CN-122004801-A - Evaluation recommendation method for cardiovascular health intervention effect by music score playing training

CN122004801ACN 122004801 ACN122004801 ACN 122004801ACN-122004801-A

Abstract

The invention discloses an evaluation recommendation method of cardiovascular health intervention effects by music score playing training, and relates to the fields of biological signal analysis and machine learning. The method comprises the steps of firstly collecting resting pulse waves of a subject as a base line, guiding the subject to play a plurality of groups of classification test music scores, synchronously recording pulse waves and behavior data before and after playing and setting a recovery period, carrying out preprocessing such as denoising, base line correction and the like on collected signals, extracting characteristics such as pulse wave morphology, pulse rate variability and the like, weighting characteristic change rates by adopting a CRITIC method, calculating a score of comprehensive intervention effect of the music score, taking the highest score as a subject personalized label, and finally constructing and training a machine learning model, taking the resting base line characteristics as input and the optimal score label as output, so as to realize personalized intelligent recommendation of the music score of a new subject. The method realizes objective and quantitative evaluation of the intervention effect, and improves pertinence and suitability of cardiovascular health intervention.

Inventors

WEI GANG
YU JIANPENG
CAO YAN
WANG YIGE
HUANG SHUANGPING

Assignees

华南理工大学

Dates

Publication Date: 20260512
Application Date: 20260210

Claims (10)

