Sleep quality analysis system and method based on medical bed sensor

Abstract

The invention belongs to the technical field of sleep quality analysis, and particularly relates to a sleep quality analysis system and method based on a medical bed sensor, so as to solve the technical problem that an existing method excessively depends on duration statistics and neglects sleep continuity and periodic rhythmicity. The analysis method comprises the following steps: S1, acquiring sign signal data acquired by a sensor on a medical bed in a sleep cycle, and dividing the sign signal data into a plurality of continuous time slices; s2, extracting physiological indexes of respiratory rate, heart rate and body movement amplitude in each time slice, and calculating variation depth of the corresponding time slice; s3, generating a variation depth baseline; calculating an unsteady state disturbance factor according to the occurrence frequency and the duration of the micro-awakening disturbance fragment; and S4, generating a sleep quality comprehensive score. According to the analysis method, the comprehensiveness and objectivity of the evaluation result are improved, and the sleep health condition is reflected more truly.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of sleep quality analysis, and particularly relates to a sleep quality analysis system and method based on a medical bed sensor. BACKGROUND

[0002] Traditional sleep quality evaluation mainly relies on the gold standard, polysomnography (PSG), which can divide sleep stages and diagnose various sleep disorders by pasting electrodes on the body of the subject to collect multiple physiological signals such as electroencephalogram, electrooculogram, electromyogram, and electrocardiogram. However, PSG equipment is expensive and complicated to operate, and the monitoring process is highly invasive, requiring the subject to be in a professional sleep laboratory environment, and the cables connected to the whole body can seriously interfere with normal sleep, resulting in monitoring results that may not fully reflect the natural sleep state.

[0003] In contrast, sleep monitoring solutions based on non-contact sensors, such as pressure, vibration, or piezoelectric sensors installed under a medical bed or in a mattress, can extract basic physiological indicators such as respiratory rate, heart rate, and body movement by sensing the weak vibration signals caused by breathing, heartbeat, and body posture changes during sleep. However, generating an evaluation report by simply counting the duration of each sleep stage is intuitive but rough, ignoring two key dimensions in the sleep process: one is the continuity and stability of sleep, i.e., avoiding transient disturbance events such as micro-awakening and body movement that disrupt the sleep structure; the other is the rhythmicity of the sleep cycle, i.e., unable to represent the degree of deviation between the actual sleep cycle structure and the standard healthy sleep cycle template. Therefore, the existing sleep quality analysis system based on a medical bed sensor has deficiencies in the comprehensiveness and accuracy of the evaluation model, and the score results are often difficult to fully reflect the true sleep quality. SUMMARY

[0004] The present application provides a sleep quality analysis system and method based on a medical bed sensor to solve the technical problem that existing methods rely too much on duration statistics and ignore the continuity and periodicity of sleep.

[0005] In a first aspect, the present application provides a sleep quality analysis method based on a medical bed sensor, comprising the following steps: S1, acquiring the physiological signal data collected by the medical bed sensor in a sleep cycle and dividing it into multiple continuous time segments; S2, extracting the physiological indicators of respiratory rate, heart rate, and body movement amplitude in each time segment, and based on a pre-set sleep stage physiological indicator correlation matrix, preliminarily determining the sleep stage of the corresponding time segment; at the same time, calculating the normalized power spectrum entropy of the physiological indicators in each time segment as the variation depth of the corresponding time segment; S3. Based on the mean variation depth of the same sleep stage in historical data and the preset stage sensitivity coefficient, a variation depth baseline is generated; when the variation depth of a certain time segment exceeds the variation depth baseline of the corresponding sleep stage, the time segment is marked as a micro-awakening disturbance segment; the non-steady-state disturbance factor is calculated according to the frequency and duration of the micro-awakening disturbance segment. S4. The determined sleep stage sequence is matched with the standard sleep cycle template by time regularization, and the cycle rhythm deviation is calculated. The distribution duration of deep sleep, light sleep and REM sleep stages in the whole sleep cycle is statistically analyzed to generate the basic evaluation coefficient. The basic evaluation coefficient, cycle rhythm deviation and non-steady-state disturbance factor are fused and non-linear weighted calculation is performed to generate a comprehensive sleep quality score.

