Sleeping point detection method
By calculating the standard deviation and number of physical movements of the electrocardiogram, breathing and triaxial acceleration signals, and combining the length and duration of the characteristic window, the sleeping point is determined, which solves the problems of high misjudgment rate and equipment limitation of sleeping point detection in the prior art. Higher accuracy and reliability are achieved.
Patent Information
- Application Number
- CN202110178505.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2021-02-07
- Publication Date
- 2024-12-10
- Estimated Expiration
- 2041-02-07
AI Technical Summary
Existing sleep point detection methods, such as polysomnography (PSG), have problems such as expensive equipment, low efficiency, limited site, needing professional doctors to judge, and high misjudgment rates, especially the lack of universality and science of the number of flips. in accordance with.
The heart rate, breathing rate and three-axis acceleration standard deviation are calculated by the subject's electrocardiogram, breathing signal and three-axis acceleration signal for a predetermined time period, and the alignment and sliding window processing are performed, the number of body movements and average values are calculated, and the length of the characteristic window is combined with the length of the characteristic window. and duration, determine the sleeping period and sleeping point, and use specific rules to determine the sleeping candidate points and the final sleeping point.
It improves the accuracy and reliability of sleeping point detection, reduces the misjudgment rate, and can be compared with the judgment results of PSG physicians, indicating that the method has high accuracy and reliability.
Smart Images
Figure CN112971724B_ABST
Abstract
Description
Technical Field
[0001] The present application relates to sleep detection technology, and in particular to a method for detecting a sleeping point. Background Art
[0002] Sleeping time occupies one-third of human life. Poor sleep quality can lead to physical and mental fatigue during the day. Sleep disorders can lead to depression, diabetes, hypertension and many other cardiovascular diseases. As the starting point for judging sleep quality, the sleep time plays a very important role in accurately judging sleep quality.
[0003] The current "gold standard" for sleep monitoring is polysomnography (PSG), which can accurately detect the subject's falling asleep point based on the monitoring of EEG, ECG, EMG, eye movement, breathing and blood oxygen signals, providing a basic guarantee for accurately judging sleep quality. However, PSG requires subjects to wear EEG electrodes in the sleep laboratory, which makes the subjects feel uncomfortable and affects their sleep to a certain extent. In addition, the use of PSG equipment is expensive, inefficient, limited in space, and requires professional doctors to judge, which greatly limits the use of PSG.
[0004] Patent application number 201510371905.4 proposes a solution for detecting the falling asleep point, which monitors the subject's tossing and turning times and heart rate within 15 minutes, and then sets a threshold for the number of tossing and heart rate. By comparing the number of tossing and heart rate within 15 minutes with the threshold, the corresponding falling asleep point is obtained. This method is too simple and can easily lead to misjudgment. Each subject has different turning habits and individual differences. The setting of the threshold for the number of tossing is extremely non-universal and has no scientific basis. Summary of the invention
[0005] In view of the above problems, the present application aims to propose a method for detecting the sleeping point.
[0006] The present application provides a method for detecting a sleeping point, which calculates the subject's heart rate, respiratory rate and standard deviation of three-axis acceleration in a predetermined time period through the subject's electrocardiogram signal, respiratory signal and three-axis acceleration signal in the predetermined time period; aligns the calculated heart rate, respiratory rate and standard deviation of three-axis acceleration to obtain aligned heart rate, respiratory rate and standard deviation of three-axis acceleration; slides the aligned heart rate, respiratory rate and standard deviation of three-axis acceleration in a feature window of a predetermined time length, and calculates the number of body movements, heart rate average value and standard deviation of respiratory rate in each feature window; determines a time period in which the number of body movements is less than a predetermined number and the duration exceeds a predetermined long time as a sleeping period; calculates a candidate sleeping point for each of the sleeping periods; and determines a sleeping point from the candidate sleeping points.
[0007] Preferably, the time length of the characteristic window of predetermined time length is 20 minutes.
