A sleep assistance method and system

Abstract

The present invention provides a sleep assistance method and system. The method includes the following steps: Step 1: Collect piezoelectric signals through a sleeping pillow embedded with a thin-film piezoelectric sensor; Step 2: Process the collected piezoelectric signals to obtain human physiological signals; Step 3: Obtain a sleep stage time series based on the obtained human physiological signals; Step 4: Determine sleep intervention measures through a BP neural network according to the sleep stage time series; Step 5: Implement the sleep intervention measures; Step 6: Evaluate the effect of the sleep intervention measures. The technical solution of the present invention is a closed loop, including the entire processes of sleep monitoring, sleep intervention, and intervention evaluation; in the sleep monitoring stage, non-invasive monitoring is achieved, and there is no physical discomfort or sleep interference caused by wearing a device for a long time; based on the non-invasive monitoring stage data, sleep intervention measures are determined, making the sleep intervention measures more targeted at the actual sleep situation.

Description

Technical Field

[0001] The present invention relates to the technical field of sleep assistance, and particularly to a sleep assistance method and system. Background Art

[0002] In today's society, sleep quality has become one of the important indicators of people's health and quality of life. People are increasingly concerned about the state of their own bodies during sleep. With the continuous progress of technology, sleep monitoring technology is constantly developing. The evolution from traditional sleep monitoring devices to more intelligent and convenient monitoring means aims to improve the accuracy and comfort of sleep monitoring. Currently, polysomnography monitoring is the most effective monitoring means for sleep disorders, but the subjects need to be continuously observed and recorded for 8 hours, and the long-term monitoring causes limitations in the examination.

[0003] In order to improve the convenience of monitoring, a large amount of research has been done, and a series of problems existing in traditional medical detection instruments have been successfully solved, such as complex operation, high professional requirements, the need for patients to go to the hospital for on-site monitoring, untimely monitoring, bulky instruments, poor portability, and inability to meet the needs of telemedicine services. However, there are still some problems to be solved. For example, long-term wearing of the device will cause physical discomfort, the wearing of the device will cause a certain degree of interference to sleep, and it is impossible to simultaneously meet the requirements of no load, convenience, and accuracy.

[0004] In addition, one of the main purposes of sleep monitoring is to timely intervene in sleep to improve sleep quality when poor sleep quality is found. Real-time intervention mainly uses the real-time data of non-invasive monitoring and uses means such as slow waves, sounds, and vibrations to intervene in sleep. According to sleep characteristics, it is mainly divided into two stages of intervention, one is the intervention in the falling asleep stage, and the other is the intervention in the sleep stage.

[0005] The intervention in the falling asleep stage is mainly to solve the problem of difficulty in falling asleep. Different methods are adopted for different groups of people, and these methods mainly include white noise, pink noise, light music, vibration and other means. The intervention in the sleep stage is to adopt different methods according to different sleep stages and for different groups of people, and these methods mainly include slow waves, vibration for snoring cessation, smell, room temperature adjustment, etc.

[0006] Currently, various technologies for sleep staging and sleep intervention based on sleep staging have emerged like mushrooms after a spring rain. However, such technologies either perform sleep intervention after staging based on electroencephalogram signals and cannot achieve non-invasive sleep monitoring, or only perform sleep intervention on the existing sleep staging data and are separated from sleep monitoring. In view of this, a technology that can achieve non-invasive sleep monitoring and perform sleep intervention based on the results of non-invasive sleep monitoring is urgently needed. Summary of the Invention

[0007] To solve at least one of the above technical problems, the present invention provides a sleep assistance method and system.

[0008] In a first aspect of the present invention, a sleep assistance method is provided, including:

[0009] Step 1: Collect piezoelectric signals through a sleeping pillow embedded with a thin-film piezoelectric sensor;

[0010] Step 2: Process the collected piezoelectric signals to obtain human physiological signals;

[0011] Step 3: Obtain a sleep stage time series based on the obtained human physiological signals;

[0012] Step 4: Determine sleep intervention measures through a BP neural network according to the sleep stage time series;

[0013] Step 5: Implement sleep intervention measures;

[0014] Step 6: Evaluate the effect of the sleep intervention measures.

[0015] Preferably, the thin-film piezoelectric sensor is embedded in the upper side of the sleeping pillow, and a flexible thin-film piezoelectric sensor is used.

[0016] Preferably, in any of the above solutions, Step 2 includes:

[0017] Step 21: Process the collected piezoelectric signals to obtain a respiration rate;

[0018] Step 22: Process the collected piezoelectric signals to obtain a heart rate;

[0019] Step 23: Process the collected piezoelectric signals to obtain body movement.

[0020] Preferably, in any of the above solutions, Step 21 includes:

[0021] Step 211: Separate the respiration signal from the piezoelectric signals;

[0022] Step 212: Find the peak value of each cycle of the respiration signal. When it is determined that the peak value is greater than a set respiration threshold, mark it as a respiration peak and record the index of the respiration peak;

[0023] Step 213: According to the formula

[0024]

[0025] calculate the respiration rate, where Y represents the respiration rate, with the unit of breaths per minute, d i represents the distance interval between adjacent respiration peak indices, f srepresents the sampling frequency in Hz, and n represents the total number of indices of the respiratory peak.