[0006] Furthermore, in S1, the vital sign data is divided into non-overlapping window functions with a time segment length of 30 seconds to generate a series of continuous time segment sequences.

[0007] Furthermore, physiological indicators of respiratory rate, heart rate, and body movement amplitude were extracted for each time segment, including: Bandpass filtering is applied to the signal in each time segment to separate heartbeat and respiratory signals; heart rate and respiratory rate are obtained by detecting the peak points of the filtered signals and calculating the peak intervals; the standard deviation of the original signal amplitude within a time segment is calculated as the body motion amplitude index for the corresponding time segment.

[0008] Further, the normalized power spectral entropy of the physiological indicators within each time segment is calculated, including: Perform a Fast Fourier Transform on the heart rate and respiratory rate change sequence within the time segment to obtain the power spectral density distribution; then, extract the energy values ​​of each frequency component of the power spectrum. Divide by total energy The normalized probability is obtained. This forms a probability distribution; according to the Shannon entropy formula... Calculate the information entropy of the probability distribution, and use the information entropy as the mutation depth, where, This represents each individual frequency component in the power spectral density distribution. It represents the total number of all frequency components in the power spectral density distribution.

[0009] Further, when generating the variation deep baseline, at least 50 historical segments of the same sleep stage as the current time segment in the historical database are queried, and an arithmetic mean of variation deep of the historical segments is calculated; the arithmetic mean is multiplied by an adjustment factor, and the adjustment factor is equal to 1 plus a sensitivity coefficient of the corresponding sleep stage, wherein the sensitivity coefficient of the light sleep stage is 0.2, the sensitivity coefficient of the deep sleep stage is 0.3, and the sensitivity coefficient of the rapid eye movement sleep stage is 0.25, and the variation deep baseline of the sleep stage is generated.

[0010] Further, according to the occurrence frequency and the duration of the micro-waking disturbance segment, a non-steady-state disturbance factor is calculated, including: The total number of micro-waking disturbance segments in the entire sleep cycle is counted, and divided by the total sleep hours to obtain the occurrence frequency ; The average duration of all micro-waking disturbance segments is calculated ; The non-steady-state disturbance factor is calculated by normalizing and weighting the sum of the occurrence frequency and the average duration , and the calculation formula is: ; Wherein, and are preset normalization reference values of the frequency and the duration, and are corresponding weight coefficients, and .

[0011] Further, when calculating the cycle rhythm deviation degree, the “deep sleep”, “light sleep”, “rapid eye movement” and “wakefulness” in the sleep stage are respectively mapped to the numerical values 4, 3, 2 and 1, so as to form a sleep stage numerical sequence; The sleep stage numerical sequence is time-regularly matched with a standard sleep cycle template sequence which is also subjected to numerical mapping, to obtain a cumulative regular path cost; The cumulative regular path cost is divided by the length of the matching path, and is normalized, and the value is taken as the cycle rhythm deviation degree.

[0012] Further, when generating the basic evaluation coefficient, the proportions of the deep sleep, the light sleep and the rapid eye movement sleep stage in the total sleep time in the entire sleep cycle are counted, and are respectively denoted as , , ; The proportions are weighted and summed by a preset weight to obtain the basic evaluation coefficient , and the calculation formula of the basic evaluation coefficient is: ; wherein, 、 、 are weight coefficients of each sleep stage, and the sum is 1.

[0013] Further, the calculation formula when generating the comprehensive score of sleep quality is: ; wherein, is the comprehensive score, is a basic evaluation coefficient, is a periodic rhythm deviation degree, is a non-steady-state disturbance factor; 、 、 are weight coefficients of each item, and are non-linear adjustment indexes.

[0014] In a second aspect, the present application provides a sleep quality analysis system based on a sensor on a medical bed, comprising a memory and a processor, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the sleep quality analysis method based on the sensor on the medical bed is realized.