[0008] Preferably, a time period in which the number of body movements is less than 30 and the duration is more than 40 minutes is determined as the sleeping period.
[0009] Preferably, the calculated alignment of the heart rate, respiratory rate and triaxial acceleration standard deviation is an alignment within a time window with a width of 1 sec.
[0010] Preferably, a sliding window is performed on the aligned heart rate, respiratory rate and triaxial acceleration standard deviation in units of 1 sec.
[0011] Preferably, for each sleeping period, a candidate sleeping point is calculated by using the heart rate mean value and the breathing rate standard deviation.
[0012] Preferably, in each sleep onset period, the time point when at least one of the heart rate mean value or the respiratory rate standard deviation drops to 70% of its maximum value is a candidate sleep onset point.
[0013] Preferably, the electrocardiogram signal, respiratory signal, and triaxial acceleration signal in the predetermined time period are the electrocardiogram signal, respiratory signal, and triaxial acceleration signal from 18:00:00 on the first day to 10:00:00 on the second day.
[0014] Preferably,
[0015] The sleep period in which there is both a sleep point generated by a decrease in the mean heart rate and a sleep point generated by a decrease in the standard deviation of the respiratory rate is a common sleep period;
[0016] If there is no common sleep period, the most forward candidate sleeping point is determined as the sleeping point;
[0017] If there is only one common sleep period, calculate the time error of the sleep point generated by the decrease in the heart rate average value and the sleep point generated by the decrease in the respiratory rate standard deviation. If the time error is less than or equal to 10 minutes, determine the relatively early sleep candidate point in the common sleep period as the sleep point; if the time error is greater than 10 minutes, check whether there is a sleep period before the common sleep period. If so, determine the most early sleep candidate point in the previous sleep period as the sleep point. If not, determine the early sleep candidate point in the common sleep period as the sleep point;
[0018] If there are at least two common sleep periods, first determine whether there is a common sleep period before 02:00:00 the next day;
[0019] If there is a common sleep period before 02:00:00 of the next day: select the most preceding common sleep period whose time error of the candidate sleeping point is less than or equal to 10 minutes from the common sleep periods before 02:00:00 of the next day, record the time error, if the time error is 0, determine whether there is another sleeping period before the common sleep period, if so, retain the preceding candidate sleeping point in the preceding sleeping period as the sleeping point, if not, retain the preceding candidate sleeping point in the common sleep period as the sleeping point; if the sleeping point errors of the common sleep periods before 02:00:00 of the next day are all greater than 10 minutes, record the time error, and output the preceding candidate sleeping point in the sleeping period with the smallest time error as the sleeping point;
[0020] If there is no common sleep period before 02:00:00 of the next day: select the earliest common sleep period with a time error of the candidate sleeping point less than or equal to 10 minutes from the common sleep periods after 02:00:00 of the next day, record the time error, if the time error is 0, determine whether there is a sleep period before the common sleep period, if so, retain the earlier sleep candidate point in the previous sleep period as the sleep point, if not, retain the earlier sleep candidate point in the common sleep period as the sleep point; if the time errors of the candidate sleeping points in the common sleep periods after 02:00:00 of the next day are all greater than 10 minutes, record the time error, and determine the earlier sleep candidate point in the common sleep period with the smallest time error as the sleep point.
[0021] The sleep onset point detection method of the present application determines the sleep onset period and candidate sleep onset point by calculating the average heart rate, standard deviation of respiratory rate and number of body movements, then determines the sleep onset point by using the candidate sleep onset point judgment rule, and compares the algorithm judgment result with the PSG physician judgment result, indicating that this method has high accuracy and reliability. BRIEF DESCRIPTION OF THE DRAWINGS
[0022] Figure 1 This is a flow chart of the sleeping point detection method of the present application;
[0023] Figure 2 A flowchart for confirming a sleeping point of the sleeping point detection method of the present application;
[0024] Figure 3 The original ECG signal, respiratory signal and triaxial acceleration signal captured from 18:00 on the first day to 10:00 on the second day in one embodiment;
[0025] Figure 4 for Figure 3 A local magnified view of the signal;
[0026] Figure 5 This is a diagram of the sleep onset detection result of the first example (one sleep onset period);
[0027] Figure 6 is a sleeping point detection result diagram of the second example (multiple sleeping periods);
[0028] Figure 7 This is a diagram of the sleeping point detection result of the third example (the sleeping point does not exist). DETAILED DESCRIPTION
[0029] Below, the present application is described in detail with reference to the accompanying drawings.
