User fatigue state evaluation method, device, equipment and medium

Abstract

The application discloses a user fatigue state evaluation method, device, equipment and medium, the user fatigue state evaluation method collects user physiological signals through a physiological signal sensor and pre-processes, extracts multi-dimensional physiological characteristic indexes in parallel based on the pre-processed physiological signals, dynamically adjusts the weight of each dimension according to the signal quality evaluation result and / or physiological state trigger condition corresponding to each dimension physiological characteristic index, compares the real-time extracted physiological characteristic indexes of each dimension with the user individual physiological baseline to obtain a normalized score, combines the adjusted weight to calculate a fatigue comprehensive score, performs time sequence smoothing processing and abnormal fluctuation suppression processing on the fatigue comprehensive score to obtain a corrected fatigue comprehensive score, determines the user fatigue level according to the mapping relationship between the corrected fatigue comprehensive score and the preset fatigue level threshold, and outputs a fatigue evaluation result. The scheme significantly improves the accuracy and stability of fatigue state evaluation.

Description

Technical Field

[0001] This invention relates to the field of smart home technology, and in particular to a method, apparatus, device, and medium for assessing user fatigue status. Background Technology

[0002] With the fast pace of modern life and increasing work pressure, fatigue has become a significant health issue affecting modern people. Long-term fatigue accumulation can lead to health risks such as cardiovascular disease and weakened immune function. Smart home devices, as frequently used devices in daily life, provide an ideal platform for achieving seamless fatigue monitoring. Existing fatigue assessment methods for smart homes primarily determine the degree of fatigue by collecting and analyzing users' physiological signals.

[0003] However, existing technologies mostly employ single-indicator assessment methods, relying solely on one physiological indicator from heart rate, respiration, or body movement to determine fatigue. Because a single physiological indicator is susceptible to interference from various factors, its assessment has a high rate of false positives and false negatives, failing to meet the needs of medical-grade or near-medical-grade applications. When the monitoring environment is complex, with factors such as clothing obstruction, multiple people interfering, or environmental vibrations, the accuracy of single-indicator methods decreases, and the reliability of the assessment is significantly reduced. Furthermore, existing technologies use uniform physiological indicator thresholds to determine fatigue for all users. Since there are significant individual differences in the physiological foundations of different users, the same physiological indicator value may represent different fatigue states for different users. Using absolute thresholds cannot accommodate the physiological differences among users, leading to significant biases in the assessment results.

[0004] In summary, existing fatigue assessment methods for smart homes have low accuracy and insufficient stability and reliability, making it difficult to meet users' needs for precise fatigue monitoring. Summary of the Invention

[0005] The embodiments of the present invention provide a method, apparatus, device and medium for assessing user fatigue status, aiming to solve the technical problem of low assessment accuracy of existing fatigue status assessment methods for smart homes.

[0006] In a first aspect, embodiments of the present invention provide a method for assessing user fatigue state. The method includes: collecting a user's physiological signals through a physiological signal sensor and preprocessing the physiological signals to obtain preprocessed physiological signals; extracting multi-dimensional physiological feature indicators in parallel based on the preprocessed physiological signals; dynamically adjusting the weights of each dimension of physiological feature indicators according to the signal quality assessment results and / or physiological state triggering conditions corresponding to each dimension of physiological feature indicators; comparing the real-time extracted physiological feature indicators of each dimension with the user's personalized physiological baseline to obtain a normalized score for each dimension, and calculating a comprehensive fatigue score by combining the adjusted weights of the physiological feature indicators; performing time-series smoothing and abnormal fluctuation suppression processing on the comprehensive fatigue score to obtain a corrected comprehensive fatigue score; and determining the user's fatigue level and outputting a fatigue assessment result based on the mapping relationship between the corrected comprehensive fatigue score and a preset fatigue level threshold.

[0007] Secondly, embodiments of the present invention also provide a user fatigue state assessment device for performing the user fatigue state assessment method described above.

[0008] Thirdly, embodiments of the present invention also provide a computer device, the computer device including a memory and a processor connected to the memory; the memory is used to store a computer program; the processor is used to run the computer program stored in the memory to perform the steps of the above-described user fatigue state assessment method.

[0009] Fourthly, embodiments of the present invention also provide a computer-readable storage medium storing a computer program, the computer program including program instructions, which, when executed by a processor, can implement the steps of the above-described user fatigue state assessment method.

[0010] Compared with the prior art, the beneficial effects of the present invention are: In the technical solution of this invention, the user fatigue state assessment method collects and preprocesses user physiological signals using physiological signal sensors. Based on the preprocessed physiological signals, multi-dimensional physiological feature indicators are extracted in parallel. The weights of each dimension are dynamically adjusted according to the signal quality assessment results and / or physiological state triggering conditions corresponding to each dimension of physiological feature indicators. The real-time extracted physiological feature indicators of each dimension are compared with the user's personalized physiological baseline to obtain a normalized score. A comprehensive fatigue score is calculated by combining the adjusted weights. The comprehensive fatigue score is then subjected to time-series smoothing and abnormal fluctuation suppression processing to obtain a corrected comprehensive fatigue score. The user's fatigue level is determined based on the mapping relationship between the corrected comprehensive fatigue score and a preset fatigue level threshold, and the fatigue assessment result is output. This solution significantly improves the accuracy and stability of fatigue state assessment. Attached Figure Description

[0011] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0012] Figure 1 A flowchart of the user fatigue state assessment method provided by the present invention; Figure 2 The first sub-flowchart of the user fatigue state assessment method provided by the present invention; Figure 3 This is a second sub-flowchart of the user fatigue state assessment method provided by the present invention; Figure 4 The third sub-flowchart of the user fatigue state assessment method provided by the present invention; Figure 5 The fourth sub-flowchart of the user fatigue state assessment method provided by the present invention; Figure 6 The fifth sub-flowchart of the user fatigue state assessment method provided by the present invention; Figure 7 The sixth sub-flowchart of the user fatigue state assessment method provided by the present invention; Figure 8 The seventh sub-flowchart of the user fatigue state assessment method provided by the present invention; Figure 9 The eighth sub-flowchart of the user fatigue state assessment method provided by the present invention; Figure 10 A schematic block diagram of a unit of the user fatigue state assessment device provided by the present invention; Figure 11 A schematic block diagram of a computer device provided for an embodiment of the present invention. Detailed Implementation

[0013] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0014] It should be understood that, when used in this specification and the appended claims, the terms "comprising" and "including" indicate the presence of the described features, integrals, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or collections thereof.

[0015] It should also be understood that the terminology used in this specification is for the purpose of describing embodiments only and is not intended to limit the invention. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0016] It should also be further understood that the term "and / or" as used in this specification and the appended claims refers to any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0017] In order to solve the technical problem of low accuracy in existing fatigue state assessment methods for smart homes, this invention discloses a user fatigue state assessment method.

[0018] Reference Figures 1 to 9 The user fatigue assessment method includes the following steps: S110. Collect the user's physiological signals through a physiological signal sensor, and preprocess the physiological signals to obtain preprocessed physiological signals; S120. Based on the preprocessed physiological signals, extract multi-dimensional physiological feature indicators in parallel. S130. Based on the signal quality assessment results and / or physiological state triggering conditions corresponding to each dimension of physiological characteristic indicators, dynamically adjust the weights of each dimension of physiological characteristic indicators. S140. Compare the real-time extracted physiological characteristic indicators of each dimension with the user's personalized physiological baseline to obtain the normalized score of each dimension, and calculate the comprehensive fatigue score by combining the adjusted weights of the physiological characteristic indicators. S150. Perform time-series smoothing and abnormal fluctuation suppression processing on the fatigue comprehensive score to obtain the corrected fatigue comprehensive score. S160. Based on the mapping relationship between the corrected fatigue comprehensive score and the preset fatigue level threshold, determine the user's fatigue level and output the fatigue assessment result.

[0019] The user fatigue assessment method of the present invention is applied within the system of smart home devices. Specifically, the physiological signal sensor employs a frequency modulated continuous wave (FMCW) millimeter-wave radar. This radar module is embedded in the smart home device, such as the backrest of a smart sofa or the upper half of a smart mattress, and is precisely pointed at the user's chest cavity to synchronously capture raw physiological signals related to heartbeat, breathing, and micro-body movements.

[0020] Preprocessing the physiological signal includes filtering the original analog signal through a bandpass filter to remove DC drift below a certain preset range and high-frequency noise above a certain preset range, followed by analog-to-digital conversion to obtain a digitized preprocessed physiological signal.

[0021] Based on the preprocessed physiological signals, multi-dimensional physiological characteristic indicators are extracted in parallel, specifically including: identifying R-wave peaks from heart rate signals and calculating RR interval sequences to obtain heart rate variability characteristic indicators RMSSD and LF / HF ratio; detecting respiratory cycles from respiratory signals and calculating respiratory interval standard deviation and respiratory variability coefficient as respiratory regularity characteristic indicators; detecting body movement events from body movement signals, statistically analyzing body movement frequency and amplitude per unit time, and generating body movement pattern scores by combining behavioral patterns such as restlessness and posture changes.

[0022] Based on the signal quality assessment results and physiological state triggering conditions corresponding to each dimension of physiological characteristic indicators, the weights of each dimension of physiological characteristic indicators are dynamically adjusted. During implementation, the signal quality assessment value of each dimension is first calculated. If the signal quality of a certain dimension is lower than the set threshold, its basic weight is reduced. At the same time, when physiological state triggering conditions such as the LF / HF ratio exceeding the preset value, the standard deviation of the respiratory interval exceeding the preset interval, or the body movement frequency exceeding the preset frequency per minute are detected, the weight of the corresponding dimension is increased accordingly, and upper and lower limit constraints are applied to the final weights to ensure that the weight allocation is reasonable.

[0023] The real-time extracted physiological characteristic indicators of each dimension are compared with the user's personalized physiological baseline. The user's personalized physiological baseline is a personal health record established by the system after continuously collecting data during the initial use phase of the user. By calculating the degree of deviation of the current indicator from the baseline value, a normalized score for each dimension is generated. Then, the score is weighted and summed with the adjusted weights to obtain a comprehensive fatigue score in the range of 0 to 100.

