Mental stress assessment method based on non-contact sensor and HRV data analysis

Abstract

The invention relates to a mental stress assessment method based on a non-contact sensor and HRV data analysis. The method comprises the following steps: acquiring a historical RR interval data set and a target day RR interval data set, and correspondingly generating a plurality of different interval characteristics based on RR interval data; calculating a quality score of the RR interval data; carrying out average calculation on the quality scores of all RR interval data contained in each sleep stage to obtain a stage average value corresponding to each sleep stage label; calculating a stability score value corresponding to each sleep stage based on the average value of each stage, and determining an optimal sleep stage; extracting all target day RR interval data of the optimal sleep stage, and generating a target day HRV value; generating a historical HRV sequence based on the historical RR interval data set; and performing numerical processing on the target day HRV value based on the constructed detrending model and the historical HRV sequence to generate a mental assessment HRV sequence, and generating a mental stress assessment result based on the mental assessment HRV sequence.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of mental stress assessment, and in particular to a mental stress assessment method based on a non-contact sensor and HRV data analysis. BACKGROUND

[0002] Mental stress assessment is generally achieved by quantifying the psychological load state of the physiological indicators of the human body. The core principle is that stress activates the sympathetic nervous system, suppresses parasympathetic nervous activity, and causes heart rate variability (HRV) to decrease. Therefore, mental stress assessment can be performed by model data analysis of heart rate variability. HRV is an important physiological indicator for assessing the function of the autonomic nervous system and the level of mental stress. HRV reflects the millisecond-level fluctuations of consecutive RR intervals, which directly reflects the balance between the sympathetic nervous system (excitation) and the parasympathetic nervous system (inhibition). The RR interval refers to the time difference between adjacent R-wave peaks in the QRS complex of an electrocardiogram.

[0003] In the prior art, there are various mental stress assessment schemes based on HRV. These schemes use contact sensors, such as finger clamps (photoelectric pulse wave PPG) or non-contact sensors (smart mattress piezoelectric film), to collect RR interval data. Then, HRV is calculated based on the RR interval data, and finally, the mental stress assessment result is obtained through time domain analysis and frequency domain analysis.

[0004] For contact methods, the device generally disrupts the user's body comfort, which affects the user's short-term emotional fluctuations and the final assessment result. The data collection duration of contact devices is limited by the user's tolerance, and most can only collect short-term data for a few minutes. The mental stress level represented by small data is more likely to deviate due to the user's short-term emotional fluctuations and physical energy consumption. The stability of the data cannot be optimal, which leads to a large deviation in the subsequent calculation of HRV, and the reference value of the mental stress assessment result is weakened.

[0005] Although the non-contact method can extend the data collection duration to some extent and obtain more data volume due to its comfort advantage, it is worth noting that the accuracy of data analysis by the non-contact method depends heavily on the quality of the RR interval data processing model. The existing non-contact method has defects such as inaccurate RR interval data acquisition and data processing model algorithm that does not exclude HRV periodic interference factors. Both of the above problems will lead to inaccurate HRV data, which will affect the scientificity and accuracy of mental stress assessment. SUMMARY

[0006] The purpose of the present application is to provide a mental stress evaluation method based on a non-contact sensor and HRV data analysis, which can calculate HRV based on high-quality RR interval data and dynamically correct periodic interference offset factors in the RR interval data processing process, and the mental stress evaluation results obtained based on the HRV data analysis method have higher scientific accuracy.

[0007] In order to achieve the above-mentioned purpose, the technical scheme adopted by the present application is as follows:

[0008] In the first aspect, the present application provides a mental stress evaluation method based on a non-contact sensor and HRV data analysis, which comprises the following steps:

[0009] Step 02, obtaining a historical RR interval data set and a target day RR interval data set, the RR interval data set containing a plurality of RR interval data; generating a plurality of different interval features based on the RR interval data;

[0010] Step 04, based on each interval feature of an RR interval data and the score weight assigned to each interval feature, calculating the quality score of the RR interval data;

