A method for quantifying psychological stress relief effect based on pulse wave and drumming behavior

Abstract

The present application relates to the technical field of psychological stress relief effect evaluation, and particularly relates to a psychological stress relief effect quantification method based on pulse wave and drumming behavior, comprising: collecting PPG signals of a subject and extracting physiological features, and constructing a training stress prediction model; guiding the subject to complete drumming training, and generating a drumming action time sequence; based on the drumming action time sequence and the synchronously collected piezoelectric sensor signals, calculating and forming a behavior feature vector of the subject; outputting a predicted stress value change amount before and after drumming through the stress prediction model, quantifying the contribution of different drumming behavior features to stress relief; and based on the contribution of different drumming behavior features to stress relief, quantitatively evaluating the effect of drumming behavior on psychological stress relief. The present application realizes quantitative evaluation of the stress relief effect of different drumming behaviors by fusing physiological features and behavior features, and can realize objective and accurate quantitative evaluation of the psychological stress relief effect of drumming behavior.

Description

Technical Field

[0001] This invention relates to the field of evaluation technology for psychological stress relief effects, specifically to a quantitative method for psychological stress relief effects based on pulse waves and drumming behavior. Background Technology

[0002] With the accelerating pace of modern society, individuals face significantly increased pressure from work, life, and other aspects. While moderate stress can motivate individuals and improve work efficiency, prolonged or excessive stress can seriously threaten physical and mental health. It can easily trigger anxiety, depression, and other psychological discomforts, and may also induce chronic diseases such as diabetes and cardiovascular disease. Therefore, it has become one of the important public health issues affecting human health in modern society. Thus, developing scientific and effective methods for relieving psychological stress and establishing a precise effectiveness evaluation system are of significant practical importance.

[0003] Music therapy is a widely used stress relief intervention, which can be divided into passive and active methods based on the participant's mode of engagement. Passive music therapy focuses on "listening and receiving," with participants passively receiving musical stimuli. Active music therapy, on the other hand, emphasizes the direct participation of participants in the music production process, such as playing instruments or improvising, resulting in stronger interactivity and immersion, and more significant stress-relieving effects. Drumming, as a typical example of active music therapy, has advantages such as low entry barriers, wide audience reach, and direct emotional release, making it a highly promising way to relieve psychological stress. However, the lack of systematic and objective quantitative evaluation methods for the stress-relieving effects of drumming therapy currently hinders its standardized application and promotion.

[0004] Existing methods for assessing the effectiveness of stress relief are mainly divided into two categories: subjective assessment and objective measurement. Subjective assessment relies on participants' self-reports, such as questionnaires and mood rating scales, which are subject to significant subjective bias, individual differences, and reporting bias, failing to comprehensively and accurately reflect the true effect of stress relief. Objective measurement is achieved by detecting physiological signals or indicators, commonly including cortisol, α-salivary amylase, electroencephalogram (EEG) signals, and electrocardiogram (ECG) signals. However, these commonly used objective indicators generally have significant drawbacks: the collection process for biochemical indicators such as cortisol and α-salivary amylase is highly invasive and lacks real-time accuracy, making dynamic assessment impossible; the equipment for collecting EEG and ECG signals is bulky, complex to operate, expensive, and has stringent requirements for the collection environment, making it difficult to apply to real-time monitoring in everyday scenarios. These shortcomings make it difficult for existing objective assessment methods to meet the quantitative needs of everyday stress relief methods such as drumming.

[0005] Pulse wave signals, as important physiological signals reflecting the state of the human cardiovascular system and autonomic nervous system, have unique advantages in acquisition: the corresponding sensors are easy to integrate and wear, inexpensive, and the acquisition process is non-invasive and convenient. Furthermore, they can reflect real-time fluctuations in autonomic nervous function caused by stress changes, effectively compensating for the shortcomings of existing objective assessment indicators. Based on this, a practical and scalable objective quantitative method for assessing the effects of psychological stress relief can be established by combining the active stress-relieving characteristics of drumming behavior with the convenient detection advantages of pulse wave signals. Summary of the Invention

[0006] To address the technical problems existing in the prior art, this invention provides a method for quantifying the psychological stress relief effect based on pulse wave and drumming behavior. Through physiological signal acquisition, drumming behavior feature extraction, and stress prediction model construction, based on the contribution of different drumming behavior features to stress relief, and through comprehensive analysis based on drumming behavior features and changes in key physiological features, an objective and accurate quantitative assessment of the psychological stress relief effect is achieved.

[0007] The objective of this invention can be achieved by adopting the following technical solutions:

[0008] A method for quantifying the stress-relieving effects based on pulse waves and drumming behavior includes the following steps:

[0009] S1. Collect PPG signals from subjects in relaxed and stressed states, preprocess the collected PPG signals and extract physiological features, select a subset of key physiological features and train a stress prediction model.

[0010] S2. Guide the subject to complete drumming training through a preset drumming training program, obtain the drumming time, drumming force and drumming position of each drumming event, and generate a drumming action time sequence.

[0011] S3. Based on the time sequence of drumming action and the synchronously acquired piezoelectric sensor signal, calculate the drumming rhythm characteristics, drumming force characteristics, and drumming position characteristics to form the behavioral feature vector of the subject.

[0012] S4. Collect PPG signal features of subjects before and after drumming training, extract key physiological feature values ​​and standardize them, and output the predicted pressure value changes before and after drumming through the pressure prediction model.

[0013] S5. Based on the predicted change in stress value and the behavioral feature vector of the subjects, quantify the contribution of different drumming behavior characteristics to stress relief.