[0015] The beneficial effects are: the present application calculates the normalized power spectrum entropy of the physiological indicators of the sign signals as the variation depth, which is used to analyze the subtle fluctuations in the sleep process, and identifies the micro-awakening disturbance events that cause damage to the sleep continuity based on the variation depth baseline, and then generates the non-steady-state disturbance factor to evaluate the stability of the sleep. At the same time, the actual sleep stage sequence of the user is time-regularly matched with the standard sleep cycle template, and the periodic rhythm deviation degree is calculated to represent the regularity and rhythm of the sleep structure. By nonlinearly fusing and weighting the basic evaluation coefficient reflecting the sleep amount, the non-steady-state disturbance factor reflecting the stability, and the periodic rhythm deviation degree representing the rhythm, the comprehensive score of sleep quality obtained by the present application not only contains the evaluation of the sleep amount, but also integrates the quality of sleep, thereby improving the comprehensiveness and objectivity of the evaluation result, and more truly reflecting the sleep health status. BRIEF DESCRIPTION OF DRAWINGS

[0016] Figure 1 is a flowchart of the sleep quality analysis method based on the sensor on the medical bed; Figure 2 is a schematic diagram of physiological indicator extraction and stage determination; Figure 3 is a schematic diagram of signal acquisition and segmentation; Figure 4 is a schematic diagram of the comprehensive score of sleep quality; Figure 5A schematic diagram of a sleep quality analysis system based on sensors on a medical bed. DETAILED DESCRIPTION

[0017] An embodiment of the sleep quality analysis method based on sensors on a medical bed provided by the present application includes the following steps: As shown in the accompanying drawings, Figure 1 a sleep quality analysis based on sensors on a medical bed includes the following steps: S1, acquiring the vital sign signal data collected by the sensors on the medical bed within a sleep cycle, and dividing it into multiple continuous time segments.

[0018] Specifically, by deploying a piezoelectric film sensor under the medical bed mattress, continuous acquisition of pressure change signals generated by breathing, heartbeat and body movement during the entire night sleep of the user forms the original vital sign signal time series data; the original vital sign signal time series data is divided into multiple continuous time segments according to a fixed time window (for example, every 30s), and each segment is used as a basic unit for subsequent analysis.

[0019] In an optional embodiment, in S1, the vital sign signal data is divided into a series of continuous time segments with a time segment length of 30s.

[0020] Specifically, a complete sleep cycle data is obtained from the piezoelectric film sensor of the medical bed, for example, a continuous signal stream with a length of 8h, assuming a sampling frequency of 50Hz, a total of 1.44 million data points are obtained. Set a fixed time window length, i.e. 30s, corresponding to 1500 data points. As Figure 3 , the entire signal stream is sliced in a non-overlapping manner, the first time segment contains the first to the 1500th data point, the second time segment contains the 1501st to the 3000th data point, and so on until the end of the signal stream. The continuous 8h vital sign signal data is converted into a sequence composed of 960 independent time segments with a length of 30s, and each segment is a basic unit for subsequent physiological index analysis.

[0021] S2, extracting the respiratory rate, heart rate and body movement amplitude physiological indexes in each time segment, and based on the pre-set sleep stage physiological index correlation matrix, preliminarily determining the sleep stage of the corresponding time segment; at the same time, calculating the normalized power spectrum entropy of the physiological indexes in each time segment as the variation depth of the corresponding time segment.

[0022] Specifically, for each 30-second time segment of data, digital signal processing techniques are first employed. Low-frequency respiratory waves are extracted using low-pass filtering, and the respiratory rate is obtained by calculating the peak interval of the respiratory waves. Higher-frequency cardiac impulse signals are separated using band-pass filtering, and the heart rate is obtained using a peak detection algorithm. The amplitude of body movement is represented by calculating the energy or variance of the signal. For example... Figure 2 The extracted respiratory rate, heart rate, and body movement amplitude are used as a feature vector and input into a classification model that has been trained with a large amount of clinical data, namely the preset sleep stage physiological index association matrix (such as decision tree or support vector machine), so as to preliminarily determine the most likely sleep stage of the time segment, such as deep sleep, light sleep, REM sleep, or wakefulness.