[0030] The method of the present application uses the three characteristics of the mean of heart rate, the standard deviation of breathing rate, and the number of body movements within the window to identify the falling asleep point. Figure 1 shown.
[0031] 1. Intercept the physiological signals of continuous dynamic monitoring from 18:00 to 10:00 (the next day), including: ECG, respiration, triaxial acceleration signals and corresponding timestamps. If the wearing time does not cover this time period, intercept the time within this time period to provide the corresponding timestamp.
[0032] 2. Calculate the standard deviation of heart rate, respiratory rate and three-axis acceleration, and correspond the calculation results to every second through interpolation and windowing. Since the standard deviation of heart rate, respiratory rate and acceleration calculated based on the intercepted continuous dynamic physiological signal is 1 / second, and the sampling rates of ECG signal, respiratory signal and acceleration are different, the interpolation method is used to achieve the correspondence with heart rate and respiratory rate in the window every second, that is, the sampling rate of heart rate, respiratory rate and three-axis acceleration standard deviation is 1HZ.
[0033] The standard deviation of the three-axis acceleration is the sum of the standard deviations of the acceleration of each axis per second, as shown in the following formula.
[0034] d_std=sum(std(x_1sec)+std(y_1sec)+std(z_1sec))
[0035] 3. Filter outliers: If the standard deviation of acceleration is greater than 0.08, it is marked as a body movement. The number of body movements in each window is calculated by sliding the window second by second. The characteristic window length is 20 minutes. If the number of body movements in the window is ≥50, it is marked as 50 to facilitate visualization and avoid the influence of outliers, such as those caused by signal noise or irregular human activities.
[0036] 4. Calculate the mean heart rate and standard deviation of respiration rate in each feature window, and correspond the mean heart rate and standard deviation of respiration rate calculated for each window length to the number of body movements calculated in the window.
[0037] Window length body movement times: the number of times the acceleration standard deviation d_std>0.08 within the window.
[0038] 5. Calculate the possible falling asleep period (also called sleep period) by the number of body movements.
[0039] The time period when the number of body movements is less than 30 and the interval length is greater than 40 minutes is determined as a possible sleeping period. This step is to find the time period when the number of body movements is less than 30 and the interval length is greater than 40 minutes by sliding the window step by step.
[0040] 6. For each calculated sleep period, the possible sleep onset point is calculated using the mean heart rate and standard deviation of breathing rate.
[0041] 7. Confirm your sleeping point
[0042] By calculating the sleep monitoring data through the above-mentioned method of calculating the number of body movements, the average heart rate, and the standard deviation of the respiratory rate, there may be multiple sleep periods. For example, 1) during sleep, you may get up to go to the bathroom. Then, through the above-mentioned method of calculating the sleep period, this time period will be judged as the awakening period, and the previous and later periods of this time period will be judged as the sleeping period, resulting in multiple sleeping periods; 2) There may also be misjudgments. For example, if a person goes to bed very early and lies there without moving, then this time period may be misjudged as the sleeping period. Then, he does other things and then falls asleep, which will also result in multiple sleeping periods. At the same time, this also shows that the sleeping point is not in the first sleeping period; 3) There are multiple awakening and sleeping time periods during the entire sleep process, that is, there will be multiple awakenings and multiple falling asleep. Therefore, it is necessary to calibrate the possible sleeping period by the average heart rate or the standard deviation of the respiratory rate to determine the true sleeping period, and find the sleeping point during the sleeping period to avoid misjudgment as much as possible. The process of the sleeping point confirmation method is as follows Figure 2 shown.