[0014] S6. Based on the contribution of different drumming behavior characteristics to stress relief, a comprehensive analysis is conducted based on drumming behavior characteristics and changes in key physiological characteristics to quantitatively evaluate the effect of drumming behavior on psychological stress relief.

[0015] Specifically, step S1 includes:

[0016] S11. Collect PPG signals of subjects in a stable relaxation state as baseline data, and collect PPG signals of subjects during and after the standardized stress-induced task as stress state data.

[0017] S12. Preprocess the acquired PPG signal and extract physiological features for pressure identification from the preprocessed PPG signal.

[0018] S13. Perform paired-sample significance testing on various physiological characteristics under relaxed and stressed states, and screen out physiological characteristics that show statistically significant differences between relaxed and stressed states to form a candidate feature set for stress sensitivity.

[0019] S14. Based on the stress-sensitive candidate feature set, recursive feature elimination is used to filter and obtain multiple key physiological features. Based on the multiple key physiological features, a nonlinear regression model with multiple feature inputs is trained, and the trained nonlinear regression model is used as a stress prediction model.

[0020] Specifically, step S2 includes:

[0021] S21. Guide the subject to complete drumming training through a preset drumming training program. The preset drumming training program includes a drumming training program in follow-up imitation mode and a drumming training program in free mode.

[0022] S22. Distributed sensor layout and collaborative data acquisition: Piezoelectric sensors are fixed at the center and edge of the drum surface. The voltage signals of all piezoelectric sensors are collected synchronously. The force value of drumming is calculated based on the voltage signals of the piezoelectric sensors. The position tag of drumming is generated according to the area corresponding to the sensor. The timestamp of each drumming event is bound to the position tag and arranged in chronological order to form a drumming action time sequence.

[0023] Specifically, step S3 includes:

[0024] S31. Calculate rhythmic features: Based on the time sequence of the drumming action, calculate the temporal features reflecting the rhythmic pattern.

[0025] S32. Calculate the striking force characteristics. Use the energy value of the signal within a single striking window as the force value of this drumbeat. Calculate the mean, standard deviation, and accuracy of all drumbeat force values.

[0026] S33. Calculate the characteristics of the striking force, calculate the proportion of striking positions and the frequency of striking position switching, calculate the proportion of the number of times the drumhead is struck and the number of times the drum rim is struck to the total number of times, and calculate the ratio of the number of times the striking positions are different between two adjacent striking positions to the total number of times.

[0027] Specifically, step S4 includes:

[0028] S41. Collect PPG signals from the subjects before and after drumming training.

[0029] S42. Extract key physiological features from the PPG signals of the subjects before and after drumming training, and standardize each key physiological feature.

[0030] S43. Input the standardized key physiological characteristics before and after drumming training into the stress prediction model to calculate the predicted change in psychological stress value.

[0031] Specifically, step S5 includes:

[0032] S501. The behavioral feature vector of the test subjects in step S3 and the change in the predicted stress value output in step S4 are grouped according to task mode and difficulty to obtain multiple analysis groups.

[0033] S502. Correlation analysis: Calculate the correlation coefficient between the behavioral characteristics within each group and the predicted changes in psychological stress values ​​before and after training, construct a correlation coefficient matrix for each group, and determine the direction and intensity of the influence of different drumming behavioral characteristics on stress relief effect by the positive and negative values ​​and significance of the coefficients.

[0034] Specifically, step S6 includes:

[0035] S601. Feature matrix construction: Combine all extracted behavioral features with key physiological features, and use the predicted change in psychological stress value as an indicator of relief effect to construct a comprehensive feature matrix for clustering.

[0036] S602. After standardizing the comprehensive feature matrix, reduce the data dimensionality, remove redundancy, and alleviate multicollinearity through feature transformation and compression.

[0037] S603. Perform unsupervised grouping processing on the reduced-dimensional comprehensive feature matrix. By statistically analyzing the central tendency or representative characteristics of drumming behavior features in each behavior pattern category, determine the drumming behavior pattern combination corresponding to different stress relief effect levels, and obtain the stress relief effect corresponding to different drumming behavior pattern combinations.

[0038] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0039] (1) This invention provides a method for quantifying the psychological stress relief effect based on pulse wave and drumming behavior. A multi-dimensional physiological feature system is constructed based on pulse wave signals, comprehensively extracting heart rate variability features and pulse wave morphological features. The contribution of physiological features is evaluated and screened using a prediction model based on multi-feature inputs, obtaining a subset of key features highly correlated with psychological stress state. By reducing the introduction of redundant features, the system's robustness to noise and individual differences is improved while ensuring the accuracy of stress assessment, and the overall computational complexity is reduced, thereby enhancing the stability and real-time application efficiency of stress assessment. This transforms the traditional stress relief assessment method, which relies on subjective scales, into an objective quantitative assessment method based on changes in physiological signals, achieving objective, continuous, and repeatable assessment, overcoming the problems of poor consistency and insufficient comparability of existing subjective assessment methods.

[0040] (2) This invention integrates the behavioral characteristics and physiological state changes of rhythm, intensity and striking position during the drumming process to classify and analyze the effects of different drumming behavior patterns. It establishes a quantitative mapping relationship between drumming behavior characteristics and psychological stress relief effects, and can identify combinations of drumming behavior patterns with different stress relief effects. This provides an objective basis for the design of personalized and differentiated psychological stress intervention programs, and enhances the adaptability and guidance value of the system in practical application scenarios. Attached Figure Description

[0041] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the structures shown in these drawings without creative effort.

[0042] Figure 1 This is a flowchart of the overall process of a method for quantifying the psychological stress relief effect based on pulse waves and drumming behavior, as disclosed in Embodiment 1 of the present invention.