[0023] Meanwhile, taking heart rate as an example, a continuous heartbeat interval sequence is extracted within this time segment, and a fast Fourier transform is performed on this sequence to obtain the power spectral density. This power spectral density is then normalized so that the sum of the power of all frequency components is one, forming a probability distribution. The information entropy of this probability distribution is calculated based on the Shannon entropy formula, and the resulting entropy value is the variation depth of this time segment. This value can represent the complexity and uncertainty of heart rate changes.

[0024] In an optional embodiment, physiological parameters such as respiratory rate, heart rate, and body movement amplitude are extracted for each time segment, including: Bandpass filtering is applied to the signal in each time segment to separate heartbeat and respiratory signals; heart rate and respiratory rate are obtained by detecting the peak points of the filtered signals and calculating the peak intervals; the standard deviation of the original signal amplitude within a time segment is calculated as the body motion amplitude index for the corresponding time segment.

[0025] The generated 30-second time segment data is then processed for signal separation and index extraction. Specifically, a bandpass filter of 0.8 to 2.5 Hz is used to extract the weak vibration signal caused by heartbeats (i.e., cardiac impact signal); simultaneously, another bandpass filter of 0.1 to 0.5 Hz is used to extract the respiratory signal caused by chest cavity fluctuations. In the filtered cardiac impact signal, a peak detection algorithm is used to identify all heartbeat peaks within 30 seconds. For example, if 35 peaks are detected, the average time interval between peaks (e.g., 0.86 s) is used to calculate a heart rate of approximately 70 beats per minute. The same method is applied to the respiratory signal; if 8 respiratory peaks are detected, the respiratory rate is calculated to be approximately 16 breaths per minute. Returning to the unfiltered original 30-second signal segment, the standard deviation of the amplitude of 1500 data points is calculated. This standard deviation is used as an index of body movement amplitude for that segment. For example, the standard deviation is 0.1 when the user's body is stable, but it may increase to 1.5 if there are larger movements such as turning over.

[0026] In an optional embodiment, calculating the normalized power spectral entropy of the physiological indicator within each time segment includes: Perform a Fast Fourier Transform on the heart rate and respiratory rate change sequence within the time segment to obtain the power spectral density distribution; then, extract the energy values ​​of each frequency component of the power spectrum. Divide by total energy The normalized probability is obtained. This forms a probability distribution; according to the Shannon entropy formula... Calculate the information entropy of the probability distribution, and use the information entropy as the mutation depth, where, This represents each individual frequency component in the power spectral density distribution. It represents the total number of all frequency components in the power spectral density distribution.

[0027] Taking a heart rate variability sequence as an example, based on the time intervals of the detected consecutive heartbeat peaks, a heartbeat interval sequence is first generated, for example, sequences of 0.85s, 0.87s, and 0.84s within a 30s segment. A Fast Fourier Transform is then performed on the interval sequence to transform the time domain to the frequency domain, resulting in a power spectral density map. This map shows the energy distribution of heart rate variability at different frequencies. Subsequently, the energy value of each frequency component is calculated. And add up the energy of all components to get the total energy. By using the energy of each component Divide by total energy This gives the proportion of that component in the total energy, i.e., the normalized probability. ,all The sum of these probabilities is 1. Substituting these probability values ​​into the Shannon entropy formula, the resulting information entropy value is the variation depth of that time segment. For example, if the energy is mainly concentrated on a few frequencies, a lower entropy value, such as 1.2, is calculated, indicating simple variability; conversely, if the energy is evenly distributed across multiple frequencies, a higher entropy value, such as 3.5, is obtained, indicating complex variability.

[0028] S3. Based on the mean variation depth of the same sleep stage in historical data and the preset stage sensitivity coefficient, a variation depth baseline is generated. When the variation depth of a certain time segment exceeds the variation depth baseline of the corresponding sleep stage, the time segment is marked as a micro-awakening disturbance segment. The non-steady-state disturbance factor is calculated according to the occurrence frequency and duration of the micro-awakening disturbance segment.