[0026] Preferably, in any of the above solutions, step 22 includes:

[0027] Step 221: Separate the heartbeat signal from the piezoelectric signal;

[0028] Step 222: Perform peak detection on the heartbeat signal and mark the detected peak as the default J peak;

[0029] Step 223: For each default J peak, take the data points within T time before and after it as a reference, and form an initial heartbeat signal template with the indexed data points and the default J peak;

[0030] Step 224: Cluster two types of heartbeat signal templates from all the initial heartbeat signal templates through a clustering algorithm;

[0031] Step 225: Take the average of the two types of heartbeat signal templates to obtain the optimal heartbeat signal template;

[0032] Step 226: Use a rectangular window to intercept the heartbeat signal with the same length as the optimal heartbeat signal template, and calculate the correlation coefficient between the optimal heartbeat signal template and the intercepted heartbeat signal;

[0033] Step 227: When the correlation coefficient is greater than the set correlation coefficient threshold, mark the intercepted heartbeat signal as the heartbeat signal including the true J peak and record its index;

[0034] Step 228: Move the rectangular window and repeat steps 226 - 228 until the heartbeat signal ends;

[0035] Step 229: According to the formula

[0036]

[0037] calculate the heart rate, where H represents the heart rate, i represents the total number of indices of the heartbeat signal including the true J peak, l represents the length of the heartbeat signal data, and f s represents the sampling frequency in Hz.

[0038] Preferably, in any of the above solutions, step 23 includes:

[0039] Step 231: Denoise the piezoelectric signal and divide it into multiple segments;

[0040] Step 232: For a segment of the piezoelectric signal, determine whether the amplitude of the signal exceeds the set body movement threshold. If so, mark it as a body movement signal and record it. If not, execute step 233;

[0041] Step 233: Calculate the energy of the signal, and determine whether there is a sudden change in the signal energy. If so, mark it as a body movement signal and record it; if not, execute Step 234;

[0042] Step 234: Read the next segment of piezoelectric signal, and repeatedly execute Step 232 - Step 234 until the piezoelectric signal ends.

[0043] Preferably, in any of the above solutions, Step 3 includes:

[0044] Step 31: Collect sound data, and identify snoring signals based on the collected sound data;

[0045] Step 32: Input the respiratory rate, heart rate, and body movement signals obtained in Step 2, and the snoring signals identified in Step 31 into the trained neural network to obtain a sleep stage time series.

[0046] Preferably, in any of the above solutions, the sleep stages include four types: wakefulness, light sleep, deep sleep, and rapid eye movement.

[0047] Preferably, in any of the above solutions, Step 4 includes:

[0048] Step 41: Divide the sleep stage time series according to sleep cycles;

[0049] Step 42: Process each sub - sleep stage time series of the divided sleep cycles to obtain sub - sleep stage time series of equal length;

[0050] Step 43: Input the sub - sleep stage time series of equal length into a BP neural network to determine sleep intervention measures.

[0051] Preferably, in Step 41, the sleep cycle is divided based on the rapid eye movement sleep stage. That is, starting from the sleep stage time series, the first rapid eye movement sleep stage and other sleep stages before it are divided into the first sleep cycle, the second rapid eye movement sleep stage and other sleep cycles after the first rapid eye movement sleep stage are divided into the second sleep cycle, the third rapid eye movement sleep stage and other sleep cycles after the second rapid eye movement sleep stage are divided into the third sleep cycle, and so on.

[0052] Preferably, in Step 42, for each sub - sleep stage time series of a sleep cycle, it is proportionally shortened according to the proportion of each sleep stage in it to obtain sub - sleep stage time series of equal length.

[0053] Preferably, the sleep intervention measures include at least one of playing white noise, playing pink noise, playing light music, playing low - frequency slow waves, vibrating, adjusting the smell, and adjusting the room temperature.

[0054] Preferably, in any of the above solutions, in step 6, the sleep quality score is calculated according to the proportion of each sleep stage in the sleep stage time series, and the implementation effect of the sleep intervention measure is evaluated according to the sleep quality score.

[0055] The second aspect of the present invention provides a sleep assistance system for implementing the sleep assistance method. The sleep assistance system includes:

[0056] A data acquisition module, which includes a pillow embedded with a thin-film piezoelectric sensor, configured to collect piezoelectric signals, and a sound sensor, configured to collect sound signals.

[0057] A data processing module, configured to process the collected piezoelectric signals and sound signals, determine the sleep intervention measure, and evaluate the implementation effect of the sleep intervention measure.

[0058] An intervention implementation module, configured to implement sleep intervention measures.