[0043] Figure 2 This is a diagram showing the effect of PPG signal preprocessing in Embodiment 1 of the present invention;

[0044] Figure 3 This is a diagram showing the localization effect of various pulse wave morphological feature points in Embodiment 1 of the present invention;

[0045] Figure 4 This is a schematic diagram illustrating the significance test of characteristic changes before and after pressure induction in Embodiment 1 of the present invention;

[0046] Figure 5This is a schematic diagram showing the importance of key physiological features selected by the output of the nonlinear regression model based on multiple feature inputs in Embodiment 1 of the present invention.

[0047] Figure 6 This is a schematic diagram illustrating the effect of the nonlinear regression model based on multiple feature inputs on the test set in Embodiment 1 of the present invention;

[0048] Figure 7 This is a schematic diagram of the drumming training system in Embodiment 1 of the present invention;

[0049] Figure 8 This is a diagram illustrating the percussion instrument in Embodiment 1 of the present invention;

[0050] Figure 9 This is a schematic diagram of signal extraction for the drumming task in Embodiment 1 of the present invention;

[0051] Figure 10 This is a diagram showing the matching results of the drumming signal in Embodiment 1 of the present invention;

[0052] Figure 11 This is a box plot comparing pressure values ​​before and after drumming training in Embodiment 1 of the present invention;

[0053] Figure 12 This is a heatmap showing the correlation between behavioral characteristics and changes in psychological stress in Embodiment 1 of the present invention. Detailed Implementation

[0054] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments, and the implementation of the present invention is not limited thereto. 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.

[0055] Example 1:

[0056] This embodiment provides a method for quantifying the psychological stress relief effect based on pulse wave and drumming behavior. By collecting pulse wave signals of subjects before and after stress induction and processing them, heart rate variability (HRV) features and morphological features related to the pulse wave signals are obtained. Highly important features are selected through recursive feature elimination (RFE). A stress prediction model is constructed and trained. The trained stress prediction model is applied to the drumming task. The stress prediction model obtains the intervention effect of the drumming task based on the changes in physiological signals before and after the intervention. At the same time, behavioral features in the drumming task are extracted, and the importance of behavioral features in the intervention is evaluated.

[0057] like Figure 1The diagram shows the overall workflow of a method for quantifying the psychological stress relief effect based on pulse waves and drumming behavior. The method, as described in this invention, includes the following steps:

[0058] S1. Collect PPG signals from subjects in relaxed and stressed states, preprocess the collected PPG signals and extract physiological features, select a subset of key physiological features and train a stress prediction model.

[0059] S11. Data Acquisition: Collect PPG signals of subjects in a stable relaxed state as baseline data, and collect PPG signals of subjects during and after the standardized stress-induced task as stress state data.

[0060] In a standardized experimental environment, PPG signals were collected from subjects in both a resting, relaxed state and a stress state induced by a standardized stress task. PPG signals, or photoplethysmography, are a biosignal that uses optical principles to detect changes in blood volume. Simply put, it reflects heart rate and blood flow by measuring changes in skin light absorption. Specifically, after entering the experimental area, subjects sat quietly for 5 minutes to eliminate environmental stress. Then, they placed their index finger on the photodetector of a wearable pulse wave detection device, maintaining a static posture to collect stable PPG signals as baseline data for the resting, relaxed state. Subjects were then guided to complete a standardized stress-inducing task, during which PPG signals were collected simultaneously as stress state data. Performance on the stress-inducing task was also recorded.

[0061] S12. Signal preprocessing and feature extraction: The acquired PPG signal is preprocessed, and physiological features for stress identification are extracted from the preprocessed PPG signal.

[0062] S121. The acquired PPG signal is preprocessed, including noise reduction, filtering, and baseline drift correction, to obtain a high-quality pulse wave waveform. Noise filtering is performed because the original PPG signal contains interference such as power frequency noise and electromyographic noise. This invention uses a Butterworth bandpass filter to filter the signal, ensuring the integrity of the pulse wave characteristics.

[0063] Baseline drift correction addresses the issue of low-frequency, slow signal shift caused by respiratory movements and subtle changes in body posture. This invention employs cubic spline interpolation for baseline drift correction. First, the filtered PPG signal is smoothed using a sliding window to fit a baseline drift trend curve. Then, the difference between the original filtered signal and the drift trend curve yields the purified PPG signal after baseline drift elimination. This method accurately matches the low-frequency, slow drift trend, preserving key morphological features such as pulse wave peaks, troughs, and rising slope to the greatest extent possible, ensuring the accuracy of subsequent cycle identification and feature extraction. Cycle segmentation and filtering utilize a peak detection algorithm to extract peak values ​​from the first-order differential signal, identifying the point of fastest rise in the PPG signal. For each valid peak, the adjacent zero-crossing point to its left (the inflection point where the first-order differential signal changes from negative to positive) is traced back. The signal index corresponding to this zero-crossing point is the starting index of a cycle. Then, using all detected cycles, a median waveform is calculated as a "standard template." The Pearson correlation coefficient between each cycle and the standard template is used to eliminate abnormal cycles whose correlation is lower than the average correlation coefficient - 2 standard deviations or higher than the average correlation coefficient + 2 standard deviations. For example... Figure 2 As shown, this demonstrates the effect of PPG signal preprocessing.

[0064] S122. Extract the set of physiological features for stress identification from the preprocessed PPG signal. The physiological features for stress identification include the time domain features, frequency domain features and pulse wave morphological features (such as peaks, troughs and area) of heart rate variability (HRV).

[0065] Extracting temporal features of heart rate variability (HRV), including but not limited to: mean normal heartbeat interval ( The formulas for calculating each time domain feature are as follows: standard deviation of normal heartbeat interval (SDNN), root mean square of the difference between consecutive normal heartbeat intervals (RMSSD), number of heartbeats with a difference between consecutive normal heartbeat intervals greater than 50ms (NN50) and the proportion of NN50 to the total number of differences (PNN50).