[0043] 1) Search the standard deviation of breathing rate (std_BR) and the mean heart rate (mean_HR) in each sleep period to find possible candidate points for falling asleep.
[0044] If there is no candidate point, that is, during the sleep period, the standard deviation of the respiratory rate (std_BR) or the mean heart rate (mean_HR) does not drop to 70% of the maximum standard deviation of the respiratory rate or the maximum mean heart rate during the whole process, then the sleep onset point does not exist;
[0045] During the sleep period, if there is at least one point where the standard deviation of the respiratory rate (std_BR) or the mean heart rate (mean_HR) drops to 70% of the maximum standard deviation of the respiratory rate or the maximum mean heart rate during the whole process, there is a candidate point for falling asleep.
[0046] 2) Determine whether there is a common sleeping point during the sleep period.
[0047] The common sleep period refers to a point in a sleep period when the standard deviation of respiratory rate (std_BR) and the mean heart rate (mean_HR) both drop to 70% of the maximum standard deviation of respiratory rate or the maximum mean heart rate during the whole process.
[0048] Only one public sleep period
[0049] Calculate the bias (error) of the current common period, that is, calculate the point in the current common sleep period when the heart rate average value or the respiratory rate standard deviation drops to 70% of the maximum respiratory rate standard deviation or the maximum heart rate average value in the whole process.
[0050] a) If the difference is less than or equal to 10 minutes, then the earlier sleep onset point in the common sleep period is output, that is, the time point when the heart rate average value or the respiratory rate standard deviation first drops to 70% of the maximum heart rate average value or the maximum respiratory rate standard deviation is the sleep onset point;
[0051] b) If the difference is greater than 10 minutes, find out whether there is another sleep period before the common sleep period. If not, keep the current result, that is, the time point when the average heart rate or standard deviation of breathing rate in the common sleep period first drops to 70% of its maximum heart rate average or maximum breathing rate standard deviation is the sleep onset point; if there is another sleep period before the common sleep period, the time point when the average heart rate or standard deviation of breathing rate in the previous sleep period first drops to 70% of its maximum heart rate average or maximum breathing rate standard deviation is the sleep onset point.
[0052] There are at least 2 public sleep periods
[0053] Since most people fall asleep before 02:00:00, priority is given to whether there is a common sleep period before 02:00:00. If so, and a reasonable sleeping point can be found through the algorithm (the sleeping time is within the range of sleep rules), the common sleep period after 02:00:00 will no longer be considered.
[0054] ①There is a public sleep period before 02:00:00
[0055] a) If the error of the sleeping point is less than or equal to 10 minutes during the common sleep period, the earlier point will be retained (recorded as the first common sleep period: the first sleep period with the sleeping point error less than or equal to 10 minutes) and the error (bias) will be recorded.
[0056] With error (bias ≠ 0, and less than or equal to 10 minutes): The time point when the heart rate or respiration in the first common sleep period first drops to 70% of the maximum heart rate or maximum respiration rate is the sleep onset point. For example, if there is at least one common sleep period before 02:00:00 with a sleep onset error of less than or equal to 10 minutes and not 0, the time point when the average heart rate or standard deviation of respiration rate in the earlier sleep period first drops to 70% of its maximum heart rate average or maximum respiration rate standard deviation is selected as the sleep onset point.
[0057] No error (bias = 0): If the time point when the average heart rate or standard deviation of breathing rate in the first common sleep period drops to 70% of its maximum average heart rate or maximum standard deviation of breathing rate coincides with the time point, check whether there is a sleep period before the first common sleep period. If so, output the earlier sleep point in the previous sleep period (the sleep period before the first common sleep period) (the time point when the average heart rate or standard deviation of breathing rate first drops to 70% of its maximum average heart rate or maximum standard deviation of breathing rate is the sleep point); if not, output the sleep point of the first common period (the time point when the average heart rate or standard deviation of breathing rate first drops to 70% of its maximum average heart rate or maximum standard deviation of breathing rate is the sleep point).