[0029] Specifically, historical sleep data is analyzed in advance to calculate the average level of variation depth under different sleep stages such as deep sleep and light sleep. This average level is then multiplied by a set stage sensitivity coefficient (e.g., 1.2 for deep sleep and 1.5 for light sleep) to generate a specific variation depth threshold, i.e., variation depth baseline, for each sleep stage.

[0030] When analyzing the current sleep, the depth of variation calculated for each time segment is compared with the baseline of the corresponding sleep stage. If the depth of variation of the current segment is greater than the corresponding baseline, the sleep is considered to be disturbed, and the time segment is marked as a micro-wake disturbance segment.

[0031] After analyzing the entire sleep cycle, the total number of all segments marked as micro-awakening disturbances is counted and divided by the total sleep duration to obtain the average number of disturbances per hour. The total duration of the time segments is summed, and a comprehensive non-steady-state disturbance factor is calculated by using a weighting formula, such as adding the normalized average number of disturbances to the proportion of the total disturbance duration. The higher the value of the non-steady-state disturbance factor, the more unstable and fragmented the sleep process is.

[0032] In an optional embodiment, when generating the variation depth baseline, at least 50 historical segments in the historical database that belong to the same sleep stage as the current time segment are queried, and the arithmetic mean of the variation depth of the historical segments is calculated; the arithmetic mean is multiplied by an adjustment factor, which is equal to 1 plus the sensitivity coefficient of the corresponding sleep stage, wherein the sensitivity coefficient of the light sleep stage is 0.2, the sensitivity coefficient of the deep sleep stage is 0.3, and the sensitivity coefficient of the REM sleep stage is 0.25, to generate the variation depth baseline of the sleep stage.

[0033] Specifically, assuming the current time segment is classified as deep sleep with a variation depth value of 2.1, the database storing the user's historical sleep data is accessed to retrieve all past time segments marked as deep sleep. The 60 most recent segments are then selected randomly or chronologically. The variation depth values ​​of these 60 historical segments are read (e.g., 1.9, 2.2, 2.0, etc.), and their arithmetic mean is calculated, assuming a result of 2.05. Since the current stage is deep sleep, the corresponding sensitivity coefficient is 0.3, therefore the adjustment factor is calculated to be 1.3. The variation depth baseline is obtained by multiplying the arithmetic mean by the adjustment factor, resulting in 2.665. This value of 2.665 is the personalized judgment threshold for the current deep sleep segment, used for subsequent comparisons to determine whether the physiological state of the current segment is stable.

[0034] In an optional embodiment, the nonsteady-state perturbation factor is calculated based on the frequency and duration of the micro-awakening perturbation segments, including: The frequency of micro-awakening disturbances is calculated by dividing the total number of micro-awakening disturbances throughout the entire sleep cycle by the total number of sleep hours. ; Calculate the average duration of all micro-awakening perturbation segments. ; By analyzing the frequency of occurrence and average duration Perform normalized weighted summation to calculate the unsteady-state perturbation factor. The calculation formula is: ; in, and These are the preset normalized baseline values ​​for frequency and duration, respectively. and For the corresponding weight coefficients, and .

[0035] Specifically, after the entire night's sleep monitoring is completed, all 30-second segments identified as micro-awakening disturbances are summarized. Assuming a 7.5-hour sleep period, a total of 15 micro-awakening disturbance segments were detected. Frequency of occurrence... That is, 2 times per hour. Since each segment is 30 seconds long, and the 15 segments may occur sporadically, their average duration is calculated. If they are 15 independent 30-second segments, the average duration is... That is, 30 seconds. Set the baseline values ​​and weights required for standardization, such as frequency baseline values. 4 times / hour, duration baseline For 60 seconds, weight It is 0.7. The value is 0.3. Substitute this value into the formula to calculate the unsteady-state disturbance factor. The calculated result is 0.5.