[0011] Step 06, based on the pre-established mapping relationship between sleep staging labels and RR interval data, combining the historical RR interval data set and the target day RR interval data set, and averaging the quality scores of all RR interval data contained in each sleep staging, obtaining the staging mean value corresponding to each sleep staging label;

[0012] Step 08, based on each staging mean value, calculating the stability score value corresponding to each sleep staging, the stability score value being used to represent the RR interval data stability of the sleep staging, and determining the sleep staging with the maximum stability score value as the optimal sleep staging;

[0013] Step 10, extracting all target day RR interval data of the optimal sleep staging from the target day RR interval data set, and generating a target day HRV value based on the target day RR interval data; generating a historical HRV sequence based on the historical RR interval data set;

[0014] Step 12, based on the constructed detrended model and the historical HRV sequence, numerically processing the target day HRV value to generate a mental evaluation HRV sequence, and generating a mental stress evaluation result based on the mental evaluation HRV sequence.

[0015] As a preferred embodiment of the present application, the calculation method of the quality score Q is as follows:

[0016]

[0017] Wherein, Q is the quality score, F is the interval feature, and W is the score weight assigned to each interval feature.i w is the pre-allocated score weight, i is the category to which the interval feature belongs, and m is the total number of interval feature categories. i w is the pre-allocated score weight, i is the category to which the interval feature belongs, and m is the total number of interval feature categories.

[0018] As a preferred embodiment of the present application, when calculating the stability score value corresponding to each sleep stage, the steps of calculating the standard deviation of the duration and the standard deviation of the mean stage are further included, specifically:

[0019] All sleep stage durations corresponding to the sleep stage to be calculated on all detection days are obtained from the historical RR interval data set, and the standard deviation of the duration is calculated, which is used to represent the duration stability corresponding to the sleep stage. All mean stage values corresponding to the sleep stage to be calculated on all detection days are calculated, and the standard deviation of the mean stage is calculated, which is used to represent the RR interval quality stability corresponding to the sleep stage.

[0020] As a preferred embodiment of the present application, the determination method of the optimal sleep stage is specifically: the standard deviation of the duration and the standard deviation of the mean stage of a sleep stage to be calculated are standardized, and each is given a score weight, and weighted calculation is performed to obtain a stability score value; the sleep stage to be calculated with the maximum stability score value is selected as the optimal sleep stage.

[0021] As a preferred embodiment of the present application, the detrended model is composed of a short-period prediction part and a long-period prediction part.

[0022] As a preferred embodiment of the present application, the short-period prediction part and the long-period prediction part can be constructed based on any one of a sine wave function, a sawtooth wave function, a triangular wave function, or a Fourier series.

[0023] As a preferred embodiment of the present application, the function type of the prediction part is determined based on the curve shape of the historical HRV sequence.

[0024] As a preferred embodiment of the present application, the specific steps of numerically processing the target day HRV value based on the constructed detrended model and the historical HRV sequence are:

[0025] Generating a historical HRV sequence based on a historical RR interval data set;

[0026] Generating a detrended numerical sequence based on the detrended model;

[0027] Subtracting the detrended numerical sequence from the historical HRV sequence to obtain a target day detrended HRV sequence;

[0028] Adding a target day HRV value to each value in the target day detrended HRV sequence to generate a final mental assessment HRV sequence.

[0029] In a second aspect, the present application further provides a storage medium, which stores a computer instruction program, and the computer instruction program, when executed by a processor, causes the processor to perform the steps of the mental stress evaluation method based on the non-contact sensor and HRV data analysis.

[0030] In a third aspect, the present application further provides a computer device, which comprises at least one memory and at least one processor, and the memory stores a computer instruction program, and the computer instruction program, when executed by the processor, causes the processor to perform the steps of the mental stress evaluation method based on the non-contact sensor and HRV data analysis.

[0031] In summary, the present application has the following beneficial effects:

[0032] The technical solution provided by the present application can accurately locate the "golden sleep staging", and compared with the traditional contact type device, the present application selects "historical data + real-time data" for long-period analysis during data analysis; the quality score of the multi-dimensional stability of the RR interval data is obtained, and the quality score result is used to select the "optimal adaptive sleep staging" for the user, so that the source data quality is preliminarily improved; and the accuracy of the final mental stress evaluation result is greatly improved.