[0058] b) Before 02:00:00, the error of falling asleep at the point in the common sleep period is greater than 10 minutes, and the bias is recorded. Find the common sleep period with the smallest bias, and output the earlier point in the common sleep period (the time point when the heart rate average or respiratory rate standard deviation first drops to 70% of its maximum heart rate average or maximum respiratory rate standard deviation is the falling asleep point).
[0059] ②The public sleep period starts after 02:00:00
[0060] a) If the error of the sleep onset point is less than or equal to 10 minutes during the common sleep period, keep the earlier point (the first sleep period after 02:00:00 with a sleep onset point less than or equal to 10 minutes, recorded as the first common period) and record the bias.
[0061] With error (bias ≠ 0, and less than or equal to 10 minutes): The time point when the average heart rate or standard deviation of breathing rate in the first common sleep period first drops to 70% of its maximum heart rate average or maximum breathing rate standard deviation is the sleep onset point. For example, if there is at least one sleep onset point error in the common sleep period after 02:00:00 that is less than or equal to 10 minutes and not 0, the time point when the average heart rate or standard deviation of breathing rate in the earlier sleep period first drops to 70% of its maximum heart rate average or maximum breathing rate standard deviation is selected as the sleep onset point.
[0062] No error (bias = 0): If the time point when the average heart rate or standard deviation of breathing rate in the first common sleep period drops to 70% of its maximum average heart rate or maximum standard deviation of breathing rate coincides with the time point, check whether there is a sleep period before the first common sleep period. If so, output the earlier sleep point in the previous sleep period (the sleep period before the first common sleep period) (the time point when the average heart rate or standard deviation of breathing rate first drops to 70% of its maximum average heart rate or maximum standard deviation of breathing rate is the sleep point); if not, output the sleep point of the first common period (the time point when the average heart rate or standard deviation of breathing rate first drops to 70% of its maximum average heart rate or maximum standard deviation of breathing rate is the sleep point).
[0063] b) After 02:00:00, if the error of falling asleep during the common sleep period is greater than 10 minutes, record the bias. Find the common sleep period with the smallest bias and output the earlier point in the common sleep period (the time point when the heart rate average or respiratory rate standard deviation first drops to 70% of its maximum heart rate average or maximum respiratory rate standard deviation is the falling asleep point).
[0064] No public sleeping period
[0065] No common sleep period means that in the same sleep period, the heart rate average or respiratory rate standard deviation does not appear at the same time (i.e., only one of them appears). At this time, the time point when the heart rate average or respiratory rate standard deviation drops to 70% of its maximum heart rate average or maximum respiratory rate standard deviation is the sleep onset point.
[0066] Combination Figure 5-7 The example shown is used for explanation.
[0067] 1. Intercept the physiological signal data from 18:00 to 10:00 (the next day), including: ECG signal, respiratory signal and triaxial acceleration signal. The original signal data is visualized as follows: Figure 3 As shown, the local amplified signal is Figure 4 shown.
[0068] 2. According to the intercepted physiological signals, calculate the respiratory rate, heart rate and acceleration standard deviation, and correspond the calculation results to every second through interpolation and window. Since the value of the heart rate, respiratory rate and acceleration standard deviation calculated based on the intercepted continuous dynamic physiological signals is 1 / second, the heart rate, respiratory rate and standard deviation are corresponded to every second in the window through interpolation, that is, the sampling rate of the heart rate, respiratory rate and three-axis acceleration standard deviation is 1HZ.
[0069] 3. Calculate the number of body movements in each window: acceleration standard deviation > 0.08 is marked as one body movement. The number of body movements in each window is calculated by sliding the window second by second. The characteristic window length is 20 minutes. If the number of body movements in the window is ≥ 50, it is marked as 50 for easy visualization and to avoid the influence of outliers, such as those caused by signal noise or irregular human activities.