[0066] Mean of normal heartbeat interval :

[0067] (1)

[0068] Normal heart rate interval standard deviation (SDNN):

[0069] (2)

[0070] The root mean square (RMSSD) of the difference between consecutive normal heartbeat intervals:

[0071] (3)

[0072] The number of heartbeats with a difference of more than 50 ms between adjacent normal heartbeat intervals (NN50) among all heartbeat intervals:

[0073] (4)

[0074] The proportion of NN50 to the total number of differences:

[0075] (5)

[0076] Where, N For the i-th effective heartbeat interval, This is an indicator function; the condition in parentheses is 1 if true, and 0 otherwise. This represents the total number of effective heartbeat intervals.

[0077] The frequency domain features of heart rate variability (HRV) are extracted and calculated using Fast Fourier Transform (FFT), including but not limited to: low-frequency power (LF, 0.04-0.15Hz, reflecting sympathetic activity), high-frequency power (HF, 0.15-0.4Hz, reflecting parasympathetic activity), the LF / HF ratio (reflecting the sympathetic-parasympathetic balance), normalized low-frequency power LF_NU, and normalized high-frequency power HF_NU. The formula for calculating frequency band power is as follows:

[0078] (6)

[0079] Where P is the target frequency band power. , The upper and lower boundaries of the frequency band are defined by PSD(f), which is the power spectral density function.

[0080] Morphological features of the pulse wave were extracted, and cycle-by-cycle feature point identification was performed on the preprocessed PPG signal. The extracted feature points included the main peak of the pulse wave, the dicrota peak of the pulse wave, the descent mid-wave, and the start and end points of the pulse wave cycle. The localization effect of each pulse wave morphological feature point is shown in the figure. Figure 3 As shown. Based on the feature points, the morphological characteristic values ​​of each cycle are calculated, including but not limited to: the amplitude H of the main peak of the pulse wave. A Pulse wave diabetic peak value H B Enhancement index (RI, the ratio of dicrotic wave amplitude to main wave amplitude), and peak rise time (T) OA The time from the start of the cycle to the peak of the main peak, and the time of descent of the main peak (T). AO The time from the peak to the end of the cycle, and the area of ​​the rising branch (R). A1 The area of ​​the region enclosed by the starting point of the cycle and the peak), and the area of ​​the descending branch (R). A2The area enclosed by the main peak and the end point of the cycle, and the ratio of the area of ​​the rising branch to the area of ​​the falling branch, R. A12 Reflection time (T) AB The distance from the peak of the main wave to the peak of the diphtheria wave, and the reflection time ratio (R) AB The ratio of the distance from the peak point of the main wave to the peak point of the diphtheria wave to the pulse interval, etc., are calculated. At the same time, the mean, standard deviation and root mean square of the morphological features within the complete length of the PPG signal are calculated. The calculation formula is similar to formula (1)-(3).

[0081] Define HRV eigenvalues ​​as a set Morphological eigenvalues ​​are represented as a set Then the set of physiological characteristic values ​​F is represented as:

[0082] (7)

[0083] S13. Perform paired sample significance tests on various physiological characteristics under relaxed and stressed states, screen out physiological characteristics that show statistically significant differences between relaxed and stressed states, and construct a set of candidate features that are stress-sensitive, so as to reduce the computational complexity of subsequent recursive feature elimination.

[0084] Specifically, by performing paired-samples t-tests on physiological characteristics under relaxed and stressed states, features showing statistically significant differences (p < 0.05) between the two states were selected to form the stress-sensitive candidate feature set F_candidate. Figure 4 The diagram illustrates the significance test of characteristic changes before and after pressure induction. The physiological characteristics in F_candidate are plotted on the ordinate, and the p-value of the paired-samples test is plotted on the abscissa. A significance threshold of p=0.05 is also marked. The diagram shows... as well as The distribution of P-values ​​before and after stress induction shows that the P-values ​​of many features such as PRV_RMSSD, PRV_SDNN, and LF_HF are much less than 0.05, indicating that these features have highly significant statistical differences under relaxed and stressed states. On the other hand, the P-values ​​of features such as morph_RI_rms and LF are greater than 0.05, showing no statistically significant differences. This figure provides direct statistical basis for screening candidate feature sets that are sensitive to stress.

[0085] S14. Initial screening of key features and construction of the stress prediction model: Based on a set of stress-sensitive candidate features, recursive feature elimination is used to screen and obtain multiple key physiological features. A multi-feature input nonlinear regression model is trained based on these key physiological features to obtain the trained nonlinear regression model, which serves as the stress prediction model. The key feature set is then used as input to predict psychological stress using the stress prediction model.

[0086] Specifically, recursive feature elimination is performed using a stress-sensitive candidate feature set to select several key physiological features with the highest discriminative power for stress states. Using these key physiological features as input, a nonlinear regression model based on multiple feature inputs is trained. Five-fold cross-validation is used to evaluate the model performance, and the average prediction accuracy (R²) and mean squared error (MSE) are recorded to ensure that the model has stable generalization ability. The trained nonlinear regression model serves as the stress prediction model.