[0036] S4. The determined sleep stage sequence is matched with the standard sleep cycle template by time regularization, and the cycle rhythm deviation is calculated. The distribution duration of deep sleep, light sleep and REM sleep stages in the whole sleep cycle is statistically analyzed to generate the basic evaluation coefficient. The basic evaluation coefficient, cycle rhythm deviation and non-steady-state disturbance factor are fused and non-linear weighted calculation is performed to generate a comprehensive sleep quality score.

[0037] Specifically, firstly, the circadian rhythm deviation is calculated: the sleep stage determination results of the whole night (e.g., the sequence of light sleep-deep sleep-light sleep-REM sleep) are taken as an actual sleep rhythm sequence; an ideal sleep cycle template sequence is built in, which is constructed from data of healthy people. This template usually contains several complete 90-minute sleep cycles; a time warping algorithm is used to calculate the minimum matching cost between the actual sequence and the ideal template sequence. After normalization, the cost is the circadian rhythm deviation, which represents the difference between the user's sleep cycle structure and the standard health model.

[0038] Secondly, generate basic evaluation coefficients: traverse all time segments throughout the night, count the segments that are determined to be deep sleep, light sleep, and REM sleep, and multiply each count value by the duration of a single segment (e.g., 30 seconds) to calculate the total duration of each sleep stage; use the ratio of key indicators such as deep sleep duration and REM sleep duration to the total sleep duration as core parameters to form basic evaluation coefficients.

[0039] Finally, a nonlinear weighted calculation is performed: the three dimensions of indicators obtained above—basic evaluation coefficient, circadian rhythm deviation, and non-steady-state disturbance factor—are standardized; the three standardized values ​​are then input into a preset nonlinear fusion model, such as a small neural network or a fuzzy logic system. This model can intelligently weight the indicators according to different combinations. For example, when both the circadian rhythm deviation and non-steady-state disturbance factor are poor, the final score will be lowered, even if the deep sleep duration reaches the target. The model will output a comprehensive sleep quality score from 0 to 100.

[0040] In an optional embodiment, when calculating the circadian rhythm deviation, the sleep stages “deep sleep”, “light sleep”, “REM sleep”, and “wake” are mapped to the values ​​4, 3, 2, and 1, respectively, thereby forming a numerical sequence of sleep stages. The cumulative normalized path cost is obtained by performing time-warping matching between the numerical sequence of sleep stages and the standard sleep cycle template sequence that has also undergone numerical mapping. The cumulative normalized path cost is divided by the length of the matching path and normalized. This value is used as the periodic rhythm deviation.

[0041] Specifically, the analysis results of sleep stages throughout the night, such as wakefulness, light sleep, deep sleep, and REM sleep, are converted into a numerical sequence of sleep stages. According to the mapping rules, this sequence may be 1, 3, 3, 4, 2, etc. Simultaneously, a template sequence representing the ideal sleep rhythm is built-in; for example, a complete 90-minute cycle might be represented as 1, 3, 3, 4, 4, 4, 3, 2, 2. A time warping algorithm is used to compare the user's actual sleep sequence with the ideal template sequence. This algorithm finds an optimal alignment path that minimizes the sum of numerical differences between corresponding points in the two sequences. This minimum sum of differences is the cumulative warping path cost, assumed to be 250. The algorithm also records the length of the optimal alignment path, assumed to be 300. Normalizing the cost by dividing by the path length yields a value of 0.83, which is the circadian rhythm deviation. The smaller the value, the closer the user's sleep cycle structure is to the ideal pattern.

[0042] In an optional embodiment, when generating the baseline evaluation coefficient, the proportion of total sleep time in deep sleep, light sleep, and REM sleep stages throughout the entire sleep cycle is calculated and denoted as follows: , , ; The basic evaluation coefficient is obtained by weighting and summing the proportions using preset weights. Basic evaluation coefficient The calculation formula is: ; in, , , These are the weighting coefficients for each sleep stage, and their sum is 1.