[0033] In addition, the present application creatively constructs a personalized detrended model to perform dynamic detrended processing operation, and excludes the adverse effects of the HRV periodic interference factors that have not been considered in the traditional RR interval data processing model on the final mental stress evaluation result, so that the accuracy of the evaluation result is greatly improved. BRIEF DESCRIPTION OF DRAWINGS

[0034] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings in the following description are some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained without creative labor on the basis of these drawings.

[0035] Figure 1 The method flowchart of the present application is shown in the figure.

[0036] Figure 2 The method flowchart of the present application is shown in the figure. DETAILED DESCRIPTION

[0037] The subject matter described herein will now be discussed with reference to example implementations. It should be understood that these implementations are discussed solely for the purpose of illustrating aspects of the subject matter described herein and are not a limitation of the scope, applicability, or examples set forth in the claims. Changes in the function and arrangement of elements discussed can be made without departing from the scope of the subject matter described herein. Various examples can omit, substitute, or add various procedures or components as appropriate. For instance, the methods described can be performed in an order different than that described, and / or various steps can be added, omitted, or combined. Also, features described with respect to some examples can be combined in other examples.

[0038] In the present application, the non-contact sensor (such as the piezoelectric film embedded in the mattress) generates electrocardiogram information by capturing the body surface micro-vibration caused by heartbeat (Ballistocardiography, BCG) and converting it into heart rate voltage signals, then identifying the electrocardiogram information through the signal processor, obtaining the J wave peak value in the waveform (corresponding to the mechanical characteristics of the blood ejection impact on the blood vessels during the systolic period), and finally calculating the time interval of adjacent J waves as the equivalent RR interval. In medicine, the electrocardiogram signal can also be collected by a contact electrode, and the R wave vertex in the QRS complex is repositioned to calculate the time difference of adjacent R waves as the RR interval, i.e. the electrocardiogram signal analysis method. It should be clear that the equivalent RR interval obtained by the non-contact sensor provided in the present application is only different from the traditional electrocardiogram acquisition method in terms of implementation path, and both can obtain accurate and usable RR interval data.

[0039] As can be seen from the above, the RR interval data output by the vibration sensor is consistent with the physical meaning of the RR interval of the traditional electrocardiogram, and in a broad sense, both represent the time difference between adjacent heartbeats. HRV refers to the small difference (unit: millisecond) in RR interval fluctuation over time, reflecting the autonomous regulation ability of cardiac rhythm, and detecting HRV requires being performed in a resting state. Therefore, in the present embodiment, the non-contact sensor used is a vibration sensor in a sleep bed mattress, which collects the heartbeat vibration characteristics of the user's body during the sleep stage and analyzes based on the existing signal processing device, and can output all RR interval data of the user from falling asleep to waking up. In the actual data collection process, the vibration sensor reports data once every certain interval (for example, 10S), and the data is analyzed and preprocessed by the processor in the back end.

[0040] The historical heartbeat interval data set refers to a data set containing a certain number of days, for example, 30 days, all RR interval data corresponding to the sleep stage of each day are recorded completely; the target day corresponds to the latest day, in the present application, the HRV data of the target day is used for the mental stress assessment to reflect the user's recent mental stress state, wherein the mental stress assessment using the HRV data can be realized by the existing means, and the present application focuses on how to obtain the HRV data with the highest scientificity and reference degree. As can be seen, the target day RR interval data set contains all RR interval data corresponding to the sleep stage of the target day; in addition, each RR interval data contains multiple feature dimensions, which together reveal the complexity of the cardiac rhythm and the autonomic nervous regulation state.

[0041] As shown in Figure 1 and Figure 2 In the present method, first, the historical RR interval data set and the target day RR interval data set are obtained; then, based on each feature dimension of a RR interval data and the scoring weight corresponding to each feature dimension, the quality score Q of the RR interval data is calculated.