[0024] The fatigue comprehensive score is subjected to time-series smoothing and abnormal fluctuation suppression processing. Specifically, the continuous scores are smoothed using an exponentially weighted moving average algorithm. In this embodiment, the smoothing coefficient can be set to 0.3. At the same time, the change between two adjacent scores is monitored. If the difference exceeds 20 points, the outlier suppression mechanism is activated to correct the current score to the maximum allowable change value, thereby obtaining the corrected fatigue comprehensive score.

[0025] Based on the mapping relationship between the revised fatigue comprehensive score and the preset fatigue level threshold, a score of 0-25 is judged as Level 0 (full of energy), 26-50 as Level 1 (mild fatigue), 51-75 as Level 2 (moderate fatigue), and 76-100 as Level 3 (severe fatigue). The fatigue level is then transmitted to the user's smartphone application interface via Bluetooth for display, thus completing the output of the fatigue assessment results.

[0026] In one embodiment, step S110 includes: S111. Collect raw physiological signals by pointing a millimeter-wave radar sensor at the user's chest cavity location. The raw physiological signals include heart rate signals, respiratory signals, and body movement signals. S112. Perform bandpass filtering on the original physiological signal to remove DC components and high-frequency noise; S113. Perform analog-to-digital conversion on the filtered physiological signal to obtain the preprocessed digital physiological signal.

[0027] Specifically, raw physiological signals are collected by pointing a millimeter-wave radar sensor at the user's chest cavity. The millimeter-wave radar sensor uses frequency-modulated continuous wave radar, operating in the 24GHz to 77GHz frequency band. The millimeter-wave radar sensor emits electromagnetic wave signals, which are reflected when they encounter the user's body. The reflected signals carry information about the micro-movements of the user's chest cavity, including heartbeat micro-movements caused by heartbeats, respiratory fluctuations caused by lung respiration, and body displacement caused by changes in body posture.

[0028] Millimeter-wave radar sensors receive reflected signals and extract raw physiological signals containing physiological information by mixing, amplifying, and processing the reflected signals using intermediate frequency (IF) signals. These raw physiological signals include three types: heart rate, respiration, and body movement. The heart rate signal has a frequency range of approximately 0.8 Hz to 2 Hz, corresponding to a heart rate of 48 to 120 beats per minute; the respiration signal has a frequency range of approximately 0.1 Hz to 0.5 Hz, corresponding to a respiratory rate of 6 to 30 breaths per minute; and the body movement signal has a wider frequency range and can be distinguished based on the type of body movement.

[0029] Millimeter-wave radar sensors are pointed at the user's chest cavity to collect signals. The chest cavity is the torso region containing the heart and lungs, where both heartbeat and respiratory signals can be acquired simultaneously, with relatively high signal strength. Millimeter-wave radar sensors can employ multi-antenna array configurations, using beamforming technology to enhance signal directivity to the chest cavity, improving the signal-to-noise ratio and spatial resolution. Multi-antenna arrays also enable user differentiation in scenarios with simultaneous multi-user monitoring. When multiple users are present on a sofa, the millimeter-wave radar sensor can distinguish the physiological signals of different users based on spatial angle.

[0030] The raw physiological signal is bandpass filtered to remove DC components and high-frequency noise. The bandpass filter range is set from 0.5Hz to 10Hz. The lower cutoff frequency of 0.5Hz is used to remove DC components and extremely low-frequency drift. DC components mainly originate from the hardware bias of the millimeter-wave radar sensor and static reflections from the environment, while extremely low-frequency drift mainly originates from the user's slow body movement and changes in ambient temperature. The upper cutoff frequency of 10Hz is used to remove high-frequency noise, which mainly originates from electromagnetic interference, sensor electronic noise, and environmental vibration interference. Bandpass filtering can be implemented using digital filters, such as Butterworth filters, Chebyshev filters, or elliptic filters. Butterworth filters have the flattest amplitude-frequency response within the passband, making them suitable for applications requiring high phase linearity. Chebyshev filters have a steeper transition band response, making them suitable for applications requiring high frequency selectivity. Elliptic filters have the steepest transition band response, but ripples exist in both the passband and stopband. The filter order for bandpass filtering can be selected according to actual needs. The higher the filter order, the better the filtering effect, but the computational complexity also increases accordingly.

[0031] After bandpass filtering, the DC component and high-frequency noise in the original physiological signal are effectively removed. According to actual measurements, the signal-to-noise ratio of the pre-processed physiological signal can be improved by approximately 8.6 dB compared to the original signal. Furthermore, bandpass filtering can also employ adaptive filtering algorithms, dynamically adjusting filtering parameters based on real-time signal characteristics. When abnormal energy is detected in certain frequency components of the signal, the adaptive filtering algorithm can automatically enhance the suppression of those frequency components, further improving the filtering effect. Adaptive filtering algorithms can employ least mean squares, recursive least squares, or Kalman filtering algorithms, among others.

[0032] The filtered physiological signal is converted from analog to digital (ADC) to obtain a preprocessed digital physiological signal. ADC converts the analog signal into a digital signal for subsequent digital signal processing algorithms. The ADC resolution can be selected according to actual needs; higher resolution results in higher digital signal accuracy, but also increases the data volume. A 16-bit resolution is suitable, meeting the accuracy requirements of most physiological signal processing. The ADC sampling rate is consistent with that of the millimeter-wave radar sensor. After ADC, the preprocessed digital physiological signal is obtained, represented as a discrete-time series, denoted as x[n], where n is the time index and x[n] is the signal amplitude at the nth sampling point. The preprocessed digital physiological signal is stored in the system memory for subsequent feature extraction modules to access.

[0033] In one embodiment, in step S120, the multi-dimensional physiological characteristic indicators include heart rate variability characteristic indicators, respiratory regularity characteristic indicators, and body movement characteristic indicators. The extraction steps for each characteristic indicator are as follows: The steps for extracting heart rate variability characteristic indicators include: S1211. Separate the heartbeat signal component from the preprocessed heart rate signal; S1212. Identify the position of the R-wave peak in the heartbeat signal using a threshold detection algorithm; S1213. Based on the position of the R-wave peak, calculate the time interval between adjacent R-wave peaks to obtain the RR interval sequence; S1214. Calculate heart rate variability indices based on the RR interval sequence, wherein the heart rate variability indices include RMSSD index and LF / HF ratio index.

[0034] The heart rate signal component needs to be separated from the preprocessed heart rate signal. The preprocessed heart rate signal contains heart rate, respiratory, and other noise components, requiring signal separation methods to extract the heart rate signal component from the mixed signal. Adaptive filtering is used for heart rate component separation, leveraging the differences in frequency characteristics between the heart rate and respiratory signals. The frequency range of the heart rate signal is approximately 0.8 Hz to 2 Hz, corresponding to a heart rate of 48 to 120 beats per minute, while the frequency range of the respiratory signal is approximately 0.1 Hz to 0.5 Hz, corresponding to a respiratory rate of 6 to 30 breaths per minute. The adaptive filtering method estimates and subtracts the respiratory signal component from the mixed signal by establishing a reference signal model, retaining the heart rate signal component. The filter coefficients of the adaptive filtering method can be dynamically adjusted according to signal characteristics to track the time-varying characteristics of the heart rate signal. Adaptive filtering algorithms can employ least mean squares, recursive least squares, or Kalman filtering algorithms, among others. After adaptive filtering, the separated heartbeat signal components are obtained. The characteristics of the R-wave peak in the heartbeat signal components are more obvious, which facilitates the subsequent identification of the R-wave peak position. Heartbeat signal component separation can also employ blind source separation methods, such as independent component analysis (ICA), which decompose the mixed signal into multiple statistically independent source signals from which heartbeat signal components are selected. Blind source separation methods do not require a pre-established reference signal model and are suitable for scenarios where signal characteristics are unknown.

[0035] The R-wave peak position in the heartbeat signal is identified using a threshold detection algorithm. The R-wave peak is the highest amplitude peak in the electrocardiogram (ECG) signal, corresponding to the ventricular depolarization process and is the most prominent feature point in the heartbeat signal. The threshold detection algorithm first performs envelope detection on the heartbeat signal components, extracting the envelope line, which reflects the amplitude change trend of the heartbeat signal. Then, an adaptive threshold is set, which can be dynamically adjusted according to the signal amplitude. When an overall decrease in signal amplitude is detected, the adaptive threshold is lowered accordingly; when an overall increase in signal amplitude is detected, the adaptive threshold is raised accordingly. The heartbeat signal envelope line is compared with the adaptive threshold. When the envelope line crosses the adaptive threshold from below, it is marked as a candidate R-wave peak position. The candidate R-wave peak positions are verified under three conditions: amplitude, time interval, and waveform morphology. The amplitude condition requires that the amplitude of the candidate R-wave peak exceeds the adaptive threshold; the time interval condition requires that the time interval between adjacent candidate R-wave peaks be within a reasonable range, such as 0.4 seconds to 1.5 seconds, corresponding to a heart rate of 40 to 150 beats per minute; and the waveform morphology condition requires that the candidate R-wave peak exhibit typical R-wave morphological characteristics. The verified candidate R-wave peak positions are confirmed as the final R-wave peak positions. These positions are recorded as timestamps for subsequent RR interval calculations. R-wave peak position identification can also employ wavelet transform, utilizing its multi-resolution analysis capabilities to detect R-wave peak features at different scales, thereby improving the accuracy and robustness of R-wave peak identification.

[0036] The RR interval sequence is obtained by calculating the time interval between adjacent R-wave peaks based on their positions. The RR interval refers to the time interval between two adjacent R-wave peaks, reflecting the changes in the heartbeat cycle. The RR interval is calculated as the difference between the timestamps of adjacent R-wave peak positions, i.e.: in For the timestamp of the i-th R-wave peak, This is the timestamp of the (i+1)th R-wave peak. This refers to the i-th RR interval. In practice, the analysis window for the RR interval sequence is set to 5 minutes, and the sliding step size is set to 1 minute. That is, the characteristic indicators of the RR interval sequence are calculated every 1 minute, and each analysis uses the RR interval data within the most recent 5 minutes. The number of RR intervals within the 5-minute analysis window is approximately 300 to 600, depending on the user's heart rate level.