[0070] 4. Calculate the mean heart rate and standard deviation of respiration rate in each feature window, and correspond the mean heart rate and standard deviation of respiration rate calculated for each window length to the number of body movements calculated in the window.
[0071] Window length body movement times: the number of times the acceleration standard deviation d_std>0.08 within the window.
[0072] 5. Calculate the possible sleep period (also called sleeping period) by the number of body movements. The time period with less than 30 body movements and a length greater than 40 minutes is considered as the possible sleep period. This step is to find the time period with less than 30 body movements and a length greater than 40 minutes by sliding the window step by step.
[0073] 6. For each calculated sleep period, the possible sleep onset point is calculated by the heart rate average and the respiratory rate standard deviation. That is, the heart rate average and the respiratory rate standard deviation in each sleep period are calculated, and the possible sleep onset point in the sleep period is determined by the heart rate average or the respiratory rate standard deviation.
[0074] 7. Confirm your sleeping point
[0075] According to the above rules, the sleep onset point is confirmed. The analysis results of each monitoring data are represented by four sub-graphs. The first two are the mean heart rate (meanHR) and the standard deviation of the respiratory rate (stdBR), the third is the number of body movements (NumACC), and the fourth is the three-axis acceleration signal (ACC). The yellow line represents the detection result of the automatic detection algorithm of the sleep onset point, the black line represents the PSG physician marking result, and the red and green lines of the third sub-graph represent the start and end time points of each sleep period.
[0076] 1) Wang XX, male, height: 180, weight: 75kg, age 30, intercepted the ECG, respiration and triaxial acceleration signals from 18:00 to 10:00 (the next day) and processed and analyzed the signals to confirm the falling asleep point, such as Figure 5 shown.
[0077] The physician marked the sleep onset time as 00:10:00, and the sleep onset time automatically detected as 00:14:49, with a difference of about 5 minutes in the sleep onset time, indicating that the algorithm has high accuracy and reliability.
[0078] 2) Hao XX, male, height 175cm, weight 80kg, age 45, intercepted the ECG, respiration and triaxial acceleration signals from 18:00 to 10:00 (the next day) and processed and analyzed the signals to confirm the falling asleep point, such as Figure 6 shown.
[0079] The doctor marked the sleep time as 20:32:30, and the automatic sleep time detection was 20:19:38, with a difference of about 13 minutes, indicating that the algorithm has high accuracy and reliability. The automatic sleep time detection algorithm detected two sleep periods and marked the second sleep period. The sleep time determined was 20:19:38, which is very close to the doctor's marked time, indicating that the algorithm is more reasonable to mark the time point in the second sleep period.
[0080] 3) Wang XX, male, height: 170cm, weight: 65kg, age: 38 years old, intercept the ECG, respiration and triaxial acceleration signals from 18:00 to 10:00 (the next day) of his monitoring data, process and analyze the signals, and confirm the sleeping point, such as Figure 7 shown.
[0081] Depend on Figure 7 Analysis shows that the number of body movements (NumACC) in this data is ≥50 within the intercepted data segment. According to the sleep point judgment rule, there is no sleep period, that is, there is no sleep point. It may be because there is a problem with the collected monitoring data, such as large signal noise or the signal collection sensor is detached, or it may be that the person did not sleep all night.
[0082] Unless otherwise defined, all technical and / or scientific terms used in this application have the same meaning as commonly understood by one of ordinary skill in the art to which the invention relates. The materials, methods and examples mentioned in this application are illustrative only and not limiting.
[0083] Although the present invention has been described in conjunction with specific embodiments, within the spirit of the invention of this application, those skilled in the art may make appropriate substitutions, modifications and changes, and such substitutions, modifications and changes still fall within the scope of protection of this application.