[0042] Specifically, the calculation method of the quality score is:

[0043]

[0044] Wherein, Q is the quality score, F i is the interval feature, w i is the pre-allocated scoring weight, i is the type to which the interval feature belongs, and m is the total number of interval feature types. It can be seen that for each RR interval data, a quality score is obtained, which is used to select the subsequent optimal sleep staging.

[0045] In the present method, all RR interval data contained in the optimal sleep staging are regarded as high-quality RR interval data. When calculating the subsequent HRV, only the high-quality RR interval data is used. Each interval feature F i corresponds to a feature dimension. Specifically, based on the existing means, for example, waveform symmetry feature F1, spectral energy ratio feature F2 and template matching correlation feature F3 can be generated. Among them, the waveform symmetry feature F1 can measure the symmetry and regularity of the RR interval change, which can be quantitatively calculated based on the Poincaré Plot to obtain a specific numerical value; the spectral energy ratio feature F2 can evaluate the balance state of the autonomic nervous system, which can be obtained based on the Fourier transform calculation; the template matching correlation feature F3 can evaluate the shape quality and reliability of the current sequence; the closer the numerical value of the above three interval features to the midpoint of the specified range, the higher the scoring weight w iThe larger the value, the greater the influence on the calculation of the quality score Q. Based on this, the quality and reliability of the RR interval data can be comprehensively evaluated from multiple dimensions, reducing the adverse effects of unqualified data caused by atrial fibrillation, premature beat, strenuous exercise or device noise on the overall quality superiority of the data, which is reflected at the data level as follows: the larger the quality score Q, the higher the quality superiority of the RR interval data.

[0046] As can be seen from the above formula, for an RR interval data, its quality score Q is calculated by weighting each interval feature F i and the scoring weight w i corresponding to each interval feature, and here the contribution of different interval features when performing mental stress evaluation is considered. The initial value of each weight can be set based on artificial experience, and can be adaptively adjusted later, for example, adjusted according to numerical fluctuations or adjusted according to user feedback. For example, when adjusting according to numerical fluctuations, historical data can be considered to analyze the interval features F i of the user's data in the past 30 days, and features with large numerical fluctuations are selected to reduce the weight w i . In this process, attention needs to be paid to weight constraints, such as not less than 0.2, to avoid a single feature dominating the score.

[0047] The variability coefficient is used to quantify the fluctuation of the feature, and features with CV>0.2 are considered to have high fluctuation and need to reduce the weight;

[0048] If , the weight w is reduced, and the amplitude is:

[0049]

[0050] wherein, is the weight reduction amount, . If , set to satisfy the constraint.

[0051] Afterwards, it is known that a person will experience multiple sleep stages (sleep epochs) during the whole sleep process from falling asleep to waking up, such as light sleep, deep sleep or REM (rapid eye movement sleep), which can be defined as a sleep epoch in the technical solution by itself, and any two (or more) adjacent sleep epochs (such as deep sleep + REM) can also be defined as a sleep epoch in the technical solution, so based on actual needs, multiple different sleep epochs can be divided from the whole sleep stage of a user, and it should be noted that any two (or more) adjacent sleep epochs are selected as candidate sleep epochs because physiological signals will present a unique and mixed change pattern during sleep stage transition (such as from deep sleep to REM). This pattern is neither purely deep sleep nor purely REM. For example, creating a mixed sleep epoch of “deep sleep + REM” can analyze this transition process as a whole, so as to more accurately describe and understand the data characteristics in this process.

[0052] Taking the sleep epoch of deep sleep as an example, there is almost a deep sleep every day in the historical data of 30 days, and assuming that there is at least one deep sleep every day in the previous 30 days, at least 30 groups of data can be obtained for the sleep epoch of deep sleep (the sleep epoch is one time period in the whole sleep stage, and the time length of the time period can be set by an artificial system, which corresponds to one physiological sleep stage or a combination of multiple adjacent physiological sleep stages. The physiological sleep stage can be obtained based on the prior art, and the information obtained includes the start time and duration of the physiological sleep stage; more specifically, a sleep epoch contains multiple RR interval data, each RR interval data is a time value, and since the total time length of different sleep epochs can be different, the number of RR interval data contained is also different).