[0037] During RR interval sequence calculation, abnormal RR intervals need to be detected and removed. Abnormal RR intervals may be caused by incorrect R-wave peak identification or arrhythmias. The detection criteria for abnormal RR intervals include: an RR interval less than 0.3 seconds or greater than 2.0 seconds, and the difference between the current RR interval and the previous RR interval exceeding 20% ​​of the previous RR interval. Detected abnormal RR intervals are removed from the RR interval sequence and do not participate in subsequent heart rate variability index calculations. RR interval sequences can also undergo interpolation processing to convert non-uniformly sampled RR interval sequences into uniformly sampled time series, facilitating subsequent frequency domain analysis. Interpolation methods can include linear interpolation, spline interpolation, or cubic interpolation.

[0038] Heart rate variability (RMSSD) indices are calculated based on RR interval sequences. These indices include the RMSSD index and the LF / HF ratio. The RMSSD index reflects the parasympathetic nervous system's ability to rapidly regulate heart rate; RMSSD values ​​tend to decrease under fatigue conditions. The formula for calculating the RMSSD index is: Where N represents the number of RR intervals analyzed, ranging from 300 to 600, corresponding to 5 to 10 minutes of monitoring data. For the i-th RR interval, This refers to the (i+1)th RR interval. For example, when monitoring a user's RR interval sequence over a 5-minute period, if N equals 400 RR interval data points are measured, the sum of squared differences between adjacent intervals is 5800 ms. 2 If the baseline RMSSD is 4.5ms, then the current RMSSD is reduced by about 15%, indicating that there may be mild fatigue.

[0039] The LF / HF ratio reflects the balance between the sympathetic and parasympathetic nervous systems. An elevated LF / HF ratio indicates relatively hyperactive sympathetic nerve activity, commonly seen in fatigue. The LF / HF ratio is calculated using power spectrum estimation in the frequency domain. First, the power spectrum of the RR interval sequence is estimated. Power spectrum estimation methods can include Fourier transform, autoregressive modeling, or wavelet transform. The frequency band is divided into three bands: the VLF band (0.003Hz to 0.04Hz), the LF band (0.04Hz to 0.15Hz), and the HF band (0.15Hz to 0.4Hz). The power value of each band is calculated by integrating the power of each band. The power of the LF band is denoted as... HF band power is denoted as The formula for calculating the LF / HF ratio is: For example, a user calculated the LF band power to be 1200ms during continuous monitoring. 2The HF band power is 400ms. 2 If the LF / HF ratio is 1200 divided by 400, then the LF / HF ratio is 3.0. An LF / HF ratio exceeding the threshold of 2.5 indicates that the user may be in a state of fatigue.

[0040] In addition, heart rate variability indicators can also include other indicators such as the SDNN index, pNN50 index, and VLF power index. The SDNN index is the standard deviation of the RR interval, reflecting the overall level of heart rate variability. The pNN50 index is the percentage of adjacent RR intervals with a difference greater than 50 ms, reflecting parasympathetic nerve activity. The VLF power index is the power in the very low frequency band, reflecting long-term regulatory mechanisms such as thermoregulation and hormone secretion. Multiple heart rate variability indicators can be used in combination to form a more comprehensive heart rate variability assessment system.

[0041] The steps for extracting respiratory rhythm characteristic indicators include: S1221. Identify the start and end points of the respiratory cycle from the preprocessed respiratory signal; S1222. Calculate the time interval between adjacent respiratory cycles to obtain the respiratory interval sequence; S1223. Calculate respiratory regularity indicators based on the respiratory interval sequence, wherein the respiratory regularity indicators include respiratory interval standard deviation and respiratory variability coefficient.

[0042] The preprocessed respiratory signal is the respiratory component of the digital physiological signal after bandpass filtering and analog-to-digital conversion. The frequency range of the respiratory signal is approximately 0.1Hz to 0.5Hz, corresponding to a respiratory rate of 6 to 30 breaths / minute. The start and end points of the respiratory cycle are identified using a respiratory event detection algorithm. This algorithm first performs envelope detection on the preprocessed respiratory signal to extract the envelope, which reflects the amplitude variation trend of the respiratory signal. Then, a peak detection algorithm is used to identify the inspiratory peak and expiratory trough of the respiratory signal. The inspiratory peak corresponds to the end of the inspiratory process, and the expiratory trough corresponds to the end of the expiratory process. The start point of the respiratory cycle is defined as the time point between two consecutive expiratory troughs, and the end point of the respiratory cycle is defined as the time point of the next expiratory trough. A complete respiratory cycle includes one inspiratory and one expiratory process. The peak detection algorithm sets an adaptive threshold, with an initial value set at 30% to 50% of the maximum value of the respiratory signal envelope. The adaptive threshold can be dynamically adjusted according to the signal amplitude. The detected inspiratory peaks and expiratory troughs are validated under three conditions: amplitude, time interval, and waveform morphology. The amplitude condition requires the amplitude of the inspiratory peak and expiratory trough to exceed an adaptive threshold. The time interval condition requires the time interval between adjacent inspiratory peaks or expiratory troughs to be within a reasonable range, such as 2 to 10 seconds, corresponding to a respiratory rate of 6 to 30 breaths / minute. The waveform morphology condition requires the inspiratory peak and expiratory trough to exhibit typical respiratory waveform characteristics. The validated positions of the inspiratory peak and expiratory trough are confirmed as the final start and end points of the respiratory cycle. These start and end points are recorded as timestamps for subsequent respiratory interval calculations. The start and end points of the respiratory cycle can also be identified using wavelet transform. Leveraging the multi-resolution analysis characteristics of wavelet transform, respiratory cycle features can be detected at different scales, improving the accuracy and robustness of respiratory cycle identification. Wavelet transform is particularly suitable for scenarios where respiratory signals exhibit baseline drift or amplitude variations.

[0043] The respiratory interval sequence is obtained by calculating the time interval between adjacent respiratory cycles. The respiratory interval refers to the time interval between the start of two adjacent respiratory cycles, reflecting changes in respiratory rhythm. The respiratory interval is calculated as the difference between the timestamps of the start of adjacent respiratory cycles, i.e.: in The timestamp of the start of the j-th respiratory cycle. This is the timestamp of the start of the (j+1)th respiratory cycle. This represents the j-th respiratory interval. The analysis window for the respiratory interval sequence is set to 10 minutes, meaning that each analysis uses respiratory interval data from the most recent 10 minutes. The number of respiratory intervals within the 10-minute analysis window is approximately 60 to 120, depending on the user's respiratory rate level.

[0044] During the calculation of respiratory interval sequences, it is necessary to detect and remove abnormal respiratory intervals. Abnormal respiratory intervals may be caused by errors in respiratory cycle identification or abnormal breathing such as apnea or deep breathing. The detection criteria for abnormal respiratory intervals include: a respiratory interval less than 1 second or greater than 15 seconds, and a difference between the current respiratory interval and the previous respiratory interval exceeding 30%. Detected abnormal respiratory intervals are removed from the respiratory interval sequence and do not participate in subsequent calculations of respiratory regularity indicators. The respiratory interval sequence can also undergo interpolation processing to convert the non-uniformly sampled respiratory interval sequence into a uniformly sampled time series, facilitating subsequent frequency domain analysis. Interpolation methods can include linear interpolation, spline interpolation, or cubic interpolation. The respiratory interval sequence can also be time-synchronized with the RR interval sequence in the heart rate variability characteristic index to ensure consistency between the two physiological characteristic indicators in the time dimension, facilitating subsequent multi-dimensional feature fusion.

[0045] Respiratory regularity indices are calculated based on respiratory interval sequences. These indices include the standard deviation of respiratory intervals and the coefficient of variability. The standard deviation of respiratory intervals reflects the dispersion of respiratory intervals; an increased standard deviation under fatigue indicates respiratory rhythm disorder. The formula for calculating the standard deviation of respiratory intervals is: Where M represents the total number of respiratory intervals, ranging from 60 to 120, corresponding to 10 minutes of monitoring data. For the j-th breathing interval, This refers to the average respiratory interval. For example, if a user's respiratory data is monitored for 10 minutes, and a total of 80 valid respiratory intervals (M) are detected, then the average respiratory interval is... The baseline SDRI is 4.0 seconds, and the standard deviation of the respiratory interval is 0.32 seconds. If the user's baseline SDRI is 0.2 seconds, then the current SDRI increases by 60%, indicating a significant disturbance in the respiratory rhythm and possibly a state of fatigue. The normal range for the standard deviation of the respiratory interval is less than 400 ms. When the standard deviation of the respiratory interval exceeds 400 ms, it indicates a significant disturbance in the respiratory rhythm.

[0046] The respiratory variability coefficient reflects the relative variability of respiratory intervals, eliminating the influence of the absolute value of respiratory rate on the degree of variability, and facilitating comparisons between different users. The formula for calculating the respiratory variability coefficient is: Wherein, SDRI is the standard deviation of respiratory interval. The respiratory interval is the average breathing interval. The normal range for the respiratory variability coefficient is less than 40%. When the respiratory variability coefficient exceeds 40%, it indicates a significant disturbance in respiratory rhythm, possibly indicating fatigue. Respiratory regularity indicators may also include baseline respiratory rate, respiratory amplitude variability, and the inspiratory-to-expiratory ratio (IPR). The baseline respiratory rate reflects the user's average respiratory rate, the respiratory amplitude variability reflects the degree of variation in respiratory depth, and the IPR reflects the ratio of inspiratory to expiratory time. Multiple respiratory regularity indicators can be used in combination to form a more comprehensive respiratory regularity assessment system. Respiratory regularity indicators can also be analyzed in conjunction with heart rate variability indicators, such as calculating the respiratory sinus arrhythmia index, which reflects the modulating effect of respiration on heart rate and is an important indicator for assessing autonomic nervous system function.

[0047] The steps for extracting body movement characteristic indicators include: S1231. Detect body movement events from the preprocessed body movement signals; S1232. Merge short-term physical events with adjacent time sequences to obtain a complete sequence of physical events; S1233. Identify the body movement patterns in the complete sequence of body movement events, including the restless body movement pattern, the posture changing body movement pattern, the head-down body movement pattern, and the trunk tilting body movement pattern. S1234. Calculate body movement characteristic indicators based on the body movement events and body movement patterns. The body movement characteristic indicators include body movement frequency indicators, body movement amplitude indicators, and body movement pattern scoring indicators.