[0059] Preferably, the data processing module includes:

[0060] A human physiological signal processing sub-module, configured to obtain a respiration rate, a heart rate, and body movement signals according to the piezoelectric signals, and identify snoring signals according to the sound signals;

[0061] A sleep stage time series determination sub-module, configured to obtain a sleep stage time series based on the respiration rate, the heart rate, the body movement signals, and the snoring signals;

[0062] A sleep intervention measure determination sub-module, configured to determine sleep intervention measures according to the sleep stage time series;

[0063] An evaluation sub-module, configured to calculate a sleep quality score according to the sleep stage time series, and further evaluate the implementation effect of the sleep intervention measure according to the sleep quality score.

[0064] The sleep assistance method and system of the present invention have the following beneficial effects:

[0065] 1. It is a closed loop, including the entire processes of sleep monitoring, sleep intervention, and intervention evaluation;

[0066] 2. In the sleep monitoring stage, non-invasive monitoring is achieved, and there is no physical discomfort or sleep interference caused by wearing the device for a long time. It has the advantages of no load, convenience, and accuracy;

[0067] 3. Based on the non-invasive monitoring stage data, sleep intervention measures are determined, making the sleep intervention measures more targeted at the actual sleep situation;

[0068] 4. Evaluating the implementation effect of sleep intervention measures can timely detect problems and make improvements. BRIEF DESCRIPTION OF THE DRAWINGS

[0069] Figure 1 It is a schematic flowchart of a preferred embodiment of the sleep assistance method according to the present invention.

[0070] Figure 2 It is a schematic flowchart of step 21 of the sleep assistance method according to the present invention.

[0071] Figure 3 It is a schematic flowchart of step 22 of the sleep assistance method according to the present invention.

[0072] Figure 4 It is a schematic flowchart of step 23 of the sleep assistance method according to the present invention.

[0073] Figure 5 It is the sleep assistance method according to the present invention as Figure 1 shown in the schematic flowchart of step 3 of the embodiment.

[0074] Figure 6 It is the sleep assistance method according to the present invention as Figure 1 shown in the schematic flowchart of step 4 of the embodiment.

[0075] Figure 7 It is a schematic structural diagram of a preferred embodiment of the sleep assistance system according to the present invention.

[0076] Figure 8 It is the sleep assistance system according to the present invention as Figure 7 shown in the schematic structural diagram of the pillow of the embodiment.

[0077] Figure 9A It is a comparison chart of the breathing signal and heartbeat signal of Subject No. 1 on a certain night.

[0078] Figure 9B It is a comparison chart of the breathing signal and heartbeat signal of Subject No. 2 on a certain night. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0079] To better understand the present invention, the present invention will be described in detail below in conjunction with specific embodiments.

[0080] Embodiment 1

[0081] As Figure 1 shown, a sleep assistance method includes:

[0082] Step 1: Collect piezoelectric signals through a pillow embedded with a thin-film piezoelectric sensor;

[0083] Step 2: Process the collected piezoelectric signals to obtain human physiological signals;

[0084] Step 3: Obtain a sleep stage time series based on the obtained human physiological signals;

[0085] Step 4: Determine sleep intervention measures through a BP neural network according to the sleep stage time series;

[0086] Step 5: Implement sleep intervention measures;

[0087] Step 6: Evaluate the effect of the sleep intervention measures.

[0088] Preferably, in Step 1, the thin-film piezoelectric sensor is embedded in the upper side of the sleeping pillow, and a flexible thin-film piezoelectric sensor is used.

[0089] Step 2 includes:

[0090] Step 21: Process the collected piezoelectric signals to obtain a respiration rate;

[0091] Step 22: Process the collected piezoelectric signals to obtain a heart rate;

[0092] Step 23: Process the collected piezoelectric signals to obtain body movement.

[0093] Specifically, the original piezoelectric signals collected by the sleeping pillow embedded with the thin-film piezoelectric sensor are mixed signals, which include physiological signals such as cardioballism, respiration, and body movement, and are accompanied by electromagnetic interference and power frequency noise. Except for the body movement signal, both the cardioballism signal and the respiration signal are low-frequency signals, and the highest frequency does not exceed 10 Hz, while the power frequency noise and electromagnetic interference have higher frequencies, generally above 50 Hz. Therefore, to obtain the respiration rate and heart rate, the original piezoelectric signals are first processed (such as using Fourier transform, fast Fourier transform, etc.) to obtain frequency domain signals, the frequency bands outside 0 - 20 Hz are set to zero, and then inverse transformation is performed to obtain the piezoelectric signals after noise reduction. The piezoelectric signals after noise reduction are still mixed signals containing physiological signals such as cardioballism, respiration, and body movement.

[0094] As Figure 2 shown, Step 21 includes:

[0095] Step 211: Separate the respiration signal from the piezoelectric signals;

[0096] Step 212: Find the peak value of each cycle of the respiration signal. When it is determined that the peak value is greater than the set respiration threshold, mark it as a respiration peak and record the index of the respiration peak;

[0097] Step 213: According to the formula

[0098]

[0099] Calculate the respiratory rate, where Y represents the respiratory rate in breaths per minute, d i represents the distance interval between adjacent respiratory peak indices, f s represents the sampling frequency in Hz, and n represents the total number of indices of respiratory peaks.

