

[0119] The above embodiments are only used to illustrate the technical concept and features of the present invention, and their purpose is to enable those of ordinary skill in the art to understand the content of the present invention and implement it accordingly, and cannot be used to limit the protection scope of the present invention. Any equivalent changes or modifications made according to the essence of the content of the present invention should be covered by the protection scope of the present invention.

Claims

1. A device with sleep assessment and intervention functions, characterized in that Set on a pillow or a lumbar cushion, including: An information acquisition module, configured to acquire physiological data, head posture data, and voice data of a human body, and send the acquired data to a sleep assessment module; A sleep assessment module, configured to extract multi-modal sleep features based on the acquired data, and perform sleep assessment according to the multi-modal sleep features; wherein, the head posture data is used to analyze the source of the physiological data and assign dynamic weights to complete an adaptive sleep staging algorithm; A sleep intervention module, configured to obtain a sleep intervention plan according to the sleep assessment result, and activate different working modes of a vibration motor and / or a speaker according to the sleep intervention plan to achieve the effect of intervening in the user's sleep.

2. The device with sleep assessment and intervention functions according to claim 1, characterized in that, The physiological data includes at least one of electroencephalogram (EEG) signals, electrodermal activity (EDA) signals, electromyogram (EMG) signals, heart rate signals, blood oxygen signals, or body temperature signals in addition to EEG signals; The physiological data is acquired through a multi-functional flexible sensor array implanted into a flexible fabric layer. The multi-functional flexible sensor includes EEG electrodes and at least one of EMG electrodes, EDA electrodes, blood oxygen sensors, or body temperature sensors; Among them, the EEG electrodes, EMG electrodes, and EDA electrodes are designed into a flexible structure with needle-like protrusions to ensure good contact between the flexible electrodes and the scalp and reduce the electrode contact impedance.

3. The device with sleep assessment and intervention functions according to claim 1, characterized in that, The head posture data is acquired through a flexible sensor array implanted into the flexible fabric layer. The flexible sensor array is a sensor array composed of flexible pressure sensors, and the change in the head posture is detected by measuring the pressure distribution in the array.

4. The device with sleep assessment and intervention functions according to claim 1, characterized in that, The working mode of the information acquisition module is as follows: For the EEG signals in the physiological data, the EEG signals are filtered, and then the discrete wavelet transform is used to reconstruct the EEG signals to remove the noise in the EEG signals; For the head posture data, excitation and amplification are required during acquisition. The excitation is used to generate a measurable output for the sensor, and the amplification is used to increase the amplitude of the measurable value to improve the signal-to-noise ratio; For the voice data, preprocessing is performed in a frame-by-frame and windowing manner to approximate the voice data as a stationary signal, which is convenient for subsequent model feature extraction; The processed physiological data, head posture data, and voice data are normalized to map the data to the same range to eliminate the dimension difference.

5. The device with sleep assessment and intervention functions according to claim 4, characterized in that, The expression for reconstructing the EEG signals using the discrete wavelet transform is: cA j-1 [n] = ∑ k h[n - 2k]cA j [k] + ∑ k g[n - 2k]cD j [k] where h[n - 2k] and g[n - 2k] are reconstruction filters; cA j [k] is an approximation coefficient; cD j [k] is a detail coefficient, both obtained by layer-by-layer decomposition of wavelet transform; c is the coefficient of each part; k is the summation index.

6. The device with sleep assessment and intervention functions according to claim 4, characterized in that, The preprocessing in the frame-by-frame and windowing manner includes: The continuous voice data is segmented into small segments of a fixed length, and the formula is expressed as: where X[n] is the voice data represented by a discrete time series, N is the total length of the signal, and L is the frame length of each frame during frame segmentation; S is the frame shift, i.e., the interval between two adjacent frames, and f i is the i-th frame after frame division; After frame segmentation, each frame is windowed to prevent sudden changes at both ends of the frame and spectral leakage problems after Fourier transform.

7. The device with sleep assessment and intervention functions according to claim 1, characterized in that, The working mode of the sleep assessment module is as follows: For the head posture data, a brain region adaptive analysis algorithm is used to calculate the posture of the human body during sleep to obtain the brain region source of the current EEG signals, and a projection layer is used to map the EEG features and other features to the same feature space; For the physiological data, a self-attention mechanism is used to extract the shallow temporal features of the physiological data to characterize the current sleep state of the human body; For voice data, a snoring detection algorithm is used to extract snoring features for assisting the judgment of the sleep assessment model, and then a projection layer is used to map the voice features and other features into the same feature space; Through the cross-attention mechanism, the head pose data is used to adaptively judge the signal acquisition situation, and dynamic weights are assigned to physiological data of different modalities to complete the multi-modal fusion of various physiological data and voice data, and multi-modal sleep features are extracted; Sleep stage labels are obtained according to the multi-modal sleep features, and a sleep assessment report is generated.

8. A device with sleep assessment and intervention functions according to claim 1, characterized in that, The working mode of the sleep intervention module is as follows: According to the sleep assessment result, a dynamic matching algorithm is used to obtain the plan with the highest similarity between the current human sleep state features and the plan features in the preset plan library, and the working modes of the vibration motor and / or the speaker are controlled according to the obtained plan; Among them, the dynamic matching algorithm is the Wasserstein distance algorithm, the Hellinger distance algorithm, the JS divergence algorithm, the KL divergence algorithm or the total variation distance algorithm.

9. The device with sleep evaluation and intervention functions according to claim 8, characterized in that, For different plans, the control methods of the vibration motor are as follows: During the light sleep stage, extremely low-frequency vibrations below 60 Hz are used to provide a comfortable background feeling for the human body and help enter a deeper sleep state faster; During the moderate sleep stage, vibrations of 60 - 150 Hz are used. Such vibrations can maintain a certain degree of comfort and will not easily interrupt the established stable sleep pattern; During the deep sleep stage, vibrations of 90 - 300 Hz are adopted. Such vibration intensity is sufficient to provide additional sense of security and support, and at the same time will not cause awakening; During the rapid eye movement sleep stage, brain activity increases and muscle tension weakens, making it easy to wake up. Therefore, the vibration intervention is suspended during this period; In addition, if severe sleep apnea is detected in the human body, the vibration motor will emit a strong vibration to wake up the user and prevent serious accidents.

10. The device with sleep assessment and intervention functions according to claim 8, characterized in that, For different plans, the control methods of the speaker are as follows: During the light sleep stage, the brain gradually transitions from the waking state to the light sleep state. Background sounds with an evenly distributed energy spectrum are played at this stage, which helps to smoothly enter a deeper sleep; During the moderate sleep stage, a person's consciousness further weakens, and spindle waves and K-complex waves begin to appear. Low-frequency vibrations or soft natural sounds are used at this stage, which helps to stabilize sleep; During the deep sleep stage, the brain generates a large number of δ waves. Extremely low-frequency sounds are used at this stage to provide additional sense of security for the human body; During the rapid eye movement sleep stage, brain activity increases and muscle tension weakens, making it easy to wake up. Low-frequency binaural beats are played during this period to induce the brain to produce θ brain waves, thereby optimizing sleep quality.

Images