[0087] like Figure 5 The diagram shows the importance of key physiological features selected from the output of a nonlinear regression model based on multiple feature inputs. The horizontal axis represents key physiological features of PPG, and the vertical axis represents importance. It visually demonstrates the importance ranking and corresponding importance values ​​of the 15 key physiological features with the highest discriminative power for stress states after recursive feature elimination. Normalized high-frequency power HF_NU, high-frequency power HF, low-frequency power to high-frequency power ratio LF_HF, normalized low-frequency power LF_NU, root mean square of heart interval PRV_RMSSD, standard deviation of heart interval PRV_SDNN, mean of heart interval PRV_AVG. : standard deviation of reflection time morph_T AB _std, reflection time ratio to standard deviation morph_R AB _std, standard deviation of the ratio of rising branch area to falling branch area morph_R A12 _std, root mean square of periodic time morph_T_rms, standard deviation of rising branch time morph_T OA _std, descent branch time standard deviation morph_T AO The values ​​for _std, morph_RI_std (standard deviation of enhancement index), and morph_RI_rms (root mean square of enhancement index) are used. Higher values ​​indicate a greater contribution of these features to the prediction of psychological stress. Key physiological features are the core input features for building stress prediction models.

[0088] like Figure 6 As shown, this diagram illustrates the performance of the nonlinear regression model based on multiple feature inputs on the test set. The solid line represents the true value, and the dashed line represents the predicted value. The small deviation between the predicted and true values ​​indicates that the stress prediction model has extremely high prediction accuracy and good generalization ability on the test set, and can accurately predict the psychological stress state of the subjects based on physiological characteristics.

[0089] S2. Guide the subject to complete drumming training through a preset drumming training program, obtain the drumming time, drumming force and drumming position of each drumming event, and generate a drumming action time sequence.

[0090] S21. Guide the subject to complete drumming training through a preset drumming training program, which includes a follow-and-imitate mode drumming training program and a free mode drumming training program.

[0091] In this embodiment, the pre-follow-the-imitation drumming training program provides a standard rhythm sequence and guides the subject to drum by moving light dots. The drumbeat sequence is displayed on the screen as an animated series of colored dots, with different colored dots indicating different striking positions. Figure 7 As shown in the diagram, the drumming training system uses a red dot 72 to indicate the drumhead and a blue dot 71 to indicate the drum rim. The size of the dots indicates the intensity of the accent (large dot = accent, small dot = soft). A fixed vertical line on the left side of the display screen serves as the "drumming trigger line." When the drumbeat sequence moves to the trigger line in sync with the music rhythm, the subject is guided to strike the drum synchronously, achieving rhythmic imitation. The free creation mode of the drumming training program plays background music with a fixed rhythm. The display screen only shows the area markings of the drumhead and drum rim, without providing specific drumbeat sequence guidance. Subjects are encouraged to choose their striking position, intensity, and rhythm freely based on their own feelings, creating their own drumming routines.

[0092] S22. Distributed sensor layout and collaborative data acquisition: Piezoelectric sensors are fixed at the center and edge of the drum surface. The voltage signals of all piezoelectric sensors are collected synchronously. The force value of drumming is calculated based on the voltage signals of the piezoelectric sensors. The position tag of drumming is generated according to the area corresponding to the sensor. The timestamp of each drumming event is bound to the position tag and arranged in chronological order to form a drumming action time sequence.

[0093] Specifically, piezoelectric sensors are arranged in the center and rim areas of the drumhead to sense vibrations. The sensor outputs are connected to a microcontroller via a signal conditioning module for signal preprocessing and analog-to-digital conversion. The subject holds wooden drumsticks with both hands and strikes the drum according to the animation guidance on the display screen or according to their own creative needs. The microcontroller simultaneously collects voltage signals from multiple sensors and transmits them in real-time to a computer terminal for storage via a serial port. Figure 8 The image shows a display of a percussion instrument.

[0094] Specifically, bandpass filtering is applied to all sensor signals to remove environmental noise and high-frequency interference. The potential peak drumming moment t0 is detected from the piezoelectric signal using a thresholding method. Using t0 as a reference, a fixed time window is selected, and the signal within this window is extracted. The signal energy value within this window is calculated as the force value A of this drumming event. For a single drumming event, the amplitude percentage of the sensor signal peak is calculated; the region corresponding to the sensor with the largest amplitude percentage is taken as the location label for this strike.

[0095] The timestamp of each drumming event is bound to a location tag and arranged in chronological order to form a two-dimensional "time-location" drumming action time sequence:

[0096] (8)

[0097] in Let i be the timestamp of the i-th drumbeat. Let i be the position label of the i-th drumbeat. Let the force be the i-th time. This represents the total number of taps during the training process.

[0098] like Figure 9 As shown in the diagram, the signal extraction diagram of the drumming task shows the data on the change of the amplitude of the original vibration signal collected by the piezoelectric sensors at the drum edge and drum core over time. It can be seen that the signal response peaks of the sensors at different time points clearly reflect the striking actions of the test subject at different times and positions, providing reliable original signal data support for the subsequent extraction of drumming time, force, and position characteristics.

[0099] S3. Based on the time sequence of drumming action and the synchronously acquired piezoelectric sensor signal, calculate the drumming rhythm characteristics, drumming force characteristics, and drumming position characteristics to form a behavioral feature vector.

[0100] like Figure 10 As shown in the diagram, the drumming signal matching effect of the drumming task is displayed, showing the sensor amplitude changes of the drum center and drum edge over time. The detection threshold and effective window boundary are marked, and the drum center and drum edge strike marks detected by the algorithm are presented simultaneously. It can be seen that the actual strike signal peaks all exceed the threshold and fall within the effective window. The marks correspond precisely to the signal peaks, indicating that the algorithm can accurately identify the time and position of the drumming, achieving accurate detection and matching of the drumming action.

[0101] S31. Calculate the rhythmic characteristics of the drumming action. Based on the time sequence of the drumming action, calculate the temporal characteristics that reflect the rhythmic characteristics, including the time interval between adjacent drumming actions, rhythmic complexity, and rhythmic accuracy.