[0048] The preprocessed body movement signal is the body movement component in the digital physiological signal after bandpass filtering and analog-to-digital conversion. Body movement signal detection employs a threshold detection method. First, the instantaneous energy of the body movement signal is calculated as the square of the signal amplitude. Then, an adaptive threshold is set, initially 1.5 to 2.5 times the baseline instantaneous energy of the body movement signal. This threshold can be dynamically adjusted based on the ambient noise level. When the instantaneous energy of the body movement signal exceeds the adaptive threshold, a body movement event is detected, and the start timestamp of the event is recorded. The end time of the body movement event is determined by the moment the instantaneous energy falls below the adaptive threshold, and the end timestamp of the event is recorded. The duration of the body movement event is the difference between the end timestamp and the start timestamp.

[0049] Merging short-duration physical movement events with adjacent time sequences yields a complete sequence of physical movement events. Since a single complete physical movement action by a user may manifest as multiple adjacent short-duration physical movement events in the signal, these short-duration physical movement events need to be merged into a complete physical movement event. The merging conditions include temporal adjacency and interval time conditions. The temporal adjacency condition requires that the two physical movement events be adjacent on the time axis, meaning there are no other physical movement events between the end time of the preceding physical movement event and the start time of the following physical movement event. The interval time condition requires that the time interval between the two physical movement events be less than a preset merging threshold, which can be set to 0.5 seconds to 2 seconds. When two physical movement events simultaneously meet both the temporal adjacency and interval time conditions, the two physical movement events are merged into a single complete physical movement event. The start time of the merged complete physical movement event is the start time of the preceding physical movement event, and the end time is the end time of the following physical movement event. The merging process is iterated until merging can no longer continue, ultimately resulting in a complete sequence of physical movement events. Body movement event merging can also employ clustering algorithms, such as density-based clustering, which groups temporally adjacent body movement events into the same cluster, with each cluster corresponding to a complete body movement event. Clustering algorithms can automatically determine the number of body movement events to be merged without requiring a preset merging threshold, making them suitable for scenarios with complex body movement patterns.

[0050] This study identifies body movement patterns within a complete sequence of body movement events. These patterns include four types: restless body movement, posture changing body movement, head-down body movement, and trunk-tilting body movement. Body movement pattern recognition employs a pattern recognition algorithm. First, it extracts body movement pattern features from the complete body movement event. These features include duration, amplitude, frequency, and waveform characteristics. Duration features represent the duration of the complete body movement event; amplitude features represent the maximum amplitude of the body movement signal; frequency features represent the number of body movements per unit time; and waveform features represent the waveform shape descriptor of the body movement signal. Then, a classifier is used to classify the body movement pattern features. The classifier can be a rule-based classifier or a machine learning classifier. The rule-based classifier determines body movement patterns according to preset rules. For example, the rules for determining an uneasy body movement pattern are that the movement duration is less than 5 seconds and the movement frequency is higher than a preset threshold; the rules for determining a posture-changing body movement pattern are that the movement amplitude is greater than a preset threshold and the movement duration is between 5 and 30 seconds; the rules for determining a head-down body movement pattern are that the body movement signal shows a specific slowly decreasing waveform and lasts for more than 10 seconds; and the rules for determining a trunk-tilting body movement pattern are that the body movement signal shows a specific tilting waveform and lasts for more than 5 seconds. The machine learning classifier learns the mapping relationship between body movement pattern features and body movement pattern types through training data. The classifier can use algorithms such as support vector machines, random forests, gradient boosting trees, or neural networks. The identified body movement patterns are attached to the complete body movement event in the form of labels for subsequent calculation of body movement feature indicators.

[0051] Body movement characteristic indicators are calculated based on body movement events and patterns. These indicators include body movement frequency, body movement amplitude, and body movement pattern scoring. The body movement frequency indicator is defined as the number of body movements within a preset time window, typically set to 10 minutes. The formula for calculating the body movement frequency indicator is as follows: in, The number of physical events within a preset time window. The preset time window duration (in minutes) is defined by MF, which stands for Body Movement Frequency. For example, if a user experiences 18 body movement events within 10 minutes (MF = 18 events), exceeding the threshold of 15 events per 10 minutes, it indicates a significant pattern of restless body movement and potential fatigue. The normal range for body movement frequency is less than 15 events per 10 minutes. When the body movement frequency exceeds 15 events per 10 minutes, it suggests frequent body movement and potential fatigue. Body movement amplitude reflects the intensity of body movement events. It is calculated as the average or maximum amplitude of all body movement events in a complete sequence. The formula for calculating body movement amplitude is: Where K represents the number of whole-body motion events. Let MA be the amplitude of the k-th complete physical activity event. The physical activity amplitude index (MA) may increase or decrease under fatigue conditions, requiring assessment in conjunction with the user's personalized baseline. The physical activity pattern scoring index reflects the frequency and severity of abnormal physical activity patterns. The formula for calculating the physical activity pattern scoring index is: Where K represents the number of whole-body motion events. Here are the weights for the movement patterns corresponding to the k-th complete movement event: the weight for the restless movement pattern is 1.0, the weight for the posture change movement pattern is 0.8, the weight for the head-down movement pattern is 1.2, and the weight for the trunk-tilt movement pattern is 1.0. This is a body movement pattern rating index. The normal range for the body movement pattern rating index is less than 0.6. When the body movement pattern rating index exceeds 0.6, it indicates that the user has an abnormal body movement pattern and may be in a state of fatigue.

[0052] In addition, physical activity indicators may include other indicators such as duration of physical activity, interval of physical activity, and energy of physical activity. Duration of physical activity reflects the average duration of a physical activity event, interval of physical activity reflects the time interval between adjacent physical activity events, and energy of physical activity reflects the total energy of a physical activity event. Multiple physical activity indicators can be used in combination to form a more comprehensive assessment system. Physical activity indicators can also be analyzed in conjunction with heart rate variability and respiratory regularity indicators. For example, when the physical activity frequency indicator increases while the LF / HF ratio in the heart rate variability indicator also increases, it can more accurately determine that the user is in a state of fatigue.

[0053] In one embodiment, step S130 includes: S131. Calculate the signal quality assessment value corresponding to each dimension of physiological characteristic index; S132. Set the basic weights for each dimension of physiological characteristic indicators; S133. Calculate the total quality score based on the signal quality assessment values ​​and basic weights of the physiological characteristic indicators of each dimension. S134. Based on the total quality score, the basic weights of the physiological characteristic indicators of each dimension are normalized and adjusted to obtain adaptive weights; S135. Perform boundary constraint processing on the adaptive weights to ensure that the weights of each dimension are within the preset weight range.

[0054] Signal quality assessment values ​​were calculated for each dimension of physiological characteristic indicators. These values ​​quantify the reliability of each indicator, ranging from 0 to 1, with higher values ​​indicating better signal quality. For heart rate variability (RR) indicators, the RR variability signal quality assessment value was calculated based on three sub-indicators: RR interval detection signal-to-noise ratio (SNR), RR interval validity, and signal stability. The RR interval detection SNR reflects the energy ratio of signal to noise during RR interval detection. It is calculated as the ratio of the R-wave peak energy to the background noise energy in the heartbeat signal component, normalized to a range of 0 to 1. RR interval validity reflects the proportion of valid RR intervals among the detected RR intervals. It is calculated as the ratio of the number of valid RR intervals to the total number of RR intervals. Valid RR intervals refer to those that passed anomaly detection.

[0055] Signal stability reflects the stability of a heart rate signal within a time window. It is calculated as the normalized reciprocal of the heart rate signal's standard deviation. Heart rate variability signal quality assessment values. The calculation formula is: Where Q1 is the signal-to-noise ratio of the RR interval detection, Q2 is the RR interval validity, and Q3 is the signal stability. The values ​​of Q1, Q2, and Q3 are all in the range of 0 to 1.

[0056] For indicators of respiratory regularity, the quality assessment value of the respiratory regularity signal is calculated based on two sub-indicators: the respiratory signal-to-noise ratio (SNR) and the respiratory event detection completeness rate. The SNR reflects the energy ratio of the effective signal to the noise in the respiratory signal, while the respiratory event detection completeness rate reflects the proportion of detected respiratory events to actual respiratory events. The respiratory regularity signal quality assessment value... The calculation formula is: Where R1 is the signal-to-noise ratio of the respiratory signal and R2 is the completeness rate of respiratory event detection. The values ​​of R1 and R2 are both in the range of 0 to 1.

[0057] For body motion characteristic indicators, the calculation of the body motion signal quality assessment value is based on two sub-indicators: signal stability and environmental interference suppression effect. Signal stability reflects the stability of the body motion signal within a time window, while environmental interference suppression effect reflects the system's ability to suppress environmental interference. Body motion signal quality assessment value The calculation formula is: Where M1 represents signal stability and M2 represents environmental interference suppression effect, the values ​​of M1 and M2 are both in the range of 0 to 1.

[0058] In addition, more sub-indicators can be incorporated into the signal quality assessment value, such as signal continuity index and signal amplitude consistency index. The comprehensive signal quality assessment value can be calculated by weighted averaging, which can further improve the accuracy of signal quality assessment.

[0059] Basic weights were set for each dimension of physiological characteristic indicators. These basic weights represent the initial weight allocation for each physiological characteristic indicator under ideal signal quality conditions. The basic weights were determined based on the contribution of each physiological characteristic indicator to the assessment of fatigue state. Heart rate variability, directly reflecting autonomic nervous system function, has the highest correlation with fatigue state; therefore, its basic weight was set to 0.4. Respiratory regularity, reflecting respiratory regulation function, has a relatively high correlation with fatigue state; its basic weight was set to 0.3. Body movement, reflecting user behavior patterns, has a relatively high correlation with fatigue state; its basic weight was also set to 0.3. The sum of the basic weights for the three dimensions is 1.0, satisfying the weight normalization condition.