[0100] Specifically, the normal respiratory rate range of humans is generally between 10 - 25 breaths per minute, corresponding to a respiratory frequency of 0.1 - 0.4 Hz. Therefore, in step 211, a respiratory signal can be separated from the noise-reduced piezoelectric signal through a band-pass filter (such as a Butterworth band-pass filter, Chebyshev filter, etc.). The energy contained in the human breathing motion is relatively large, so the amplitude and period of the separated respiratory signal are relatively obvious, and it is easier to locate the respiratory wave peaks in the time domain; at the same time, in order to eliminate the interference of small body movements (such as muscle twitches generated by the human body during sleep) on the respiratory signal, therefore, in step 212, for each peak of the found respiratory signal cycle, it is compared with a set respiratory threshold. When it exceeds the set respiratory threshold, the peak of this cycle is marked as a respiratory peak, and the index of the respiratory peak is recorded. Then, the respiratory rate is calculated according to the formula in step 213.

[0101] As Figure 3 shown, step 22 includes:

[0102] Step 221: Separate the heartbeat signal from the piezoelectric signal;

[0103] Step 222: Perform peak detection on the heartbeat signal, and mark the detected peak as the default J peak;

[0104] Step 223: For each default J peak, take the data points within T time before and after it as a reference, and form an initial heartbeat signal template with the indexed data points and this default J peak;

[0105] Step 224: Cluster two types of heartbeat signal templates from all the initial heartbeat signal templates through a clustering algorithm;

[0106] Step 225: Take the average of the two types of heartbeat signal templates to obtain the optimal heartbeat signal template;

[0107] Step 226: Use a rectangular window to intercept a heartbeat signal with the same length as the optimal heartbeat signal template, and calculate the correlation coefficient between the optimal heartbeat signal template and the intercepted heartbeat signal;

[0108] Step 227: When the correlation coefficient is greater than the set correlation coefficient threshold, mark the intercepted heartbeat signal as a heartbeat signal including the true J peak, and record its index;

[0109] Step 228: Move the rectangular window and repeat Steps 226 - 228 until the heartbeat signal ends;

[0110] Step 229: Calculate the heart rate according to the formula

[0111]

[0112] where \(H\) represents the heart rate, \(i\) represents the total number of indices of the heartbeat signal including the true J peak, \(l\) represents the length of the heartbeat signal data, and \(f\) s represents the sampling frequency, with the unit of Hz.

[0113] Specifically, the heartbeat signal reflects the mechanical pressure changes on the body surface during the heart beating process. The frequency of the heartbeat signal is generally between 1 - 10 Hz, and the frequency of the respiratory signal is 0.1 - 0.4 Hz, which is close to the frequency of the heartbeat signal. Therefore, the respiratory signal will cause greater interference to the heartbeat signal. In order to separate a better-quality heartbeat signal from the denoised piezoelectric signal, it is necessary to filter the interference signal. In this embodiment, preferably, considering that wavelet transform can perform multi-scale analysis on the low-frequency and high-frequency components of the signal and has the localization characteristic, which can effectively process non-stationary signals such as transients and mutations. Therefore, in Step 221, discrete wavelet transform (dB4, Level 6) is used to filter the interference signal of 0 - 0.5 Hz, and then a more real heartbeat signal is reconstructed.

[0114] Although the reconstructed heartbeat signal is closer to the real heartbeat signal, it is not an ideal ballistocardiogram (BCG) signal. This is because there will be a situation where the H peak value in a certain section of the reconstructed heartbeat signal is greater than the J peak value in another section, and this situation does not occur in the ideal heartbeat signal. The ballistocardiogram signal represents the time difference between two heartbeats through the JJ interval. Therefore, when calculating the heart rate, usually the J peak is found from the signal, and then the time interval between adjacent J peaks is calculated to obtain the heart rate. However, due to the situation where the H peak value in a certain section of the reconstructed ballistocardiogram signal is greater than the J peak value in another section, if the peak detection algorithm is directly used for the reconstructed ballistocardiogram signal to determine the J peak, it will lead to a large calculation error, and thus the heart rate detection is inaccurate. Therefore, the J peak in the reconstructed new impact signal is determined through Steps 222 - 228. In this embodiment, preferably, in Step 223, the \(T\) time can be set to 0.5 s; in Step 224, various clustering algorithms in the prior art can be used, such as the DBSCAN clustering algorithm, the OPTICS clustering algorithm, etc.; in Step 226, the correlation coefficient uses the Pearson correlation coefficient, and its calculation formula is:

[0115]

[0116] Among them, X(t) represents the heartbeat signal intercepted by the rectangular window, Y represents the optimal heartbeat signal template, r(t) represents the correlation coefficient, cov represents the covariance, and σ represents the standard deviation.

[0117] As Figure 4 shown, step 23 includes:

[0118] Step 231: Denoise the piezoelectric signal and divide it into multiple segments.

[0119] Step 232: For a segment of piezoelectric signal, determine whether the amplitude of the signal exceeds the set body movement threshold. If so, mark it as a body movement signal and record it. If not, execute step 233.