[0053] Then it needs to be determined which sleep epoch contains the optimal quality RR interval data for a specific user. In other words, it is necessary to find a sleep epoch with the best stability among multiple pre-defined sleep epochs.

[0054] In the following embodiments, the sleep stage with the highest data superiority (i.e., the optimal sleep stage) is found by obtaining a stability score value of sleep staging, and the RR interval data in the optimal sleep stage is selected as the basis for subsequent calculation of HRV. In actual operation, in order to distinguish different sleep stages, sleep staging labels can be established in advance, and each sleep stage includes multiple RR interval data. The steps are as follows: calculate the average value of the quality score Q of the RR interval data corresponding to each sleep stage in a detection day, and correspondingly obtain multiple staging average values Q'; calculate all staging average values Q' corresponding to all detection days in the historical RR interval data set. Based on each staging average value Q', a stability score value corresponding to each sleep stage is calculated, and the sleep stage with the maximum stability score value is assigned an optimal sleep stage label.

[0055] The specific operation steps are as follows: first, in the first aspect, the time length of all sleep stages corresponding to all detection days of the sleep stage to be calculated is obtained from the historical RR interval data set, and the time length standard deviation is calculated, which is used to represent the time length stability corresponding to the sleep stage. More specifically, for the time length stability indicator, the time length of all sleep stages corresponding to all detection days of the sleep stage to be calculated is obtained from the historical RR interval data set, and the time length standard deviation σ_duration is calculated, based on which the time length stability corresponding to the sleep stage to be calculated is determined.

[0056] In the second aspect, all staging average values Q' corresponding to all detection days of the sleep stage to be calculated are calculated, and the staging average value standard deviation is calculated, which is used to represent the RR interval quality stability corresponding to the sleep stage. More specifically, for the RR interval quality stability indicator, all staging average values Q' corresponding to all detection days of the sleep stage to be calculated are calculated, and the staging average value standard deviation σ_quality is calculated, based on which the RR interval quality stability corresponding to the sleep stage to be calculated is determined.

[0057] Then, the time length standard deviation σ_duration and the staging average value standard deviation σ_quality of a sleep stage to be calculated are standardized, and each is assigned a score weight, and weighted calculation is performed. The weighted calculation result (i.e., the stability score value) reflects the comprehensive stability level of the sleep stage, which is in the form of a score, and the higher the score, the better the staging stability. For example, for the defined sleep stage S1 (selecting deep sleep period) and sleep stage S2 (selecting deep sleep period + REM mixed period), the sleep stage time length standard deviation σ_duration of S1 is 15 min, and the staging average value standard deviation σ_quality of S1 is 0.05. The sleep stage time length standard deviation σ_duration of S2 is 20 min, and the staging average value standard deviation σ_quality of S2 is 0.1. The sleep stage time length standard deviation σ_duration and the staging average value standard deviation σ_quality of S1 are standardized, and each is assigned a score weight of 0.5, and the weighted calculation result (i.e., the stability score value) of S1 is 0.5. The sleep stage time length standard deviation σ_duration and the staging average value standard deviation σ_quality of S2 are standardized, and each is assigned a score weight of 0.5, and the weighted calculation result (i.e., the stability score value) of S2 is 0.5. S1 _duration is 15 min, and the staging average value standard deviation σ S1_quality is 0.05, dimensionless by Z-score normalization in the prior art, and magnitude normalized by reverse conversion, so that the standard deviation σ S1 _duration and σ S1 _quality is converted to 0.8 and 0.9, and the stability score value 0.86 is obtained by weighted calculation based on the assigned score weights (for example, 40% and 60%).

[0058] Based on this method, the stability score value of sleep stage S2 is calculated: the sleep stage duration standard deviation σ S2 _duration is 12 min, and the sleep stage average standard deviation σ S2 _quality is 0.06, dimensionless by Z-score normalization in the prior art, and magnitude normalized by reverse conversion, so that the standard deviation σ S2 _duration and σ S2 _quality is converted to 0.75 and 0.92, and the stability score value 0.852 is obtained by weighted calculation based on the assigned score weights (for example, 40% and 60%).