[0060] The base weights can be dynamically set according to the application scenario. For example, in a nighttime sleep monitoring scenario, the base weight of the breathing regularity characteristic indicator can be appropriately increased, and in a daytime activity monitoring scenario, the base weight of the body movement characteristic indicator can be appropriately increased. The base weights can also be personalized according to individual user differences. By analyzing users' historical data through machine learning methods, the allocation ratio of base weights for each dimension can be automatically optimized.

[0061] The total quality score is calculated based on the signal quality assessment values ​​and basic weights of the physiological characteristic indicators of each dimension. The total quality score reflects the comprehensive signal quality level of all physiological characteristic indicators under the current monitoring environment.

[0062] The overall quality score is calculated as the sum of the products of the base weights of each dimension and the corresponding signal quality assessment values. The calculation formula is: in, The basic weights for heart rate variability characteristic indicators, This is a value used to assess the quality of the heart rate variability signal. As the basic weight of respiratory regularity characteristic indicators, This is a quality assessment value for respiratory rhythm signals. The basic weights for body movement characteristic indicators. This is a value used to assess the quality of body movement signals. For example, when... It equals 0.6. It equals 0.9. When it equals 0.8, It equals 0.75. The total quality score ranges from 0 to 1, with a higher value indicating better overall signal quality.

[0063] Adaptive weights are obtained by normalizing and adjusting the base weights of each physiological characteristic indicator based on the total quality score. These adaptive weights dynamically adjust the weights of each physiological characteristic indicator according to real-time signal quality, ensuring accurate fatigue assessment results even in complex environments. The adaptive weights are calculated by multiplying the base weight of each dimension by its corresponding signal quality assessment value and dividing by the total quality score. Adaptive weights for heart rate variability characteristic indicators are also described. The calculation formula is: Adaptive weights of respiratory regularity characteristic indicators The calculation formula is: Adaptive weights of body movement characteristic indicators The calculation formula is: Continuing with the previous example, at this time... It equals 0.32; It equals 0.36; 0.32. After adjustment, the weight of the heart rate variability characteristic index decreased from 0.4 to 0.32, the weight of the respiratory regularity characteristic index increased from 0.3 to 0.36, and the weight of the body movement characteristic index increased from 0.3 to 0.32.

[0064] The adaptive weight adjustment mechanism ensures that when the signal quality of a certain dimension decreases, the weight of that dimension is automatically reduced, while the weight of dimensions with better signal quality is increased compensatorily, ensuring that the evaluation results are not excessively affected by low-quality signals.

[0065] Boundary constraints are applied to the adaptive weights to ensure that the weights of each dimension remain within a preset range. This boundary constraint process prevents the adaptive weight calculation results from exceeding a reasonable range, thus guaranteeing the rationality of the weight allocation. Specifically, for example, the weight constraint range for heart rate variability indicators is from a lower limit of 0.15 to an upper limit of 0.55, the weight constraint range for respiratory regularity indicators is from a lower limit of 0.10 to an upper limit of 0.45, and the weight constraint range for body movement indicators is from a lower limit of 0.10 to an upper limit of 0.45.

[0066] The condition judgment method for boundary constraint processing is as follows: when the adaptive weight calculation result is less than the lower limit, the weight is set to the lower limit value; when the adaptive weight calculation result is greater than the upper limit, the weight is set to the upper limit value; when the adaptive weight calculation result is between the upper and lower limits, the adaptive weight calculation result remains unchanged. After boundary constraint processing, weight normalization verification can be performed to ensure that the sum of the weights of the three dimensions is 1.0. If the sum of the weights is not equal to 1.0, the weights of each dimension are scaled proportionally to make the sum of the weights equal to 1.0.

[0067] In practice, the weight constraint range can be dynamically adjusted according to the application scenario. For example, the weight constraint range can be appropriately widened in environments with severe signal interference, and appropriately tightened in environments with stable signal quality. The weight constraint range can also be personalized according to individual user differences, and the optimal weight constraint range can be determined by analyzing historical user data.

[0068] In one embodiment, step S130 further includes: S136. Based on preset physiological state triggering rules, detect physiological state triggering conditions, including heart rate variability abnormality triggering conditions, respiratory disorder triggering conditions, and body movement abnormality triggering conditions. S137. When any of the physiological state trigger conditions are met, the adaptive weights of the corresponding dimensions are adjusted accordingly.

[0069] Physiological state triggering conditions are detected based on preset physiological state triggering rules. These rules are a pre-defined set of rules used to determine whether a user's physiological state is abnormal. Physiological state triggering conditions include three categories: abnormal heart rate variability (HRV) triggering conditions, respiratory disturbance triggering conditions, and abnormal body movement triggering conditions. The detection of abnormal HRV triggering conditions is based on the LF / HF ratio, a heart rate variability characteristic index that reflects the balance between the sympathetic and parasympathetic nervous systems. The determination rule for abnormal HRV triggering conditions is that the LF / HF ratio exceeds a preset sympathetic nerve activity threshold. When the LF / HF ratio is greater than the preset sympathetic nerve activity threshold, it is determined that the abnormal HRV triggering condition is met, indicating a significant increase in sympathetic nerve activity and that the user may be in a state of fatigue.

[0070] The detection of respiratory disturbance triggers is based on the respiratory interval standard deviation and respiratory variability coefficient indicators from the respiratory rhythm characteristic indicators. The rule for determining a respiratory disturbance trigger is that either the respiratory interval standard deviation exceeds a preset respiratory rhythm threshold, or the respiratory variability coefficient exceeds a preset respiratory variability threshold. When either the respiratory interval standard deviation or the respiratory variability coefficient exceeds the preset respiratory rhythm threshold, the respiratory disturbance trigger condition is met, indicating a significant respiratory rhythm disorder and that the user may be in a state of fatigue. The detection of abnormal body movement triggers is based on the body movement frequency indicator from the body movement characteristic indicators. The rule for determining an abnormal body movement trigger is that the body movement frequency exceeds a preset body movement frequency threshold. When the body movement frequency exceeds the preset body movement frequency threshold, the abnormal body movement trigger condition is met, indicating that the user is exhibiting frequent restless body movements and may be in a state of fatigue.

[0071] When any physiological state trigger condition is met, the adaptive weights of the corresponding dimension are adjusted. This adjustment is a further weight adjustment based on the adaptive weights, used to enhance the sensitivity to abnormal physiological states. When the heart rate variability (HRV) abnormality trigger condition is met, the adaptive weights of the HRV feature indicators are adjusted by increasing the weights of the HRV feature indicators. The adjusted weights of the HRV feature indicators are shown below. The calculation formula is: in, The adaptive weights for the heart rate variability characteristic indicators calculated above, This is the adjustment coefficient for heart rate variability (FRV). When the LF / HF ratio exceeds a preset threshold, it indicates a significant increase in sympathetic nerve activity. At this point, the weight of the FRV characteristic index is increased, making the fatigue assessment results more focused on changes in the FRV dimension. After adjustment, weight normalization is required to ensure that the sum of the weights of the three dimensions is 1.0.

[0072] When the respiratory disturbance trigger condition is met, the adaptive weights of the respiratory regularity characteristic indicators and heart rate variability characteristic indicators are adjusted. The adjustment method involves increasing the weight of the respiratory regularity characteristic indicators while simultaneously adjusting the weight of the heart rate variability characteristic indicators. The adjusted weights of the respiratory regularity characteristic indicators are shown below. The calculation formula is: in, The adaptive weights of the respiratory regularity characteristic indicators calculated above, This is the respiratory trigger adjustment coefficient. When the standard deviation of the respiratory interval exceeds a preset threshold or the coefficient of respiratory variability exceeds a preset threshold, it indicates a significant disturbance in the respiratory rhythm. In this case, the weight of the respiratory regularity characteristic indicators is increased, making the fatigue assessment results more focused on changes in the respiratory regularity dimension. Simultaneously, since there is a physiological coupling relationship between respiration and heart rate, the weight of the heart rate variability characteristic indicators also needs to be adjusted accordingly when respiration is disordered.

[0073] When the abnormal body movement trigger condition is met, the adaptive weights of the body movement characteristic indicators are adjusted by increasing the weights of the body movement characteristic indicators. The adjusted weights of the body movement characteristic indicators are then determined. The calculation formula is: in, These are the adaptive weights for the body motion characteristic indices calculated earlier. This is the body movement trigger adjustment coefficient. When the body movement frequency index exceeds a preset threshold, it indicates that the user is exhibiting frequent restless body movements. In this case, the weight of the body movement characteristic index is increased, making the fatigue assessment results more focused on changes in the body movement dimension. The weights after trigger adjustment still need to undergo boundary constraint processing to ensure that the weights of each dimension are within the preset weight range described above.

[0074] In one embodiment, step S140 includes: S141. Obtain the user's personalized physiological baseline, which includes the baseline values ​​of physiological characteristic indicators of each dimension obtained from the statistical analysis of the user's historical physiological data. S142. Compare the real-time extracted physiological characteristic indicators of each dimension with the corresponding baseline values, and calculate the degree of deviation of each physiological characteristic indicator from the baseline values. S143. Calculate the normalized score for each dimension based on the degree of deviation. S144. The normalized scores of each dimension are weighted and summed with the adjusted weights to obtain the comprehensive fatigue score.

[0075] The system acquires a personalized physiological baseline for each user. This baseline includes baseline values ​​for various physiological characteristic indicators derived from the user's historical physiological data. These baseline values ​​include those for heart rate variability, respiratory regularity, and body movement. The acquisition of the personalized physiological baseline is divided into a baseline acquisition phase and a baseline update phase. The baseline acquisition phase is conducted when the user first uses the fatigue state assessment method. The system continuously collects physiological data for a preset duration, with the daily collection time set to a fixed period during the user's resting state.

[0076] During the baseline acquisition phase, the system calculates the initial baseline values ​​for each dimension of physiological characteristic indicators. The initial baseline values ​​are calculated as the arithmetic mean or median of all valid measurements within the acquisition period. For example, the baseline values ​​for heart rate variability characteristic indicators include the baseline RMSSD value and the baseline LF / HF ratio; the baseline values ​​for respiratory regularity characteristic indicators include the baseline respiratory interval standard deviation value and the baseline respiratory variability coefficient value; and the baseline values ​​for body movement characteristic indicators include the baseline body movement frequency value.