[0102] Rhythm interval sequence, calculating the time interval between adjacent taps, rhythm interval sequence I:

[0103] (9)

[0104] Rhythmic complexity: Calculate the information entropy of the interval sequence to quantify the degree of rhythmic variation and unpredictability. Rhythmic complexity H:

[0105] (10)

[0106] In Formula 10, p(Iᵢ) represents the probability of the interval value Iᵢ appearing in the sequence. This represents the number of different interval values ​​in the sequence.

[0107] Rhythmic accuracy (for drumming training programs in imitation mode): The time interval sequence of the test subject is aligned and compared with the standard rhythm interval sequence preset by the program. If the test subject's drumbeat exists within the standard drumbeat tolerance window, the test subject's drumbeat is correct; otherwise, it is incorrect. The proportion of correct drumbeats by the test subject is calculated as the test subject's snare drum sight-reading time accuracy.

[0108] rhythm accuracy The subject's drumming time sequence was dynamically time-normalized and aligned with a standard rhythm sequence. If the subject was at the standard drumbeat... tolerance time window If a drumbeat is struck, the drumbeat is judged to be "rhythmic correct," and the rhythm accuracy rate is [not specified]. :

[0109] (11)

[0110] in The number of drumbeats indicating the correct rhythm. Indicates the standard number of drum beats. This represents the fault tolerance time parameter.

[0111] S32. Calculate the striking force characteristics. Use the energy value of the signal within a single striking window as the force value of this drumbeat. Calculate the mean, standard deviation, and accuracy of all drumbeat force values.

[0112] Calculate all drumming force values mean and standard deviation This reflects the overall strength level and fluctuations.

[0113] (12)

[0114] (13)

[0115] Accuracy of calculating drumming force: The rules for determining the accuracy of force are defined as follows:

[0116] (14)

[0117] in , Indicates the force of adjacent strikes. This represents the error tolerance threshold. It calculates the percentage of drumbeats with correct force, which is used as the accuracy of the drumbeat force. As shown in (15):

[0118] (15)

[0119] in The number of drumbeats indicating the correct dynamic range. Indicates the standard number of drumbeats.

[0120] S33. Calculate the characteristics of the striking position, calculate the proportion of striking positions and the frequency of striking position switching, calculate the proportion of the number of times the drumhead is struck and the number of times the drum edge is struck to the total number of striking positions, and calculate the ratio of the number of times the striking positions are different between two adjacent striking positions to the total number of striking positions.

[0121] Percentage of striking position: Percentage of drumhead Ratio of drum edge As shown below:

[0122] (16)

[0123] (17)

[0124] Tap position switching frequency : Reflects the frequency of changes in the tapping position of the test subject, as shown in formula (18):

[0125] (18)

[0126] in This indicates the number of times the two adjacent taps differ in location. This represents the total number of taps during the training process.

[0127] S4. Collect PPG signal features of the subjects before and after drumming training, extract the feature values ​​of key physiological features and standardize them, and use the pressure prediction model to calculate the change in predicted pressure value before and after drumming.

[0128] S41. Collect PPG signals from the subjects before and after drumming training.

[0129] Photoplethysmography (PPG) signals were collected from the subjects before drumming training. After the subjects completed the drumming training, they were immediately guided to return to a seated posture. After the training, stable PPG signals were collected as post-training physiological data.

[0130] S42. Extract key physiological characteristics from PPG signals collected from the subjects before and after drumming training, standardize each key physiological characteristic, and eliminate individual differences.

[0131] The preprocessing procedure and feature extraction algorithm for the trained PPG signal are exactly the same as those used in S102 and S103. This yields the trained feature vector. Where k is the number of key features. Let i be the i-th feature.

[0132] Each key physiological feature in the post-training feature set is standardized to eliminate individual differences.

[0133] For each key physiological characteristic Z-score standardization is performed to obtain the standardized vectors. :

[0134] ; (19)

[0135] in and These are the key physiological characteristics The arithmetic mean and standard deviation of the sample set used.

[0136] S43. Calculate the change in psychological stress value. Input the standardized key physiological characteristics before and after drumming training into the stress prediction model to calculate the predicted change in psychological stress value.

[0137] The PPG features before and after training were standardized and then input into the stress prediction model constructed by S105 to obtain two predicted stress values. and Calculate the change in psychological stress level: . It indicates a stress-relieving effect.

[0138] like Figure 11As shown in the box plot, stress levels before and after drumming training are compared. The horizontal axis represents the pre-training and post-training groups, and the vertical axis represents the predicted psychological stress levels. The box plot displays the distribution characteristics of the median, quartiles, and extreme values ​​of stress levels in the two groups. It is evident that the median and overall distribution level of stress levels after training are significantly lower than before training, directly demonstrating the stress-relieving effect of drumming training on the subjects and verifying the stress-reducing effect of drumming behavior.

[0139] S5. Based on the predicted changes in stress levels and the behavioral feature vectors of the subjects, quantify the contribution of different drumming behaviors to stress relief. Specific steps include:

[0140] S51. Data grouping: The behavioral feature vectors of the subjects in step S3 and the predicted stress value changes output in step S4 are grouped according to task mode and difficulty to obtain multiple analysis groups.

[0141] Specifically, the experimental data were divided into four analysis groups: low-difficulty follower group, medium-difficulty follower group, high-difficulty follower group, and free group.