[0120] Step 233: Calculate the energy of the signal and determine whether the signal energy has mutated. If so, mark it as a body movement signal and record it. If not, execute step 234.

[0121] Step 234: Read the next segment of piezoelectric signal and repeat steps 232 - 234 until the piezoelectric signal ends.

[0122] Specifically, in different states of human sleep, the occurrence frequency and amplitude of body movements also vary. As the depth of sleep increases, the body movement frequency of a person decreases and the body movement amplitude decreases. Usually, when a person makes a movement, the amplitude of the piezoelectric signal will suddenly increase, and when the body movement ends, the amplitude of the piezoelectric signal will return to normal. Therefore, by setting a certain body movement threshold, when it is detected that the amplitude of the piezoelectric signal is greater than the set body movement threshold, it is considered that a body movement has occurred. However, whether the body movement threshold is set appropriately will affect the recognition of body movements, and the body movement amplitude will change in different sleep stages of humans. Therefore, it is difficult to identify all body movements during sleep through a general body movement threshold. In view of this, in step 233, for signals whose amplitude does not exceed the set body movement threshold, the signal energy will be calculated, and signals with mutated signal energy will be marked as body movement signals. Large - amplitude body movements are judged by the signal amplitude, and small - amplitude body movements are judged by the signal energy. The dual judgment of the signal amplitude and the signal energy can greatly improve the accuracy of body movement signal recognition.

[0123] It should be noted that for the piezoelectric signals collected by the pillow embedded with a thin - film piezoelectric sensor, when obtaining the respiratory rate, heart rate, and body movement signals, the signals need to be divided, such as into equal - length signal segments with lengths of 30s, 40s, 60s, etc. The signal division can be performed on the initial piezoelectric signal, or on the denoised piezoelectric signal, or on the separated respiratory signal and heartbeat signal. However, for the convenience of subsequent determination of the sleep staging time series, when dividing the signals, the lengths of all signals need to be the same and the starting time points need to be the same.

[0124] AsFigure 5 As shown, step 3 includes:

[0125] Step 31: collecting sound data, and identifying snoring signals according to the collected sound data;

[0126] Step 32: Input the respiratory rate, heart rate and body movement signals obtained in step 2 and the snoring signal identified in step 31 into the trained neural network to obtain a sleep stage time series.

[0127] It should be noted that the sleep stages include four types: wakefulness, light sleep, deep sleep and rapid eye movement.

[0128] Specifically, in step 31, the microphone is used to collect sound data and identify the snoring signal. When identifying the snoring signal, the sound signal also needs to be divided, and the length of the divided signal is consistent with the length of the piezoelectric signal, and the starting time point should be consistent.

[0129] Since the signal is divided, and the information of respiratory rate, heart rate, body movement and snoring is obtained for each segment of the divided signal, the respiratory rate, heart rate, body movement signal obtained in step 2 and the snoring signal obtained in step 31 are all time series arranged in chronological order, with HX = [hx1, hx2, ..., hx N ] represents the time series of respiratory rate, with XT = [xt1, xt2, …, xt N ] represents the time series of heart rate, TD = [td1, td2, …, td N ] represents the time series of body movement, HS = [hs1,hs2,…,hs N ] represents the time series of snoring. It should be noted that the values ​​of each element in the time series of respiratory rate and heart rate are numerical data; the values ​​of each element in the time series of body movement and snoring are level data representing the level of body movement / snoring, and the value of each data represents a different level. In step 32, hx1, xt1, td1 and hs1 are input into the trained neural network to obtain the corresponding sleep stage fq1, hx2, xt2, td2 and hs2 are input into the trained neural network to obtain the corresponding sleep stage fq2, and so on. N , xt N 、td N and hs N Input the trained neural network to get the corresponding sleep stage fq N , then fq1, fq2, ..., fq N It forms the sleep stage time series FQ, that is, FQ = [fq1, fq2, ..., fq NIt should be further noted that the values of the elements in the sleep stage time series are categorical data representing sleep stages. For example, 1 represents wakefulness, 2 represents light sleep, 3 represents deep sleep, and 4 represents rapid eye movement (REM). The trained neural network in step 32 can be a neural network in the prior art, and no specific limitation is imposed on it in this application.

[0130] As Figure 6 shown, step 4 includes:

[0131] Step 41: Divide the sleep stage time series according to sleep cycles.

[0132] Step 42: Process the sub-sleep stage time series of each divided sleep cycle to obtain sub-sleep stage time series of equal length.

[0133] Step 43: Input the sub-sleep stage time series of equal length into a BP neural network to determine sleep intervention measures.