[0077] The baseline update phase occurs during subsequent use. The system periodically and dynamically updates the baseline value to adapt to long-term changes in the user's physiological state. The baseline update uses a sliding window update mechanism, and the update formula is as follows: in, For the updated baseline value, This is the current measurement value. The baseline value is the value before the update, and α is the update coefficient. The update coefficient α controls the speed of baseline updates. The larger the α value, the faster the baseline update and the higher the sensitivity to recent data; the smaller the α value, the slower the baseline update and the more stable the baseline value. Baseline updates can be set to be performed once a day, or once after a preset number of new measurements have been accumulated. The real-time extracted physiological characteristic indicators of each dimension are compared with the corresponding baseline values ​​to calculate the deviation of each physiological characteristic indicator relative to the baseline value. The deviation reflects the magnitude of change of the real-time physiological indicator relative to the user's own baseline. The greater the deviation, the higher the possible level of fatigue. The deviation is calculated by dividing the difference between the real-time indicator value and the baseline value by the baseline value to obtain the relative change ratio. For the RMSSD indicator in the heart rate variability characteristic indicators, since the RMSSD value shows a decreasing trend under fatigue, the deviation is calculated by subtracting the real-time value from the baseline value and then dividing by the baseline value. The deviation of the RMSSD indicator is then calculated. The calculation formula is: in, The baseline RMSSD value. This refers to the real-time RMSSD value. For example, if the user's baseline RMSSD is 4.5ms and the real-time RMSSD is 3.6ms, then... It equals 0.2, which is a deviation of 20%.

[0078] For the LF / HF ratio, a characteristic indicator of heart rate variability, since the LF / HF ratio tends to increase under fatigue conditions, the deviation is calculated by subtracting the baseline value from the real-time value and then dividing by the baseline value. The degree of deviation of the LF / HF ratio indicator... The calculation formula is: in, The baseline LF / HF ratio, This refers to the real-time LF / HF ratio. For example, if the user's baseline LF / HF ratio is 1.2 and the real-time LF / HF ratio is 2.0, then... It equals 0.67, which is a deviation of 67%.

[0079] For the respiratory interval standard deviation and respiratory variability coefficient, which are among the respiratory regularity indicators, since these indicators tend to increase under fatigue, the deviation is calculated similarly to the LF / HF ratio: real-time value minus baseline value, then divided by the baseline value. For the body movement frequency, which is among the body movement indicators, since body movement frequency tends to increase under fatigue, the deviation is also calculated as real-time value minus baseline value, then divided by the baseline value.

[0080] Normalized scores are calculated for each dimension based on the degree of deviation. The normalized score converts the deviation of each physiological characteristic indicator in each dimension into a unified scoring scale. The calculation of the normalized score comprehensively considers the deviation of multiple indicators within a dimension, and obtains a comprehensive normalized score for the dimension through a weighted average. Taking heart rate variability as an example, the normalized score of the heart rate variability characteristic indicator... Taking into account the deviations of both the RMSSD and LF / HF ratios, the formula for calculating the normalized score is as follows: in, The degree of deviation of the RMSSD metric. The deviation of the LF / HF ratio index is represented by α1 and α2, which are weighting coefficients, and the sum of α1 and α2 is 1.0. For example, when... Equals 0.2, When it equals 0.67, It equals 0.39. If the normalized score range is converted to 0 to 100, then... It equals 39 points.

[0081] Normalized scores of respiratory regularity characteristics The calculation formula is: in, The degree of deviation of the standard deviation of respiratory interval index. The coefficient of variation in respiration represents the degree of deviation. β1 and β2 are weighting coefficients, and the sum of β1 and β2 is 1.0. The values ​​of β1 and β2 can be equal.

[0082] Normalized scores of body movement characteristics The calculation formula is: in, The degree of deviation from the body movement frequency index. The degree of deviation of the body movement amplitude index. The deviation of the body movement pattern scoring index is represented by γ1, γ2, and γ3, which are weighting coefficients. The sum of the three is 1.0. For example, γ1 is 0.5, γ2 is 0.25, and γ3 is 0.25.

[0083] The fatigue score is obtained by weighting and summing the normalized scores of each dimension with the adjusted weights. The fatigue score reflects the user's overall fatigue level, ranging from 0 to 100 points, with higher scores indicating more severe fatigue. The formula for calculating the Fatigue Score (FDS) is as follows: in, The adjusted weights for heart rate variability characteristics. Normalized scores for heart rate variability characteristics. The adjusted weights of the respiratory regularity characteristic indicators, Normalized scores for respiratory regularity characteristics indicators. The adjusted weights for the body movement characteristic indicators. Normalized scores for body movement characteristics indicators. , , The final weights, after adaptive adjustment based on signal quality and physiological state triggering, are given above, and the sum of these three weights is 1.0. For example, when... Equals 0.32, Equals 39 points. It equals 0.36. Equals 45 points. Equals 0.32, When the score is 50, the FDS score is 44.68, which is approximately 45. A fatigue score of 45 falls within the range of 26 to 50, corresponding to Level 1 mild fatigue.

[0084] In one embodiment, step S150 includes: S151. The fatigue comprehensive score of the continuous time series is smoothed by the exponential weighted moving average algorithm to obtain the smoothed fatigue comprehensive score. S152. Calculate the difference between the current fatigue comprehensive score and the previous smoothed fatigue comprehensive score; S153. Determine whether the difference exceeds a preset fluctuation threshold; S154. When the difference exceeds the preset fluctuation threshold, outlier suppression processing is performed on the current fatigue comprehensive score to limit the score change within the preset fluctuation threshold range, thereby obtaining the corrected fatigue comprehensive score.

[0085] The fatigue score of a continuous time series is smoothed using an exponentially weighted moving average algorithm to obtain a smoothed fatigue score. The exponentially weighted moving average algorithm is a commonly used time series smoothing algorithm that assigns exponentially decaying weights to historical data, making recent data have a greater impact on the smoothing result while preserving the trend information of historical data. The calculation formula for the exponentially weighted moving average algorithm is as follows: in, The smoothed fatigue score is the overall score at the k-th time point. The original fatigue score at time point k. The smoothed fatigue score is the overall score at time point k-1, where δ is the smoothing coefficient. The smoothing coefficient δ controls the degree of smoothing; a larger δ value results in a weaker smoothing effect and a faster response to recent data, while a smaller δ value results in a stronger smoothing effect and a slower response to recent data. In the initial calculation... It can be initialized to the first original fatigue comprehensive score FDS(1), or initialized to a preset intermediate value. After exponential weighted moving average smoothing, the short-term fluctuations of the fatigue comprehensive score are effectively suppressed, and the score changes are more stable and continuous.

[0086] The difference between the current fatigue score and the previous smoothed fatigue score is calculated. This difference is used to detect abnormal fluctuations in the fatigue score. The difference is calculated as the absolute value of the difference between the current fatigue score and the previous smoothed fatigue score. The formula for calculating the difference ΔFDS is: Where FDS(k) is the current fatigue comprehensive score at the k-th time point. Let ΔFDS be the smoothed fatigue score at time point k-1, and ΔFDS be the absolute value of the difference. The difference ΔFDS reflects the magnitude of change in the current score relative to historical scores; a larger difference indicates a more drastic change. The difference can also be calculated using a relative difference method, where the difference is divided by the previous smoothed fatigue score to obtain the relative change ratio. The formula for calculating the relative difference is: Relative differences can eliminate the influence of absolute scores on the judgment of the magnitude of change, and are suitable for detecting abnormal fluctuations at different scoring levels. The difference calculation can also take into account the time interval factor. When the time interval between two adjacent scores is long, a larger difference is allowed, and when the time interval is short, a stricter difference judgment standard is adopted.

[0087] The system determines whether the difference exceeds a preset fluctuation threshold. This threshold is used to determine whether changes in the overall fatigue score constitute abnormal fluctuations. The preset fluctuation threshold can be set based on the range of the overall fatigue score and the gradual nature of fatigue. For example, a preset fluctuation threshold of 20 points is used. In this case, the criteria are: if ΔFDS is greater than 20 points, the difference exceeds the preset fluctuation threshold, indicating abnormal fluctuations in the current overall fatigue score; if ΔFDS is less than or equal to 20 points, the difference does not exceed the preset fluctuation threshold, indicating normal changes in the current overall fatigue score. The preset fluctuation threshold of 20 points is based on the gradual accumulation of fatigue, a slow process. Between two consecutive assessments, the change in the overall fatigue score typically does not exceed 20 points. If the change exceeds 20 points, it may be an abnormal value caused by signal interference, feature extraction errors, or temporary physiological fluctuations.

[0088] When the difference exceeds a preset fluctuation threshold, outlier suppression processing is applied to the current fatigue comprehensive score to limit the score variation within the preset fluctuation threshold range, resulting in a corrected fatigue comprehensive score. Outlier suppression processing avoids excessive influence of abnormal fluctuations on the fatigue assessment results by limiting the score variation. The calculation formula for outlier suppression processing is as follows: in, This is the revised overall fatigue score. The smoothed fatigue score is the overall score at the (k-1)th time point. The sign function of the difference, when FDS(k) is greater than When FDS(k) is less than 1, the value is 1. Take -1 at time, The minimum value between the absolute value of the difference and the preset fluctuation threshold, where Threshold is the preset fluctuation threshold. For example, when... When the score is 45, FDS(k) is 80, ΔFDS is 35, and Threshold is 20, outlier suppression is triggered because 35 is greater than 20. The score is 65. After suppression, the score smoothly transitions from 80 to 65, avoiding score jumps caused by misjudgments. If ΔFDS is less than or equal to 20, outlier suppression is not triggered. It equals FDS(k).

[0089] In one embodiment, step S160 includes: S161. Establish a mapping relationship between the comprehensive fatigue score and the fatigue level, wherein the fatigue level includes at least the mental well-being level, mild fatigue level, moderate fatigue level and severe fatigue level. S162. Match the corrected fatigue comprehensive score with the score interval in the mapping relationship; S163. Determine the user's current fatigue level based on the matching results; S164. Output fatigue assessment results containing the fatigue level.