1. A method for evaluating and recommending cardiovascular health intervention effects by music score playing training, which is characterized by comprising the following steps: S1, allowing a subject to sit still in a quiet environment until a physiological state is stable, collecting pulse wave signals in the quiet state as baseline reference signals, then guiding the subject to play a plurality of groups of preset test music scores in sequence, synchronously recording behavior data of each group of music scores during playing, and immediately pulse wave signals before and after the playing, wherein a recovery period is set at each group of music score playing interval to eliminate cross interference; s2, sequentially carrying out signal denoising, baseline correction, data segmentation interception and signal quality evaluation on all acquired pulse wave signals, and eliminating signals with quality not up to standard to obtain effective pulse wave signals; S3, extracting characteristic parameters related to cardiovascular health conditions based on the preprocessed effective pulse wave signals, wherein the characteristic parameters comprise pulse wave morphological characteristics and pulse rate variability characteristics; s4, weighting calculation is carried out on the extracted characteristic parameters by adopting a weight determination method based on index correlation, interference effect scores corresponding to each group of test music scores are obtained, and the test music score label with the highest score is used as a sample label of the subject; S5, constructing a machine learning model, taking characteristic parameters corresponding to the resting baseline pulse wave signals of the subjects as model input, taking the score marks with the highest scores as model output targets, and realizing personalized training score recommendation of new subjects after training the machine learning model.
2. The method for evaluating and recommending effects of a music score blowing training on cardiovascular health intervention according to claim 1, wherein the specific process of step S1 is as follows: s101, classifying music scores according to the number of beats per minute, the playing force, the playing air pressure range and the ventilation frequency to form a plurality of groups of test music scores; s102, after a subject closes eyes and sits still, acquiring a pulse wave signal in a resting state by using a photoelectric volume pulse wave tracing sensor as a baseline reference signal; s103, after the subject is still sitting and cross interference is eliminated, pulse wave signals before playing are collected, then a wind instrument is used for playing a specified music score, pulse wave signals after playing are collected immediately after the playing is finished, and the process is repeated until the playing of all test music scores is completed.
3. The method for evaluating and recommending effects of a music score blowing training on cardiovascular health intervention according to claim 1, wherein the specific process of step S2 is as follows: s201, filtering the pulse wave signal by adopting a multi-order low-pass Butterworth filter to filter high-frequency interference noise; S202, performing baseline correction on the filtered signals by using a cubic spline interpolation method, and removing baseline drift interference; S203, carrying out sectional processing on the signal after the baseline removal by adopting a sliding window method, and setting the uniform window size and step length; S204, carrying out quality evaluation on the segmented signals of each segment by adopting an attractor reconstruction method, and screening out effective signal segments meeting a preset quality threshold.
4. The method of claim 3, wherein step S204 is specifically implemented as follows: firstly, constructing a correlation signal by setting time delay based on a segmented pulse wave segment, obtaining a new variable through projection transformation, and forming an image of a reconstructed attractor by taking specific variables as horizontal coordinates and vertical coordinates respectively; then, three key feature points are determined in the reconstructed image, so that an ideal triangle is constructed, and discrete straight lines corresponding to three sides of the triangle are obtained; Dividing the track of the attractor into corresponding sections according to the three characteristic points, respectively calculating the errors of the corresponding sides of each track section and the triangle, taking the maximum value of the errors of all track sections of each side, and synthesizing the three maximum values to obtain the global error of the attractor; and finally, setting a quality threshold, judging that the quality of the pulse wave signal reaches the standard if the global error does not exceed the threshold, and judging that the quality does not reach the standard and eliminating if the global error exceeds the threshold.
5. The method for evaluating and recommending effects of a music score blowing training on cardiovascular health intervention according to claim 1, wherein the specific process of step S3 is as follows: S301, detecting a main wave crest of the screened pulse wave signals, and positioning a single complete pulse wave period by taking the identified main wave crest as a datum point; S302, extracting pulse wave morphology features in a single complete pulse wave period based on the positioned pulse wave period, and respectively taking the average value of the morphology features of all effective pulse wave periods as the standardized morphology features of corresponding samples; S303, calculating pulse rate variability characteristics of the pulse wave signals based on the pulse wave signals after peak detection, wherein the pulse rate variability characteristics cover time domain characteristics, frequency domain characteristics and nonlinear domain characteristics.
6. The method for evaluating and recommending effects of a music score blowing training on cardiovascular health intervention according to claim 5, wherein in step S301, a first derivative peak detection algorithm with an adaptive threshold is adopted to perform main wave peak detection on a pulse wave signal, specifically: Firstly, dividing a discrete signal into a plurality of equal-length paragraphs, setting a time window with a specific length at the initial part of each paragraph, estimating signal amplitude characteristics and instantaneous heart rate by searching local maxima and the like, determining an initial amplitude threshold value and a time interval threshold value, and dynamically adjusting the two threshold values according to the heart beat characteristics detected subsequently; Then scanning the whole discrete sequence, finding a zero crossing point in the differential sequence, checking whether the amplitude of the point exceeds the current amplitude threshold value in the original signal, and simultaneously confirming that the interval between the point and the last accepted peak value meets the time interval threshold value requirement, wherein only the point meeting the two conditions is judged as the peak value point in the systolic period; Finally, for each pair of adjacent peak points, a signal segment between the two peak points is extracted, and the minimum value point in the signal segment is positioned and used as a starting point of a corresponding pulse wave period, wherein the starting point is usually at the lowest point of each period and corresponds to the starting position of a signal waveform.
7. A method of recommending the evaluation of cardiovascular health intervention effects by score-playing training according to claim 5, wherein in step S303, the time domain features include: PP interval mean value, namely calculating the average value of all pulse intervals to reflect the overall rhythm of pulse waves; normal PP interval standard deviation, namely, the total variability of the pulse intervals on a longer time scale is reflected by calculating the standard deviation of all the pulse intervals; the root mean square of the interval difference of adjacent PP, reflect the short-term fluctuation characteristic by calculating the root mean square of the interval difference of adjacent pulse, mainly used for assessing parasympathetic nerve activity level; counting the percentage of the number of the adjacent pulse interval differences exceeding 20ms to the total interval number; The frequency domain features include: Low frequency characteristic value, corresponding to specific frequency range, reflecting the complex regulatory functions of the sympathetic nerve and vagus nerve, and the activity of the sympathetic nervous system is related; high frequency characteristics, namely mapping the regulation activity of the vagus nerve corresponding to another specific frequency range, and correlating with respiratory arrhythmia; the ratio of low-frequency power to high-frequency power, which reflects the balance between the sympathetic nervous system and the vagus nervous system, is an index of the overall balance of the autonomic nervous system; The nonlinear domain features include: sample entropy, namely quantifying the complexity of signals by comparing the similarity probabilities of templates with different lengths in a time sequence, wherein the dependence on the data length is low, the calculation is stable, and the larger the value is, the more complex the signal is; The multi-scale entropy is used as the expansion of the sample entropy, the sample entropy under each scale is calculated and integrated by carrying out the transformation of different scales on the original sequence, and the integral trend of the long-time scale and the detail characteristics of the short-time sequence are reflected; The approximate entropy is that the regularity and complexity of the signal are quantized through counting the occurrence probability of a new mode in the sequence, the larger the numerical value is, the harder the signal is predicted, and the weaker the regularity is.
8. The method for evaluating and recommending effects of a music score blowing training on cardiovascular health intervention according to claim 1, wherein the specific process of step S4 is as follows: s401, respectively calculating the change rate of each characteristic relative to a rest baseline before performance after each group of test music scores are performed by a subject according to the extracted standardized pulse wave morphological characteristics and pulse rate variability characteristics; s402, calculating the weight of each characteristic parameter by adopting a weight determination method based on index correlation based on the multi-dimensional characteristic change rate data set; S403, according to the determined objective weights of the characteristics, carrying out weighted summation on the multi-dimensional characteristic change rate generated after each subject plays each group of test music score, and calculating to obtain the comprehensive intervention effect score corresponding to the music score, finally, selecting the test music score with the highest comprehensive intervention effect score for each subject, and taking the unique label of the music score as the personalized training sample label of the subject.
9. The method for evaluating and recommending effects of a music score blowing training on cardiovascular health intervention according to claim 8, wherein in step S402, a weight determining method based on index correlation is adopted to calculate weights of each characteristic parameter, specifically: Firstly, constructing an evaluation matrix taking different music scores of a subject as rows and the change rate of each feature as columns; Secondly, carrying out data standardization processing on the evaluation matrix to eliminate dimension differences of all the features; then calculating a correlation coefficient matrix among the features, so as to quantify the conflict among the features, namely the contrast strength; then, combining the standard deviation of the standardized features and the conflict between the features, and calculating the information quantity contained in each feature; And finally, determining the final objective weight of each characteristic parameter according to the proportion of each characteristic information quantity to the total information quantity.
10. A method for evaluating and recommending the effects of a music score blowing exercise on cardiovascular health intervention according to claim 1, the method is characterized in that the step S5 specifically comprises the following steps: S501, integrating data of all subjects, taking standardized morphological characteristics and pulse rate variability characteristics obtained after the rest baseline pulse wave signals of each subject are processed in the steps S2 and S3 as input characteristic vectors of a machine learning model, and taking corresponding score marks with the highest scores as output target labels to form a structured feature-label data set; S502, dividing a training set and a verification set according to subjects, and ensuring that data of the same subject belongs to only one set; s503, training a random forest classification model by using training set data, optimizing super parameters by adopting grid search and five-fold cross verification, and determining an optimal model by selecting a parameter combination with highest cross verification accuracy; S504, collecting resting pulse wave signals of a new subject, inputting an optimal model after preprocessing and feature extraction, and outputting a corresponding optimal music score label to realize personalized recommendation.