[0134] Specifically, during sleep, under normal circumstances, the four sleep stages of wakefulness, light sleep, deep sleep, and rapid eye movement (REM) that appear in sequence constitute a sleep cycle. During a night's sleep, multiple sleep cycles may occur. The lengths of each sleep cycle may vary, and within each sleep cycle, the time lengths of each sleep stage may also be different. At the same time, the wakefulness sleep stage may be missing, but REM generally serves as the end of a sleep cycle. Therefore, in step 41, the sleep cycle is divided based on the REM sleep stage, that is, starting from the sleep stage time series, the first REM sleep stage and other sleep stages before it are divided into the first sleep cycle, the second REM sleep stage and other sleep cycles after the first REM sleep stage are divided into the second sleep cycle, the third REM sleep stage and other sleep cycles after the second REM sleep stage are divided into the third sleep cycle, and so on. For example, for a certain sleep stage time series FQ = [fq1, fq2,..., fq N, based on rapid eye movement as a benchmark, it is divided into three sleep cycles. The time series of sub-sleep stages corresponding to each sleep cycle are the first sub-sleep stage time series FQ1, the second sub-sleep stage time series FQ2, and the third sub-sleep stage time series FQ3 respectively. If the sleep stage time series FQ contains L sleep stage data, where the first sub-sleep stage time series FQ1 contains L1 sleep stage data, the second sub-sleep stage time series FQ2 contains L2 sleep stage data, and the third sub-sleep stage time series FQ3 contains L3 sleep stage data. It should be noted that before dividing the sleep stage time series, the sleep stage time series can be smoothed according to needs. For example, if a rapid eye movement sleep stage data suddenly appears in the light sleep stage, this rapid eye movement sleep stage data can be smoothed into light sleep stage data.

[0135] Since the time span of a night's sleep is relatively long, the number of data contained in the sleep stage time series is very large. Correspondingly, after the sleep cycle division, the number of data contained in the sleep stage time series corresponding to each sleep cycle is also very large; and because the time lengths of each sleep cycle are different, the lengths of the sleep stage time series corresponding to each sleep cycle also vary. Therefore, in step 42, for the time series of sub-sleep stages of each sleep cycle, according to the proportion of each sleep stage, it is equally proportionally shortened to obtain time series of sub-sleep stages of equal length. Specifically, it is set that the length of each shortened time series of sub-sleep stages is l + 1, and the last digit represents the length proportion of the current time series of sub-sleep stages in the entire sleep stage time series. Taking the first sub-sleep stage time series FQ1 as an example, the first l1 data are wakeful sleep stages, the next l2 data are light sleep stages, the next l3 data are deep sleep stages, and the last l4 data are rapid eye movement sleep stages; according to the proportion of each sleep stage in FQ1, it is equally proportionally shortened. Then, in the shortened first sub-sleep stage time series, the first data are wakeful sleep stages, and the next data are light sleep stages, and the next data are deep sleep stages, and the last The data for rapid eye movement sleep staging; the time series of other sub-sleep staging are similar. It should be noted that when shortening the time series of each sub-sleep staging, the number of data for each type of sleep staging should be rounded; if the data in the sleep staging time series is more regular, the length of the shortened time series of each sub-sleep staging can take a smaller value. Correspondingly, if the data in the sleep staging time series is more chaotic, the length of the shortened time series of each sub-sleep staging should take a larger value to retain more information. By shortening the time series of each sub-sleep staging, the data dimension at the input end of the BP network can be reduced, the calculation speed can be increased, and the calculation resources can be saved; the time length relationship of each sleep cycle can be reflected by the last data of the time series of each sub-sleep staging, and more data features are retained.

[0136] The BP neural network has strong non-linear mapping ability, self-learning and self-adaptive ability, generalization ability and fault tolerance ability. Therefore, the equal-length time series of each sub-sleep staging obtained in step 42 are input into the BP neural network in step 43 to determine the sleep intervention measures. The sleep intervention measures include at least one of playing white noise, playing pink noise, playing light music, playing low-frequency slow waves, vibrating, adjusting the smell, and adjusting the room temperature.

[0137] After obtaining the user's permission, implement the sleep intervention measures determined in step 4 and evaluate the implementation effect of the sleep intervention measures. Specifically, in step 6, calculate the sleep quality score according to the proportion of each sleep staging in the sleep staging time series, and evaluate the implementation effect of the sleep intervention measures according to the sleep quality score. If the sleep quality score is improved, it indicates that the sleep intervention measures are effective. If the sleep quality score is not improved, it indicates that the sleep intervention measures are ineffective, and then the parameters of the BP network need to be adjusted, and thus the sleep intervention measures given by it need to be adjusted.

[0138] Embodiment 2

[0139] As Figure 7 shown, a sleep assistance system for implementing the sleep assistance method, the sleep assistance system includes:

[0140] A data acquisition module 1, which includes a pillow embedded with a thin-film piezoelectric sensor configured to acquire piezoelectric signals, and a sound sensor configured to acquire sound signals.

[0141] A data processing module 2 configured to process the acquired piezoelectric signals and sound signals, determine the sleep intervention measures, and evaluate the implementation effect of the sleep intervention measures.

[0142] An intervention implementation module 3 configured to implement the sleep intervention measures.

[0143] Preferably, the data processing module 2 includes:

[0144] A human physiological signal processing sub-module 21, which is configured to obtain a respiration rate, a heart rate, and a body movement signal according to the piezoelectric signal, and to identify a snoring signal according to the sound signal;

[0145] A sleep stage time series determination sub-module 22, which is configured to obtain a sleep stage time series according to the respiration rate, the heart rate, the body movement signal, and the snoring signal;

[0146] A sleep intervention measure determination sub-module 23, which is configured to determine a sleep intervention measure according to the sleep stage time series;

[0147] An evaluation sub-module 24, which is configured to calculate a sleep quality score according to the sleep stage time series, and further to evaluate the implementation effect of the sleep intervention measure according to the sleep quality score.