[0090] A mapping relationship was established between a comprehensive fatigue score and fatigue levels, which were divided into four levels: high energy level, mild fatigue level, moderate fatigue level, and severe fatigue level. Each fatigue level corresponds to a different comprehensive fatigue score range, with the score ranges increasing sequentially and not overlapping. The high energy level corresponds to a comprehensive fatigue score range of 0-25 points. This level indicates that the user is in good mental condition, with high parasympathetic nerve activity, high heart rate variability, stable breathing, and active behavior, requiring no intervention. The mild fatigue level corresponds to a comprehensive fatigue score range of 26-50 points. This level indicates that the user experiences slight fatigue, with a slight decrease in parasympathetic nerve activity, mild breathing irregularities, and slight restlessness; rest is recommended. The moderate fatigue level corresponds to a comprehensive fatigue score range of 51-75 points. This level indicates that the user experiences moderate fatigue, with relatively increased sympathetic nerve activity, significant breathing disturbances, and frequent posture changes; rest is required. The severe fatigue level corresponds to a fatigue score range of 76-100 points. This level indicates that the user is severely fatigued, has a serious imbalance in the autonomic nervous system, is at risk of sleep apnea, and has abnormal body movement patterns, requiring immediate intervention.

[0091] The fatigue level thresholds are set based on heart rate variability guidelines and recommendations from the European Society of Cardiology, referencing medical standards such as a decrease in RMSSD of more than 30% indicating significant parasympathetic dysfunction and an increase in the LF / HF ratio of more than 50% from the baseline indicating relatively hyperactive sympathetic activity.

[0092] The corrected overall fatigue score is matched with the score intervals in the mapping relationship. The matching process uses interval comparison, comparing the corrected overall fatigue score with the score intervals corresponding to each fatigue level in turn. The matching process proceeds in order of fatigue level from mild to severe or from severe to mild. For example, it first checks if the corrected overall fatigue score is less than or equal to 25 points; if so, it matches the "energetic" level. If not, it checks if it is less than or equal to 50 points; if so, it matches the "mild fatigue" level. If not, it checks if it is less than or equal to 75 points; if so, it matches the "moderate fatigue" level. If none of the above conditions are met, it matches the "severe fatigue" level. The matching process ensures that each score value can only match one fatigue level, avoiding ambiguity in level determination. The score interval matching also has a boundary buffer mechanism. When the score value is close to the interval boundary, for example, within 3 points above or below the boundary, it can be marked as a transitional state, indicating to the user that the fatigue state is on the edge of the level and that subsequent changes should be monitored. The rating range matching can also be combined with the judgment of time continuity. For example, the rating is confirmed only when the rating falls into the same level range more than three times in a row, so as to avoid misjudgment of the level caused by a single abnormal rating.

[0093] The user's current fatigue level is determined based on the matching results. Each matching result directly corresponds to a specific fatigue level, reflecting the user's current fatigue state. Fatigue levels are represented by a level number and a level name, such as Level 0 (energetic), Level 1 (mild fatigue), Level 2 (moderate fatigue), and Level 3 (severe fatigue). The determination of fatigue levels can also be smoothed by incorporating historical fatigue levels. For example, if the current fatigue level differs from the previous fatigue level by more than one level, level smoothing can be performed to limit the change in fatigue level to a range of adjacent levels, avoiding drastic jumps in fatigue level.

[0094] The system outputs fatigue assessment results including fatigue levels, which can be presented in various formats. Numerical output directly shows the overall fatigue score, such as 45 points. Level output displays the fatigue level number and name, such as Level 1 (mild fatigue). Textual descriptions provide a detailed description of the fatigue state, such as "You are currently in a state of mild fatigue; it is recommended to take a rest." Visual output uses colors, icons, or progress bars to intuitively display the fatigue level; for example, green indicates high energy, yellow indicates mild fatigue, orange indicates moderate fatigue, and red indicates severe fatigue. The fatigue assessment results can be displayed through a user interface, such as a mobile application, a smart home control panel, or a vehicle display screen. The results can also trigger corresponding intervention suggestions: rest is recommended for mild fatigue, rest is necessary for moderate fatigue, and immediate intervention is required for severe fatigue. Intervention suggestions can include rest reminders, environmental adjustment suggestions, and exercise recommendations. The results can also be integrated with smart home systems to automatically adjust indoor lighting brightness, temperature, humidity, music, and other environmental parameters based on the fatigue level, providing a comfortable resting environment for the user. For example, when moderate or severe fatigue is detected, the system automatically dims the lights, plays soothing music, and adjusts the temperature to a comfortable level. Fatigue assessment results can also be sent to the user's family members or health management personnel for remote monitoring of fatigue levels. These results can be stored locally or in the cloud, creating a user's fatigue history for subsequent health trend analysis and personalized baseline updates.

[0095] Figure 10 This is a schematic block diagram of a user fatigue state assessment device 600 provided in an embodiment of the present invention. Figure 10 As shown, corresponding to the above-described user fatigue assessment method, the present invention also provides a user fatigue assessment device 600. This user fatigue assessment device 600 includes a unit for performing the above-described user fatigue assessment method, and the device can be configured in terminals such as smart chairs, smart sofas, and smart mattresses.

[0096] Specifically, please refer to Figure 10 The user fatigue assessment device 600 includes: The signal acquisition and preprocessing unit 610 is used to acquire the user's physiological signals through a physiological signal sensor and preprocess the physiological signals to obtain preprocessed physiological signals. The feature extraction unit 620 is used to extract multi-dimensional physiological feature indicators in parallel based on the preprocessed physiological signal. The weight adjustment unit 630 is used to dynamically adjust the weights of each physiological characteristic indicator based on the signal quality assessment results and / or physiological state triggering conditions corresponding to each physiological characteristic indicator. The scoring calculation unit 640 is used to compare the real-time extracted physiological characteristic indicators of each dimension with the user's personalized physiological baseline to obtain the normalized score of each dimension, and to calculate the comprehensive fatigue score by combining the adjusted weights of the physiological characteristic indicators. The smoothing and suppression unit 650 is used to perform time-series smoothing and abnormal fluctuation suppression processing on the fatigue comprehensive score to obtain a corrected fatigue comprehensive score. The level determination unit 660 is used to determine the user's fatigue level and output the fatigue assessment result based on the mapping relationship between the corrected comprehensive fatigue score and the preset fatigue level threshold.

[0097] In one embodiment, the signal acquisition preprocessing unit 610 includes: The signal acquisition unit is used to acquire raw physiological signals by pointing a millimeter-wave radar sensor at the user's chest cavity. The raw physiological signals include heart rate signals, respiratory signals, and body movement signals. The filtering unit is used to perform bandpass filtering on the original physiological signal to remove DC components and high-frequency noise. The analog-to-digital conversion unit is used to perform analog-to-digital conversion on the filtered physiological signal to obtain a preprocessed digital physiological signal.

[0098] In one embodiment, the feature extraction unit 620 extracts heart rate variability feature indicators including: A heartbeat separation unit is used to separate heartbeat signal components from the preprocessed heart rate signal; The R-wave recognition unit is used to identify the position of the R-wave peak in the heartbeat signal through a threshold detection algorithm; The interval calculation unit is used to calculate the time interval between adjacent R wave peaks based on the R wave peak position, so as to obtain the RR interval sequence; The variability index calculation unit is used to calculate heart rate variability indexes based on the RR interval sequence, wherein the heart rate variability indexes include the RMSSD index and the LF / HF ratio index.

[0099] In one embodiment, the feature extraction unit 620 extracts respiratory rhythm feature indicators including: A respiratory cycle identification unit is used to identify the start and end points of the respiratory cycle from the preprocessed respiratory signal; The respiratory interval calculation unit is used to calculate the time interval between adjacent respiratory cycles to obtain a respiratory interval sequence. A respiratory regularity index calculation unit is used to calculate respiratory regularity indexes based on the respiratory interval sequence. The respiratory regularity indexes include the respiratory interval standard deviation index and the respiratory variability coefficient index.

[0100] In one embodiment, the feature extraction unit 620 extracts body motion feature indicators including: A body movement event detection unit is used to detect body movement events from the preprocessed body movement signal; The event merging unit is used to merge short-term physical events that are temporally adjacent to each other to obtain a complete sequence of physical events; A body movement pattern recognition unit is used to identify body movement patterns in the complete body movement event sequence, including restless body movement patterns, posture changing body movement patterns, head drooping body movement patterns, and trunk tilting body movement patterns. The body movement characteristic index calculation unit is used to calculate body movement characteristic indices based on the body movement events and body movement patterns. The body movement characteristic indices include body movement frequency index, body movement amplitude index, and body movement pattern scoring index.

[0101] In one embodiment, the weight adjustment unit 630 includes: The signal quality assessment unit is used to calculate the signal quality assessment value corresponding to each dimension of physiological characteristic indicators. The basic weight setting unit is used to set the basic weights of physiological characteristic indicators in each dimension. The total quality score calculation unit is used to calculate the total quality score based on the signal quality assessment values ​​and basic weights of the physiological characteristic indicators of each dimension. An adaptive weight calculation unit is used to normalize and adjust the basic weights of the physiological characteristic indicators of each dimension based on the total quality score to obtain adaptive weights. The weight boundary constraint unit is used to perform boundary constraint processing on the adaptive weights to ensure that the weights of each dimension are within the preset weight range.

[0102] In one embodiment, the weight adjustment unit 630 further includes: The trigger condition detection unit is used to detect physiological state trigger conditions based on preset physiological state trigger rules. The physiological state trigger conditions include heart rate variability abnormality trigger conditions, respiratory disorder trigger conditions, and body movement abnormality trigger conditions. The weight trigger adjustment unit is used to trigger and adjust the adaptive weight of the corresponding dimension when any of the physiological state trigger conditions are met.