[0142] S52. Correlation Analysis: Calculate the Spearman correlation coefficient between the behavioral characteristics and the predicted stress value changes (stress relief effect) for each analysis group. To present the analysis results to the system, construct a correlation coefficient matrix: Using all behavioral characteristics and predicted stress value changes as row and column indicators, calculate and fill in their pairwise correlation coefficients. Finally, present the results as a matrix table to ensure efficient result integration and interpretation. The Spearman correlation coefficient ranges from -1 to 1. The closer the absolute value of the correlation coefficient is to 1, the stronger the linear correlation. Positive values ​​indicate a positive correlation, and negative values ​​indicate a negative correlation. Spearman rank correlation coefficient. The calculation formula is as follows:

[0143] (20)

[0144] ;(twenty one)

[0145] Where n is the number of data pairs, and They represent and The rank in X and Y, where X and Y represent the two indicators paired in this correlation analysis. It is the rank difference of each pair of data. It is the sum of squares of all rank differences.

[0146] like Figure 12The image shows a heatmap representing the correlation coefficient matrix in follow mode. It displays the Spearman correlation coefficients between the three core behavioral features—drumming rhythm accuracy, position accuracy, and intensity accuracy—and the predicted stress value changes, as well as the correlation coefficients between the features themselves. The values ​​are labeled at the corresponding positions on the heatmap. The color intensity of the matrix blocks corresponds to the absolute value of the correlation coefficients; darker colors indicate that the absolute value of the correlation coefficient between the two features is closer to 1, indicating a higher degree of linear correlation. This figure shows that higher rhythm and intensity accuracy in drumming leads to a larger stress difference and a better stress relief effect. Furthermore, a larger absolute value of the correlation coefficient between intensity accuracy and stress difference indicates a greater impact on stress relief, providing direct correlation evidence for quantifying the contribution of behavioral features to stress relief.

[0147] S6. Based on the contribution of different drumming behavior characteristics to stress relief, a comprehensive analysis is conducted based on drumming behavior characteristics and changes in key physiological characteristics to quantitatively evaluate the effect of drumming behavior on psychological stress relief.

[0148] Specifically, step S5 quantifies the contribution of each behavioral feature, while also screening out core behavioral features that significantly contribute to stress relief, eliminating redundant features with no significant impact, reducing the computational complexity of the unsupervised grouping algorithm in step S6, improving the effectiveness and relevance of the clustering results, and ensuring that the subsequent behavioral pattern categories are strongly correlated with the stress relief effect.

[0149] S61. Construction of a comprehensive feature matrix: Combine all extracted behavioral features with key physiological features, and use the predicted change in psychological stress value as an indicator of relief effect to construct a comprehensive feature matrix for clustering.

[0150] Specifically, all extracted behavioral features, physiological features, and predicted changes in stress values ​​are combined into a wide table, with each row representing one experiment and each column representing a feature or indicator.

[0151] S62. Feature dimensionality reduction: After standardizing the comprehensive feature matrix, the data dimensionality is reduced through feature transformation and compression processing to remove redundancy and alleviate multicollinearity.

[0152] S63. Cluster Analysis: For n samples containing drumming behavior characteristics, key physiological characteristics, and predicted changes in psychological stress values, the dimensionality-reduced comprehensive feature matrix is ​​first input into an unsupervised clustering algorithm. Samples with high feature similarity are automatically grouped into one category, and several unlabeled behavior pattern categories are finally output. Each category contains samples with similar features. For example, category 1 mainly consists of samples with high intensity accuracy and uniform rhythm, while category 3 mainly consists of samples with chaotic rhythm and large intensity fluctuations. The central tendency of the predicted changes in psychological stress values ​​of all samples within each behavior pattern category is calculated separately, and the corresponding stress relief effect level (high / medium / low) is matched to each behavior pattern category based on the central tendency. Finally, the stress relief effect corresponding to different combinations of drumming behavior patterns is obtained.

[0153] This invention provides a method for quantifying the psychological stress relief effect based on pulse wave and drumming behavior. It collects PPG signals from subjects in relaxed and stressed states, preprocesses the collected PPG signals to extract physiological features, selects a subset of key physiological features, and trains a stress prediction model. A multi-dimensional physiological feature system is constructed based on pulse wave signals, comprehensively extracting heart rate variability and pulse wave morphological features. The contribution of physiological features is evaluated and selected using a prediction model based on multi-feature inputs, obtaining a subset of key features highly correlated with psychological stress states. By reducing the introduction of redundant features, the system's robustness to noise and individual differences is improved while ensuring the accuracy of stress assessment, and the overall computational complexity is reduced, thereby enhancing the stability and real-time application efficiency of stress assessment. This transforms the traditional stress relief assessment method, which relies on subjective scales, into an objective quantitative assessment method based on changes in physiological signals, achieving objective, continuous, and repeatable assessment, overcoming the problems of poor consistency and insufficient comparability of existing subjective assessment methods. This invention further integrates behavioral characteristics such as rhythm, intensity, and striking position during drumming with physiological state changes, classifies and analyzes different drumming behavior patterns, establishes a quantitative mapping relationship between drumming behavior characteristics and psychological stress relief effects, and can identify combinations of drumming behavior patterns with different stress relief effects. This provides an objective basis for designing personalized and differentiated psychological stress intervention programs, and enhances the system's adaptability and guidance value in practical application scenarios.

[0154] The above embodiments are preferred embodiments of the present invention, but the embodiments of the present invention are not limited to the above embodiments. Any changes, modifications, substitutions, combinations, or simplifications made without departing from the spirit and principle of the present invention shall be considered equivalent substitutions and shall be included within the protection scope of the present invention.