[0148] As Figure 8 shown in the structure of the sleeping pillow embedded with the thin-film piezoelectric sensor, its main body uses memory foam 101, and the thin-film piezoelectric sensor 102 is embedded on the upper side of the sleeping pillow. A PCB circuit board 103 is also provided inside the sleeping pillow main body. Only a wireless transmission module may be provided on the PCB circuit board 103 for sending the physiological signals collected by the thin-film piezoelectric sensor to the data processing module disposed outside the sleeping pillow; or the data processing module may also be disposed on the PCB circuit board 103; or a part of the data processing module is disposed on the PCB circuit board 103 inside the sleeping pillow, and a wireless transmission module is provided on the PCB circuit board 103 for transmitting the output to other modules of the data processing module disposed outside the sleeping pillow. No specific limitation is made in this application. It should be noted that a sound sensor is preferably provided on the PCB circuit board 103 for collecting sound data during sleep, and further for identifying a snoring signal according to the sound data.

[0149] Preferably, in the present application, the thin-film piezoelectric sensor is a flexible thin-film piezoelectric sensor, and its main component material is polypropylene foaming material, which is made by mixing and pressing main raw materials such as polypropylene resin, polyethylene resin and natural stone powder. The principle of this thin-film piezoelectric sensor is as follows: After reasonable polarization, the PP film has obvious piezoelectric activity and the hysteresis loop of ferroelectric polymer. This material has the characteristics of both ferroelectric materials and electrets. The piezoelectric characteristics of such materials are due to the special cavity structure in the material and the space charges (macroscopic electric dipoles) with opposite polarities deposited on the opposite walls of the cavity, that is, these hole structures in the PP film can be polarized under a high-voltage electric field to generate piezoelectric effect. The thickness of this thin-film piezoelectric sensor is extremely thin, only 254 microns, and it will not interfere with sleep when embedded in a sleeping pillow, and can realize non-contact and non-sensing acquisition of the user's piezoelectric data during sleep, and then monitor breathing, heart rate and body movement.

[0150] Example 3

[0151] In order to verify the accuracy of extracting heart rate, respiratory rate and body movement signals from the piezoelectric signals collected by the thin-film piezoelectric sensor in the sleep assistance method, an experiment was carried out.

[0152] In the experiment, a sleeping pillow embedded with the thin-film piezoelectric sensor was used to collect the piezoelectric data of the subjects during sleep, and based on the collected piezoelectric data, the respiratory rate, heart rate and body movement signals of the subjects were obtained by using the method in step 2; at the same time, a PSG multi-lead sleep monitor was used to collect the respiratory rate, heart rate and body movement conditions of the subjects during sleep, and a comparison was made.

[0153] A total of 8 subjects participated in the experiment, and each subject understood the purpose, process and precautions of the experiment before the experiment. Each subject collected sleep data for no less than 5 nights, and the data collection time per night was no less than 6 hours.

[0154] Figure 9A and Figure 9B respectively show the comparison diagrams of the respiratory signals and heartbeat signals of a certain night of Subject 1 and Subject 2. Through Figure 9A and Figure 9B it can be seen that the respiratory signals and heartbeat signals obtained from the piezoelectric data collected by the sleeping pillow embedded with the thin-film piezoelectric sensor have a high overall degree of coincidence with the respiratory signals and heartbeat signals collected by the PSG multi-lead sleep monitor. The data coincidence degrees of other subjects and other times are also relatively high, and they are not shown one by one in this application.

[0155] Taking the respiratory rate, heart rate and body movement times of the subjects collected by the PSG multi-lead sleep monitor during sleep as a reference, the average errors of the respiratory rate, heart rate and body movement times collected by each subject through the sleeping pillow are shown in Table 1.

[0156] Table 1 Average errors of heart rate, respiratory rate and body movement times monitored by the sleeping pillow

[0157]

[0158]

[0159] It can be found from Table 1 that the maximum error of the respiratory rate collected by the pillow embedded with the thin-film piezoelectric sensor is 13.63%, the maximum error of the heart rate is 9.12%, and the maximum error of the body movement times is 8.26%. Therefore, in this application, the accuracy of obtaining the user's physiological parameters from the piezoelectric data collected by the pillow embedded with the thin-film piezoelectric sensor can reach more than 86%, which has a relatively high accuracy.

[0160] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit them; although the foregoing embodiments have described the present invention in detail, those skilled in the art should understand that they can modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements on some or all of the technical features, and these replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the present invention.