[0059] The stability score value of the two is compared, and the larger one is the one with high priority (0.86>0.852, S1 has high priority), and if the stability score values of the two are the same, they are considered to have the same priority. In actual application, usually dozens of sleep stages are defined, so after the calculation is completed, the priority ranking of all sleep stages can be obtained.

[0060] It should be noted that the sleep stage with the highest stability priority of different users is not necessarily the same. For other users, the mixed stage formed by "deep sleep period + REM" may be the sleep stage with the highest stability priority.

[0061] After the optimal sleep stage of a user is determined by the above steps, the historical HRV sequence generation step, the detrending step and the mental stress assessment step are executed.

[0062] In this embodiment, the detrending step can remove the interference of periodic factors such as female physiological cycle (short cycle prediction part implementation) and north-south hemisphere seasonal cycle (long cycle prediction part implementation).

[0063] Specifically, the historical HRV sequence generation step is:

[0064] All target day RR interval data corresponding to the optimal sleep stage label are extracted from the target day RR interval data set, and target day HRV values are generated based on the target day RR interval data; a historical HRV sequence is generated based on a historical RR interval data set.

[0065] It should be noted that, by the time domain analysis, frequency domain analysis, nonlinear analysis method of the prior art, the skilled in the art can obtain HRV data based on RR interval data. The most commonly used data dimension of HRV data is also the default data dimension of the skilled in the art, which is SDNN (standard deviation of RR interval). In this embodiment, a standard deviation value (unit: ms) calculated from all RR interval data values contained in the sleep stage with the highest stability priority corresponding to a detection day is the target day HRV value described above. As can be seen, based on multiple historical HRV values, a historical HRV sequence can be formed.

[0066] Then, a detrending step and a mental stress assessment step are performed:

[0067] Based on the constructed detrending model and the historical HRV sequence, the target day HRV value is numerically processed to generate a mental assessment HRV sequence, and a mental stress assessment result is generated based on the mental assessment HRV sequence.

[0068] Among them, the detrending model is composed of a short cycle prediction part and a long cycle prediction part; the short cycle prediction part and the long cycle prediction part can be constructed based on any one of a sine wave function, a sawtooth wave function, a triangular wave function or a Fourier series.

[0069] In addition, the prediction part is constructed based on any one of a sine wave function, a sawtooth wave function, a triangular wave function or a Fourier series.

[0070] Specifically, the embodiment provides the following several detrending models:

[0071] Taking the sine wave function as an example:

[0072]

[0073] Among them, the short cycle prediction part is: ;

[0074] (Amplitude): represents the intensity of short-term cycle fluctuation. For example, the 28-day menstrual cycle fluctuation amplitude of female user RMSSD is usually 2-5 ms, reflecting the influence of physiological cycle on HRV;

[0075] (Peiod): represents the length of the short-term cycle, which can be set to about 28 days (female menstrual cycle, individual differences);

[0076] (Phase): represents the starting time offset of the short-term cycle, reflecting the personalized starting point of individual physiological cycle (such as the starting time of the luteal phase of female menstrual cycle);

[0077] (baseline): represents the average value of the short-term cycle, affected by age and gender. For example, the RMSSD baseline value of older users is lower;

[0078] The long-term cycle prediction unit is: ;

[0079] (amplitude): represents the intensity of the long-term cycle fluctuation. For example, the LF / HF ratio of HRV may increase by 5-10% in the winter in the northern hemisphere, and the fluctuation is smaller (<3%) in the southern hemisphere tropics;

[0080] (period): represents the length of the long-term cycle, usually set to 365 days, capturing annual seasonal fluctuations;

[0081] (phase): represents the starting time offset of the seasonal cycle, reflecting regional differences;

[0082] (baseline): represents the average value of the long-term cycle, jointly determines the overall baseline level of the signal with the short-term baseline .

[0083] Take the sawtooth wave function as an example:

[0084]

[0085] where, is the sawtooth wave function, controls the proportion of the rising and falling stages, and the remaining parameters are the same as the above sine wave function.