Claims

1. A method for quantifying the psychological stress-relieving effect based on pulse waves and drumming behavior, characterized in that, Includes the following steps: S1. Collect PPG signals from subjects in relaxed and stressed states, preprocess the collected PPG signals and extract physiological features, select a subset of key physiological features and train a stress prediction model. S2. Guide the subject to complete drumming training through a preset drumming training program, obtain the drumming time, drumming force and drumming position of each drumming event, and generate a drumming action time sequence. S3. Based on the time sequence of drumming action and the synchronously acquired piezoelectric sensor signal, calculate the drumming rhythm characteristics, drumming force characteristics, and drumming position characteristics to form the behavioral feature vector of the subject. S4. Collect PPG signal features of subjects before and after drumming training, extract key physiological feature values ​​and standardize them, and output the predicted pressure value changes before and after drumming through the pressure prediction model. S5. Based on the predicted change in stress value and the behavioral feature vector of the subjects, quantify the contribution of different drumming behavior characteristics to stress relief. S6. Based on the contribution of different drumming behavior characteristics to stress relief, conduct a comprehensive analysis of changes in drumming behavior characteristics and key physiological characteristics, and quantitatively evaluate the effect of drumming behavior on psychological stress relief.

2. The method for quantifying the psychological stress relief effect based on pulse wave and drumming behavior according to claim 1, characterized in that, Step S1 includes: S11. Collect PPG signals of subjects in a stable relaxation state as baseline data, and collect PPG signals of subjects during and after the standardized stress-induced task as stress state data. S12. Preprocess the acquired PPG signal and extract physiological features for pressure identification from the preprocessed PPG signal. S13. Perform paired-sample significance testing on various physiological characteristics under relaxed and stressed states, and screen out physiological characteristics that show statistically significant differences between relaxed and stressed states to form a candidate feature set for stress sensitivity. S14. Based on the stress-sensitive candidate feature set, recursive feature elimination is used to filter and obtain multiple key physiological features. Based on the multiple key physiological features, a nonlinear regression model with multiple feature inputs is trained, and the trained nonlinear regression model is used as a stress prediction model.

3. The method for quantifying the psychological stress relief effect based on pulse wave and drumming behavior according to claim 2, characterized in that, The physiological features used for stress identification include temporal and frequency domain features of heart rate variability and pulse wave morphology features.

4. The method for quantifying the psychological stress relief effect based on pulse wave and drumming behavior according to claim 2, characterized in that, Step S2 includes: S21. Guide the subject to complete drumming training through a preset drumming training program. The preset drumming training program includes a drumming training program in follow-up imitation mode and a drumming training program in free mode. S22. Distributed sensor layout and collaborative data acquisition: Piezoelectric sensors are fixed at the center and edge of the drum surface. The voltage signals of all piezoelectric sensors are collected synchronously. The force value of drumming is calculated based on the voltage signals of the piezoelectric sensors. The position tag of drumming is generated according to the area corresponding to the sensor. The timestamp of each drumming event is bound to the position tag and arranged in chronological order to form a drumming action time sequence.

5. The method for quantifying the psychological stress relief effect based on pulse wave and drumming behavior according to claim 1, characterized in that, Step S3 includes: S31. Calculate the drumming rhythm characteristics. Based on the drumming action time sequence, calculate the temporal characteristics reflecting the drumming rhythm, including the time interval between adjacent drumming, rhythm complexity, and rhythm accuracy. S32. Calculate the striking force characteristics. Use the energy value of the signal within a single striking window as the force value of this drumbeat. Calculate the mean, standard deviation, and accuracy of all drumbeat force values. S33. Calculate the characteristics of the striking force, calculate the proportion of striking positions and the frequency of striking position switching, calculate the proportion of the number of times the drumhead is struck and the number of times the drum rim is struck to the total number of times, and calculate the ratio of the number of times the striking positions are different between two adjacent striking positions to the total number of times.

6. The method for quantifying the psychological stress relief effect based on pulse wave and drumming behavior according to claim 1, characterized in that, Step S4 includes: S41. Collect PPG signals from the subjects before and after drumming training. S42. Extract key physiological features from the PPG signals collected from the subjects before and after drumming training, and standardize each key physiological feature. S43. Input the standardized key physiological characteristics before and after drumming training into the stress prediction model to calculate the predicted change in psychological stress value.

7. The method for quantifying the psychological stress relief effect based on pulse wave and drumming behavior according to claim 1, characterized in that, Step S5 includes: S51. The behavioral characteristic vectors of the subjects and the changes in the predicted stress values ​​are grouped according to task mode and difficulty to obtain multiple analysis groups; S52. Correlation analysis: Calculate the Spearman correlation coefficient between the behavioral characteristics and the predicted changes in stress values ​​for each analysis group, construct the correlation coefficient matrix for each group, and determine the direction and intensity of the influence of different drumming behavioral characteristics on stress relief effect by the positive and negative values ​​and significance of the coefficients.

8. The method for quantifying the psychological stress relief effect based on pulse wave and drumming behavior according to claim 1, characterized in that, Step S6 includes: S61. Construction of a comprehensive feature matrix: Combine all extracted drumming behavior features with key physiological features, and use the predicted change in psychological stress value as an indicator of relief effect to construct a comprehensive feature matrix for clustering. S62. Standardize the comprehensive feature matrix and reduce data dimensionality, remove redundancy and alleviate multicollinearity through feature transformation and compression. S63. For multiple samples containing drumming behavior features, key physiological features, and predicted changes in psychological stress values, first input their dimensionality-reduced comprehensive feature matrix into an unsupervised clustering algorithm. Samples with high feature similarity are automatically grouped into one category, and several unlabeled behavior pattern categories are output. The central tendency of the predicted changes in psychological stress values ​​of all samples within each behavior pattern category is statistically analyzed, and the corresponding stress relief effect level is matched to each behavior pattern category based on the central tendency. Finally, the stress relief effect corresponding to different combinations of drumming behavior patterns is obtained.

Images