[0043] Specifically, after analyzing sleep data from the entire night, the total duration of each major sleep stage is calculated. For example, on a night with a total sleep time of 450 minutes, deep sleep lasts 90 minutes, light sleep lasts 240 minutes, and REM sleep lasts 120 minutes. The proportion of each stage to the total sleep time is then calculated. =0.2; ≈0.533; ≈0.267. The weights of different sleep stages on sleep quality are preset, for example, the weight of deep sleep. The weight of light sleep is 0.5. The weight of REM sleep is 0.2. The sum of the three is 1, and the initial value is 0.3. Substitute these values ​​into the formula to calculate the basic evaluation coefficient. The calculated result is approximately 0.2867.

[0044] In an optional embodiment, the formula for generating a comprehensive sleep quality score is as follows: ; in, For comprehensive scoring, Based on the evaluation coefficient, This represents the degree of deviation from the periodic rhythm. It is an unsteady-state perturbation factor; , , These are the weighting coefficients for each item. and It is a non-linear adjustment index.

[0045] Specifically, the previously calculated baseline evaluation coefficients were integrated in the final fusion stage. Circadian rhythm deviation and unsteady-state disturbance factors A total score is generated using multiple dimensions of sleep indicators. For example... Figure 4 Assuming the basic evaluation coefficients have been obtained The deviation from the periodic rhythm is 0.2867. It is 0.83, and the unsteady-state perturbation factor The value is 0.5. A corresponding weight and non-linear adjustment index are set for the indicator, for example, It is 200. It is 15. For 20, and the index It is 1.2. The value is 1.5. Nonlinear adjustments are made to the deviation and disturbance factor, and the calculated value is... Approximately equal to 0.799; calculated It is approximately equal to 0.354. Substituting all the values ​​into the formula, the user's overall sleep quality score is 38.275.

[0046] An embodiment of the sleep quality analysis system based on medical bed sensors provided by this invention: like Figure 5 As shown, the sleep quality analysis system based on medical bed sensors includes a processor and a memory. The memory stores computer program instructions, which, when executed by the processor, implement the aforementioned sleep quality analysis method based on medical bed sensors.

[0047] The sleep quality analysis system based on sensors in medical beds also includes other components well known to those skilled in the art, such as communication interfaces. Their settings and functions are known in the art and will not be described in detail here.

[0048] In addition, in the description of this specification, "multiple" means at least two, such as two, three or more, etc., unless otherwise expressly and specifically defined.

Claims

1. A sleep quality analysis method based on sensors in a medical bed, characterized in that, Includes the following steps: S1, acquire vital sign signal data collected by sensors on the medical bed during a sleep cycle, and divide it into multiple consecutive time segments; S2, extract the physiological indicators of respiratory rate, heart rate and body movement amplitude in each time segment, and preliminarily determine the sleep stage of the corresponding time segment based on the preset physiological indicator correlation matrix of sleep stage; at the same time, calculate the normalized power spectral entropy of the physiological indicators in each time segment as the variation depth of the corresponding time segment. S3. Based on the mean variation depth of the same sleep stage in historical data and the preset stage sensitivity coefficient, a variation depth baseline is generated; when the variation depth of a certain time segment exceeds the variation depth baseline of the corresponding sleep stage, the time segment is marked as a micro-awakening disturbance segment; the non-steady-state disturbance factor is calculated according to the frequency and duration of the micro-awakening disturbance segment. S4. The determined sleep stage sequence is matched with the standard sleep cycle template by time regularization, and the cycle rhythm deviation is calculated. The distribution duration of deep sleep, light sleep and REM sleep stages in the whole sleep cycle is statistically analyzed to generate the basic evaluation coefficient. The basic evaluation coefficient, cycle rhythm deviation and non-steady-state disturbance factor are fused and non-linear weighted calculation is performed to generate a comprehensive sleep quality score.

2. The sleep quality analysis method based on medical bed sensors according to claim 1, characterized in that, In S1, the vital sign data is divided into non-overlapping window functions with a time segment length of 30 seconds to generate a series of continuous time segment sequences.