[0086] Take the triangular wave function as an example:

[0087]

[0088] where, is the triangular wave function, controls the asymmetry of the rising and falling stages, and the remaining parameters are the same as the above sine wave function.

[0089] Take the complex waveform (Fourier series) as an example:

[0090]

[0091] N is the harmonic order, selected according to the complexity of the data, and the remaining parameters are the same as the above sine wave function.

[0092] The above shows multiple available detrended model wave functions, and it can be seen that y(t) is a time-based numerical sequence that reflects a trend, which needs to be removed in subsequent calculations.

[0093] As can be seen from the above, in the same way as the target day HRV data is calculated, each day in the historical data corresponds to a "historical target day" HRV data (selected from the RR interval data in the optimal sleep staging determined above), and if the corresponding 90-day time scale of the historical data, a historical HRV sequence can be fitted based on 90 HRV data (that is, 90 SDNN), which contains 90 SDNN values.

[0094] For different users, the function is different, and the configuration of the parameters is also different. The above is only an example, which can reflect personalized adaptation. Specifically, before the target day HRV data is de-trended based on the de-trended model, the user's history data is selected to select the appropriate wave function type to build the de-trended model.

[0095] The wave function type selection step can be realized by existing simulation software algorithm. The specific waveform of the function corresponding to each user may be different. Only the optimal waveform is selected to build the de-trended model of the user, the accuracy can be obtained in the subsequent de-trending operation. Specifically, the historical HRV sequence is curved, and the time-based numerical sequence y(t) obtained above can also be curved. The coordinate systems of the two are unified; different reference functions can correspond to different y(t), and similarity comparison can be performed through existing computer software programs, so as to select the y(t) function configuration closest to the historical HRV sequence curve as the de-trended numerical sequence y(t) suitable for the user.

[0096] Then, based on the de-trended model and the historical HRV sequence, the target day HRV value is numerically processed:

[0097] Generate a historical HRV sequence based on a historical RR interval data set;

[0098] Generate a de-trended numerical sequence based on the de-trended model;

[0099] Subtract the de-trended numerical sequence from the historical HRV sequence to obtain a target day de-trended HRV sequence;

[0100] Add a target day HRV value to each value in the target day de-trended HRV sequence to generate a final mental assessment HRV sequence.

[0101] The following provides a possible specific calculation method (taking a sine wave function as an example):

[0102]

[0103]

[0104]

[0105] wherein y(t) is the detrended value sequence applicable to the user, y raw (t) is the historical HRV sequence, y detrended (t) is the target day detrended HRV sequence, y user is the target day HRV value, y adjusted (t) is the mental assessment HRV sequence.

[0106] It can be seen that the historical HRV sequence y raw (t) minus the detrended value sequence y(t) can obtain the target day detrended HRV sequence y detrended (t), which eliminates the adverse effects of the periodic interference factors of HRV in the traditional model on the final mental stress assessment results. Then, for the target day detrended HRV sequence y detrended (t), each numerical element in the sequence is added to a target day HRV value y user , the purpose of this calculation step is to retain the physiological significance of the original HRV data, and finally obtain the mental assessment HRV sequence y adjusted (t).

[0107] Finally, based on the mental assessment HRV sequence y adjusted (t), the existing mental stress assessment model is used to generate the mental stress assessment results.

[0108] It should be noted that in order to ensure that the detrended model adapts to the physiological changes of the user, the dynamic optimization strategy is also introduced in this embodiment. As the user uses the smart bed every day, the physiological data of the user is continuously accumulated, and the system automatically updates the historical HRV sequence y raw (t) every day to ensure that the detrended model iterates with time and user state changes.

[0109] The above has described a plurality of embodiments of the present disclosure, and the above description is exemplary, not exhaustive, and is not limited to the disclosed embodiments. Many modifications and changes are obvious to those skilled in the art without departing from the scope and spirit of the described embodiments. The choice of terms used herein is intended to best explain the principles, practical applications, or technical improvements in the art, or to enable other ordinary skilled persons in the art to understand the embodiments disclosed herein.