Claims

1. A sleep aid method, characterized in that: include: Step 1: Collect piezoelectric signals through a pillow embedded with a thin film piezoelectric sensor; Step 2: Process the collected piezoelectric signal to obtain human physiological signals; Step 3: Obtain the sleep stage time series according to the obtained human physiological signals; Step 4: Determine sleep intervention measures through BP neural network according to the sleep stage time series; Step 5: Implement sleep interventions; Step 6: Evaluate the effectiveness of sleep interventions; Step 2 includes: Step 21: Process the collected piezoelectric signal to obtain the respiratory rate; Step 22: Process the collected piezoelectric signal to obtain the heart rate; Step 23: Process the collected piezoelectric signal to obtain body movement; Step 3 includes: Step 31: collecting sound data, and identifying snoring signals according to the collected sound data; Step 32: input the respiratory rate, heart rate and body movement signals obtained in step 2 and the snoring signal identified in step 31 into the trained neural network to obtain a sleep stage time series; Step 4 includes: Step 41: Divide the sleep stage time series according to the sleep cycle to obtain a plurality of sub-sleep stage time series; Step 42: Processing the sub-sleep stage time series of each divided sleep cycle to obtain sub-sleep stage time series of equal length; Step 43: Input the sub-sleep stage time series of equal length into the BP neural network to determine the sleep intervention measures; In step 42, the sleep stage sub-sequences of each sleep cycle are shortened in equal proportion according to the proportion of each sleep stage therein, so as to obtain sleep stage sub-sequences of equal length.

2. The sleep aid method according to claim 1, characterized in that: The thin film piezoelectric sensor is embedded in the upper side of the pillow and adopts a flexible thin film piezoelectric sensor.

3. The sleep aid method according to claim 1, characterized in that: Step 21 includes: Step 211: Separating the breathing signal from the piezoelectric signal; Step 212: Find the peak value of each cycle of the respiratory signal, and when it is determined that the peak value is greater than the set respiratory threshold, mark it as a respiratory peak value, and record the index of the respiratory peak value; Step 213: According to the formula Calculate the respiratory rate, where Y represents the respiratory rate in times / minute. d j Indicates the distance interval between adjacent respiratory peak indices, f s1 Indicates the sampling frequency in Hz. n The total number of indices representing respiratory peaks.

4. The sleep aid method according to claim 1, characterized in that: Step 22 includes: Step 221: Separating the heartbeat signal from the piezoelectric signal; Step 222: performing peak detection on the heartbeat signal, and marking the detected peak as a default J peak; Step 223: for each default J peak, use it as a reference, index the data points within T time before and after, and use the indexed data points and the default J peak to form an initial heartbeat signal template; Step 224: clustering two types of heartbeat signal templates from all initial heartbeat signal templates through a clustering algorithm; Step 225: taking the average of the two types of heartbeat signal templates to obtain the optimal heartbeat signal template; Step 226: using a rectangular window to intercept a heartbeat signal having the same length as the optimal heartbeat signal template, and calculating a correlation coefficient between the optimal heartbeat signal template and the intercepted heartbeat signal; Step 227: when the correlation coefficient is greater than the set correlation coefficient threshold, mark the intercepted heartbeat signal as a heartbeat signal including a true J ​​peak, and record its index; Step 228: Move the rectangular window and repeat steps 226 to 228 until the heartbeat signal ends; Step 229: According to the formula Calculate heart rate, where H represents the heart rate, i represents the total number of indices of heartbeat signals including the true J ​​peak, l Indicates the heartbeat signal data length. f s2 Indicates the sampling frequency in Hz.

5. The sleep aid method according to claim 1, characterized in that: Step 23 includes: Step 231: reducing noise on the piezoelectric signal and dividing it into multiple segments; Step 232: for a piezoelectric signal, determine whether the amplitude of the signal exceeds the set body motion threshold, if so, mark it as a body motion signal and record it, if not, execute step 233; Step 233: Calculate the energy of the signal and determine whether the signal energy changes suddenly. If so, mark it as a body motion signal and record it. If not, execute step 234. Step 234: read the next piezoelectric signal, and repeat steps 232 to 234 until the piezoelectric signal ends.

6. A sleep aid system, characterized in that: For implementing the sleep assistance method according to any one of claims 1 to 5, the sleep assistance system comprises: A data acquisition module, comprising a sleeping pillow embedded with a thin film piezoelectric sensor, configured to acquire piezoelectric signals, and a sound sensor, configured to acquire sound signals; A data processing module, which is configured to process the collected piezoelectric signals and sound signals, determine the sleep intervention measures, and evaluate the implementation effect of the sleep intervention measures; an intervention implementation module configured to implement sleep interventions; The data processing module comprises: a human physiological signal processing submodule, which is configured to obtain a respiratory rate, a heart rate and a body movement signal according to the piezoelectric signal, and to identify a snoring signal according to the sound signal; A sleep stage time series determination submodule, which is configured to obtain a sleep stage time series based on a respiratory rate, a heart rate, and a body movement signal, as well as a snoring signal; a sleep intervention measure determination submodule, configured to determine a sleep intervention measure according to the sleep stage time series; The evaluation submodule is configured to calculate a sleep quality score according to the sleep stage time series, and then evaluate the implementation effect of the sleep intervention measures according to the sleep quality score.

Images