[0103] In one embodiment, the scoring calculation unit 640 includes: The physiological baseline acquisition unit is used to acquire the user's personalized physiological baseline, which includes the baseline values ​​of physiological characteristic indicators of each dimension obtained by statistical analysis of the user's historical physiological data. The deviation calculation unit is used to compare the real-time extracted physiological feature indicators of each dimension with the corresponding baseline values ​​and calculate the degree of deviation of each physiological feature indicator from the baseline values. A normalized score calculation unit is used to calculate the normalized score for each dimension based on the degree of deviation. The comprehensive scoring calculation unit is used to perform a weighted summation of the normalized scores of each dimension and the adjusted weights to obtain a comprehensive fatigue score.

[0104] In one embodiment, the smoothing suppression unit 650 includes: The smoothing unit is used to smooth the fatigue comprehensive score of the continuous time series using the exponentially weighted moving average algorithm to obtain the smoothed fatigue comprehensive score. The difference calculation unit is used to calculate the difference between the current fatigue comprehensive score and the previous smoothed fatigue comprehensive score; A fluctuation threshold determination unit is used to determine whether the difference exceeds a preset fluctuation threshold. An outlier suppression unit is used to perform outlier suppression processing on the current fatigue comprehensive score when the difference exceeds a preset fluctuation threshold, thereby limiting the score change within the preset fluctuation threshold range and obtaining a corrected fatigue comprehensive score.

[0105] The aforementioned user fatigue assessment device 600 can be implemented as a computer program, which can, for example... Figure 11 It runs on the computer device shown.

[0106] Please see Figure 11 , Figure 11 This is a schematic block diagram of a computer device 500 provided in an embodiment of this application. The computer device 500 can be a terminal or a server. The terminal can be a smart home electronic device with communication functions, such as a smart chair, smart sofa, or smart mattress. The server can be a standalone server or a server cluster composed of multiple servers.

[0107] See Figure 11 The computer device 500 includes a processor 502, a memory, and a network interface 505 connected via a system bus 501. The memory may include a non-volatile storage medium 503 and internal memory 504.

[0108] The non-volatile storage medium 503 may store an operating system 5031 and a computer program 5032. The computer program 5032 includes program instructions that, when executed, cause the processor 502 to perform a user fatigue state assessment method.

[0109] The processor 502 provides computing and control capabilities to support the operation of the entire computer device 500.

[0110] The internal memory 504 provides an environment for the execution of the computer program 5032 in the non-volatile storage medium 503. When the computer program 5032 is executed by the processor 502, the processor 502 can execute a user fatigue state assessment method.

[0111] This network interface 505 is used for network communication with other devices. Those skilled in the art will understand that... Figure 11 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device 500 to which the present application is applied. The specific computer device 500 may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0112] The processor 502 is used to run a computer program 5032 stored in a memory to implement the steps of the above method.

[0113] It should be understood that in the embodiments of this application, the processor 502 may be a central processing unit (CPU), or it may be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor.

[0114] It will be understood by those skilled in the art that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program includes program instructions and can be stored in a storage medium, which is a computer-readable storage medium. The program instructions are executed by at least one processor in the computer system to implement the process steps of the embodiments of the above methods.

[0115] Therefore, the present invention also provides a storage medium. This storage medium can be a computer-readable storage medium. The storage medium stores a computer program, wherein the computer program includes program instructions. When executed by a processor, the program instructions cause the processor to perform the steps of the above-described method.

[0116] The storage medium can be any computer-readable storage medium capable of storing program code, such as a USB flash drive, portable hard drive, read-only memory (ROM), magnetic disk, or optical disk.

[0117] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0118] In the several embodiments provided by this invention, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative. For example, the division of each unit is only a logical functional division, and there may be other division methods in actual implementation. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed.

[0119] The steps in the method of this invention can be adjusted, merged, or reduced in order according to actual needs. The units in the device of this invention can be merged, divided, or reduced according to actual needs. Furthermore, the functional units in the various embodiments of this invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0120] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, a terminal, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention.

[0121] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in the present invention, and these modifications or substitutions should all be covered within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for assessing user fatigue, characterized in that, The method includes: The user's physiological signals are collected by a physiological signal sensor, and the physiological signals are preprocessed to obtain preprocessed physiological signals. Based on the preprocessed physiological signals, multi-dimensional physiological feature indicators are extracted in parallel. The weights of each physiological characteristic indicator are dynamically adjusted based on the signal quality assessment results and / or physiological state triggering conditions corresponding to each physiological characteristic indicator. The real-time extracted physiological characteristic indicators of each dimension are compared with the user's personalized physiological baseline to obtain the normalized score of each dimension, and the fatigue comprehensive score is calculated by combining the adjusted weights of the physiological characteristic indicators. The fatigue comprehensive score is subjected to time-series smoothing and abnormal fluctuation suppression processing to obtain the corrected fatigue comprehensive score. Based on the mapping relationship between the corrected fatigue comprehensive score and the preset fatigue level threshold, the user's fatigue level is determined and the fatigue assessment result is output.

2. The user fatigue state assessment method according to claim 1, characterized in that, The steps of acquiring the user's physiological signals through a physiological signal sensor and preprocessing the physiological signals to obtain preprocessed physiological signals include: Raw physiological signals are collected by pointing a millimeter-wave radar sensor at the user's chest cavity. These raw physiological signals include heart rate signals, respiratory signals, and body movement signals. The original physiological signal was subjected to bandpass filtering to remove DC components and high-frequency noise; The filtered physiological signal is converted from analog to digital to obtain the preprocessed digital physiological signal.

3. The user fatigue state assessment method according to claim 2, characterized in that, The multidimensional physiological characteristic indicators include heart rate variability (HRV) indicators, and the extraction steps for the HRV indicators include: Separate the heartbeat signal components from the preprocessed heart rate signal; The location of the R-wave peak in the heartbeat signal is identified using a threshold detection algorithm; Based on the R-wave peak position, the time interval between adjacent R-wave peaks is calculated to obtain the RR interval sequence; Heart rate variability indices are calculated based on the RR interval sequence, including the RMSSD index and the LF / HF ratio index.

4. The user fatigue state assessment method according to claim 2, characterized in that, The multidimensional physiological characteristic indicators also include respiratory regularity characteristic indicators, and the extraction steps of the respiratory regularity characteristic indicators include: Identify the start and end points of the respiratory cycle from the preprocessed respiratory signals; Calculate the time interval between adjacent respiratory cycles to obtain the respiratory interval sequence; Respiratory regularity indices are calculated based on the respiratory interval sequence, including the respiratory interval standard deviation index and the respiratory variability coefficient index.

5. The user fatigue state assessment method according to claim 2, characterized in that, The multidimensional physiological characteristic indicators also include body movement characteristic indicators, and the extraction steps of the body movement characteristic indicators include: Detect body movement events from the preprocessed body movement signals; Merging short-term body movement events with adjacent time sequences yields a complete sequence of body movement events; Identify body movement patterns in the complete sequence of body movement events, including restless body movement patterns, posture changing body movement patterns, head-down body movement patterns, and trunk-tilting body movement patterns. Based on the physical movement events and patterns, physical movement characteristic indicators are calculated, including physical movement frequency indicators, physical movement amplitude indicators, and physical movement pattern scoring indicators.

6. The user fatigue state assessment method according to claim 5, characterized in that, The step of dynamically adjusting the weights of each physiological characteristic indicator based on the signal quality assessment results and / or physiological state triggering conditions corresponding to each physiological characteristic indicator includes: Calculate the signal quality assessment value corresponding to each dimension of physiological characteristic index; Set the basic weights for each dimension of physiological characteristic indicators; The total quality score is calculated based on the signal quality assessment values ​​and basic weights of the physiological characteristic indicators of each dimension. Based on the total quality score, the basic weights of the physiological characteristic indicators of each dimension are normalized and adjusted to obtain adaptive weights; Boundary constraints are applied to the adaptive weights to ensure that the weights of each dimension are within the preset weight range.

7. The user fatigue state assessment method according to claim 6, characterized in that, The step of dynamically adjusting the weights of each physiological characteristic indicator based on the signal quality assessment results and / or physiological state triggering conditions corresponding to each physiological characteristic indicator further includes: Based on preset physiological state triggering rules, physiological state triggering conditions are detected, including heart rate variability abnormality triggering conditions, respiratory disorder triggering conditions, and body movement abnormality triggering conditions. When any of the physiological state trigger conditions is met, the adaptive weights of the corresponding dimensions are adjusted accordingly.

8. The user fatigue state assessment method according to claim 1, characterized in that, The steps of comparing the real-time extracted physiological characteristic indicators of each dimension with the user's personalized physiological baseline to obtain a normalized score for each dimension, and calculating a comprehensive fatigue score by combining the adjusted weights of the physiological characteristic indicators, include: Obtain the user's personalized physiological baseline, which includes the baseline values ​​of various physiological characteristic indicators obtained from the statistical analysis of the user's historical physiological data. The real-time extracted physiological characteristic indicators of each dimension are compared with the corresponding baseline values, and the degree of deviation of each physiological characteristic indicator from the baseline value is calculated. Calculate the normalized score for each dimension based on the degree of deviation; The fatigue score is obtained by weighting and summing the normalized scores of each dimension with the adjusted weights.

9. The user fatigue state assessment method according to claim 1, characterized in that, The steps of performing time-series smoothing and abnormal fluctuation suppression processing on the fatigue comprehensive score to obtain the corrected fatigue comprehensive score include: The fatigue score of a continuous time series is smoothed by an exponentially weighted moving average algorithm to obtain a smoothed fatigue score. Calculate the difference between the current fatigue score and the previous smoothed fatigue score; Determine whether the difference exceeds a preset fluctuation threshold; When the difference exceeds the preset fluctuation threshold, outlier suppression processing is performed on the current fatigue comprehensive score to limit the score change within the preset fluctuation threshold range, thus obtaining the corrected fatigue comprehensive score.

10. A user fatigue assessment device, characterized in that, Used to perform the user fatigue state assessment method as described in any one of claims 1 to 9.

11. A computer device, characterized in that, The computer device includes a memory and a processor connected to the memory; the memory is used to store a computer program; the processor is used to run the computer program stored in the memory to perform the steps of the method as described in any one of claims 1 to 9.

12. A computer-readable storage medium, characterized in that, The storage medium stores a computer program, which includes program instructions that, when executed by a processor, can implement the steps of the method as described in any one of claims 1 to 9.

Images