Claims

1. A mental stress assessment method based on non-contact sensor and HRV data analysis, characterized in that, The method comprises: obtaining a historical RR interval data set and a target day RR interval data set, the RR interval data set comprising a plurality of RR interval data; generating a plurality of different interval features based on the RR interval data; calculating a quality score of one RR interval data based on each interval feature of the RR interval data and a score weight assigned to each interval feature; based on a pre-established mapping relationship between sleep staging labels and RR interval data, combining the historical RR interval data set and the target day RR interval data set, and averaging the quality scores of all RR interval data included in each sleep stage to obtain a staging mean value corresponding to each sleep staging label; calculating a stability score value corresponding to each sleep stage based on each staging mean value, wherein the stability score value represents the stability of the RR interval data of the sleep stage, and the sleep stage with the maximum stability score value is determined as the optimal sleep stage; extracting all target day RR interval data of the optimal sleep stage from the target day RR interval data set, and generating a target day HRV value based on the target day RR interval data; 2. The mental stress evaluation method based on non-contact sensor and HRV data analysis according to claim 1, characterized in that, generating a historical HRV sequence based on the historical RR interval data set; ， where Q is the quality score, F i is the interval feature, w i is the pre-allocated score weight, i is the class to which the interval feature belongs, and m is the total number of interval feature classes.

3. The stress evaluation method based on non-contact sensor and HRV data analysis according to claim 2, characterized in that, based on the constructed detrended model and the historical HRV sequence, performing numerical processing on the target day HRV value to generate a mental assessment HRV sequence, and generating a mental stress assessment result based on the mental assessment HRV sequence. The calculation method of the quality score Q is: When calculating the stability score value corresponding to each sleep stage, the steps of calculating the time length standard deviation and the staging mean value standard deviation are further included, specifically:

4. The mental stress assessment method based on non-contact sensor and HRV data analysis according to claim 3, characterized in that, obtaining all sleep staging time lengths of all detection days corresponding to the sleep stage to be calculated from the historical RR interval data set, and calculating the time length standard deviation, which represents the time length stability corresponding to the sleep stage; calculating all staging mean values corresponding to the sleep stage to be calculated on all detection days, and calculating the staging mean value standard deviation, which represents the RR interval quality stability corresponding to the sleep stage. 5.The mental stress evaluation method based on non-contact sensor and HRV data analysis of claim 1, wherein, The determination method of the optimal sleep stage is specifically:

6. The mental stress assessment method based on non-contact sensor and HRV data analysis according to claim 5, characterized in that, standardizing the time length standard deviation and the staging mean value standard deviation of one sleep stage to be calculated, respectively assigning a score weight to each, performing weighted calculation to obtain a stability score value, and selecting the sleep stage to be calculated with the maximum stability score value as the optimal sleep stage.

7. The stress evaluation method based on non-contact sensor and HRV data analysis according to claim 6, characterized in that, The detrended model is composed of a short cycle prediction part and a long cycle prediction part.

8. The stress evaluation method based on non-contact sensor and HRV data analysis according to claim 7, characterized in that, The short cycle prediction part and the long cycle prediction part can be constructed based on any one of a sine wave function, a sawtooth wave function, a triangular wave function, or a Fourier series. The function type of the prediction part is determined based on the curve shape of the historical HRV sequence. The specific steps of numerically processing the target day HRV value based on the constructed detrended model and the historical HRV sequence are: generating a historical HRV sequence based on the historical RR interval data set; generating a detrended numerical sequence based on the detrended model; subtracting the detrended numerical sequence from the historical HRV sequence to obtain a target day detrended HRV sequence; For each value in the de-trended HRV sequence of the target day, add a target day HRV value to generate a final mental assessment HRV sequence.

9. A storage medium, characterized by A computer program product is stored in the memory and contains computer instructions which, when executed by the processor, cause the processor to perform the steps of the method according to any one of claims 1-8.

10. A computer device, comprising: A computer program product is stored in the memory and contains computer instructions which, when executed by the processor, cause the processor to perform the steps of the method according to any one of claims 1-8.

Images