Claims
1. A method for detecting a sleeping point, which calculates the heart rate, respiratory rate and standard deviation of three-axis acceleration of a subject in a predetermined time period through the subject's electrocardiogram signal, respiratory signal and three-axis acceleration signal in the predetermined time period; aligns the calculated heart rate, respiratory rate and standard deviation of three-axis acceleration to obtain aligned heart rate, respiratory rate and standard deviation of three-axis acceleration; slides the aligned heart rate, respiratory rate and standard deviation of three-axis acceleration in a feature window of a predetermined time length, and calculates the number of body movements, the average heart rate and the standard deviation of respiratory rate in each feature window; determines a time period in which the number of body movements is less than a predetermined number and the duration exceeds a predetermined long time as a sleeping period; calculates a candidate sleeping point for each sleeping period; and determines a sleeping point from the candidate sleeping points; For each sleep period, the candidate sleep point is calculated by the heart rate mean value and the respiratory rate standard deviation; in each sleep period, the time point when at least one of the heart rate mean value or the respiratory rate standard deviation drops to 70% of its maximum value is the candidate sleep point; The sleep period in which there is both a sleep point generated by a decrease in the mean heart rate and a sleep point generated by a decrease in the standard deviation of the respiratory rate is a common sleep period; If there is no common sleep period, the most forward candidate sleeping point is determined as the sleeping point; If there is only one common sleep period, calculate the time error of the sleep point generated by the decrease in the heart rate average value and the sleep point generated by the decrease in the respiratory rate standard deviation. If the time error is less than or equal to 10 minutes, determine the relatively early sleep candidate point in the common sleep period as the sleep point; if the time error is greater than 10 minutes, check whether there is a sleep period before the common sleep period. If so, determine the most early sleep candidate point in the previous sleep period as the sleep point. If not, determine the early sleep candidate point in the common sleep period as the sleep point; If there are at least two common sleep periods, first determine whether there is a common sleep period before 02:00:00 the next day; If there is a common sleep period before 02:00:00 of the next day: select the most preceding common sleep period whose time error of the candidate sleeping point is less than or equal to 10 minutes from the common sleep periods before 02:00:00 of the next day, record the time error, if the time error is 0, determine whether there is another sleeping period before the common sleep period, if so, retain the preceding candidate sleeping point in the preceding sleeping period as the sleeping point, if not, retain the preceding candidate sleeping point in the common sleep period as the sleeping point; if the sleeping point errors of the common sleep periods before 02:00:00 of the next day are all greater than 10 minutes, record the time error, and output the preceding candidate sleeping point in the sleeping period with the smallest time error as the sleeping point; If there is no common sleep period before 02:00:00 of the next day: select the earliest common sleep period with a time error of the candidate sleeping point less than or equal to 10 minutes from the common sleep periods after 02:00:00 of the next day, record the time error, if the time error is 0, determine whether there is a sleep period before the common sleep period, if so, retain the earlier sleep candidate point in the previous sleep period as the sleep point, if not, retain the earlier sleep candidate point in the common sleep period as the sleep point; if the time errors of the candidate sleeping points in the common sleep periods after 02:00:00 of the next day are all greater than 10 minutes, record the time error, and determine the earlier sleep candidate point in the common sleep period with the smallest time error as the sleep point.
2. The method according to claim 1, characterized in that: The time length of the characteristic window of the predetermined time length is 20 minutes.
3. The method according to claim 1, characterized in that: The period when the number of body movements was less than 30 and the duration was more than 40 minutes was considered as the sleeping period.
4. The method according to claim 1, characterized in that: The calculated alignment of the heart rate, respiratory rate and triaxial acceleration standard deviation is within a time window with a width of 1 sec.
5. The method according to claim 4, characterized in that: A sliding window is performed on the aligned heart rate, respiratory rate and triaxial acceleration standard deviation in units of 1 sec.
6. The method according to claim 1, characterized in that: The electrocardiogram signal, respiratory signal, and triaxial acceleration signal in the predetermined time period are the electrocardiogram signal, respiratory signal, and triaxial acceleration signal from 18:00:00 on the first day to 10:00:00 on the second day.