3. The sleep quality analysis method based on medical bed sensors according to claim 1, characterized in that, Physiological parameters such as respiratory rate, heart rate, and body movement amplitude were extracted for each time segment, including: Bandpass filtering is applied to the signal in each time segment to separate heartbeat and respiratory signals; heart rate and respiratory rate are obtained by detecting the peak points of the filtered signals and calculating the peak intervals; the standard deviation of the original signal amplitude within a time segment is calculated as the body motion amplitude index for the corresponding time segment.

4. The sleep quality analysis method based on medical bed sensors according to claim 3, characterized in that, Calculate the normalized power spectral entropy of the physiological indicators within each time segment, including: Perform a Fast Fourier Transform on the heart rate and respiratory rate change sequence within the time segment to obtain the power spectral density distribution; then, extract the energy values ​​of each frequency component of the power spectrum. Divide by total energy The normalized probability is obtained. This forms a probability distribution; according to the Shannon entropy formula... Calculate the information entropy of the probability distribution, and use the information entropy as the mutation depth, where, This represents each individual frequency component in the power spectral density distribution. It represents the total number of all frequency components in the power spectral density distribution.

5. The sleep quality analysis method based on medical bed sensors according to claim 1, characterized in that, When generating the variation depth baseline, at least 50 historical segments in the historical database that belong to the same sleep stage as the current time segment are queried, and the arithmetic mean of the variation depth of the historical segments is calculated. The arithmetic mean is multiplied by an adjustment factor, which is equal to 1 plus the sensitivity coefficient of the corresponding sleep stage, wherein the sensitivity coefficient of the light sleep stage is 0.2, the sensitivity coefficient of the deep sleep stage is 0.3, and the sensitivity coefficient of the REM sleep stage is 0.25, and the variation depth baseline of the sleep stage is generated.

6. The sleep quality analysis method based on medical bed sensors according to claim 5, characterized in that, Based on the frequency and duration of micro-awakening perturbation segments, the non-steady-state perturbation factor is calculated, including: The frequency of micro-awakening disturbances is calculated by dividing the total number of micro-awakening disturbances throughout the entire sleep cycle by the total number of sleep hours. ; Calculate the average duration of all micro-awakening perturbation segments. ; By analyzing the frequency of occurrence and average duration Perform normalized weighted summation to calculate the unsteady-state perturbation factor. The calculation formula is: ; in, and These are the preset normalized baseline values ​​for frequency and duration, respectively. and For the corresponding weight coefficients, and .

7. The sleep quality analysis method based on medical bed sensors according to any one of claims 1-6, characterized in that, When calculating the deviation of the circadian rhythm, the sleep stages of "deep sleep", "light sleep", "rapid eye movement" and "wakefulness" are mapped to the values ​​4, 3, 2 and 1 respectively, thus forming a numerical sequence of sleep stages; The cumulative normalized path cost is obtained by performing time-warping matching between the numerical sequence of sleep stages and the standard sleep cycle template sequence that has also undergone numerical mapping. The cumulative normalized path cost is divided by the length of the matching path and normalized. This value is used as the periodic rhythm deviation.

8. The sleep quality analysis method based on medical bed sensors according to claim 7, characterized in that, When generating the baseline assessment coefficient, the proportion of total sleep time in deep sleep, light sleep, and REM sleep stages throughout the entire sleep cycle is calculated and recorded as follows: , , ; The basic evaluation coefficient is obtained by weighting and summing the proportions using preset weights. Basic evaluation coefficient The calculation formula is: ; in, , , These are the weighting coefficients for each sleep stage, and their sum is 1.

9. The sleep quality analysis method based on medical bed sensors according to claim 8, characterized in that, The formula for generating a comprehensive sleep quality score is as follows: ; in, For comprehensive scoring, Based on the evaluation coefficient, This represents the degree of deviation from the periodic rhythm. It is an unsteady-state perturbation factor; , , These are the weighting coefficients for each item. and It is a non-linear adjustment index.

10. A sleep quality analysis system based on sensors in a medical bed, characterized in that, It includes a memory and a processor, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the sleep quality analysis method based on medical bed sensors as described in any one of claims 1-9 is implemented.

Images