Description

Evaluation recommendation method for cardiovascular health intervention effect by music score playing training Technical Field The invention relates to the fields of biological signal analysis, signal processing, machine learning and the like, in particular to an evaluation recommendation method for cardiovascular health intervention effects of music score playing training. Background Cardiovascular health is a key element of overall health management of human bodies, and particularly has outstanding prevention and intervention values in middle-aged and elderly people, chronic disease people and sub-health people. Effective cardiovascular health intervention can realize accurate regulation and scientific management by breaking through the limitation of traditional static index monitoring and combining a dynamic and individual physiological feedback mechanism. The pulse wave signal is used as a noninvasive, convenient and rich-information physiological signal, can objectively reflect a plurality of key cardiovascular function parameters such as vascular elasticity, peripheral resistance, heart rate variability and the like, becomes a core basis for real-time and dynamic evaluation of cardiovascular health status, and provides reliable physiological data support for health intervention. In the field of health intervention, music training, especially wind instrument training, has been demonstrated to indirectly improve cardiovascular function by virtue of its positive regulation of respiratory rhythm, precise control of psychological state, and is one of the important directions for non-pharmaceutical intervention. However, the existing music intervention method still has significant limitations that firstly, an intervention scheme is mostly based on unified music teaching logic or general psychological adjustment principle, a quantitative evaluation system aiming at individual cardiovascular physiological states is lacking, accurate matching of physiological characteristics and intervention contents cannot be achieved, secondly, evaluation of cardiovascular health intervention effects is mostly dependent on subjective questionnaire feedback or staged physical examination data, dynamic changes of physiological indexes in the process of blowing training are difficult to capture in real time, intervention strategies cannot be adjusted in time, thirdly, although single parameters such as heart rate variability and blood pressure are studied to be used for relaxing training feedback, the single parameters are not combined with structured and personalized music score blowing training, so that intervention form is general, individual suitability is insufficient, intervention potential of the blowing training on cardiovascular health is difficult to be fully exerted, and large-scale application of the method in personalized health management is limited. In addition, a complete technical system integrating physiological signal acquisition, feature analysis, intervention effect quantification and personalized recommendation is lacking in the prior art, the multidimensional features of pulse wave signals cannot be effectively associated with key parameters such as rhythm, dynamics and ventilation frequency of a music score, and a data-driven personalized intervention scheme is difficult to form. Therefore, a method for quantitatively evaluating the intervention effect of the music score playing training on cardiovascular health based on the individual dynamic pulse wave signals and realizing personalized intelligent music score recommendation is needed, so that the problems of insufficient pertinence of intervention, lag evaluation mode, low system integration and the like in the prior art are solved, and the cardiovascular health non-drug intervention is promoted to develop to the accurate and personalized direction. Disclosure of Invention The invention aims to overcome the defects and shortcomings of the prior art, provides an evaluation recommendation method for cardiovascular health intervention effects by music score playing training, combines individual dynamic pulse wave signals to match personalized playing training music scores for subjects, greatly improves pertinence and suitability of cardiovascular health intervention, and simultaneously can realize evaluation recommendation for cardiovascular health intervention effects by fusing modern signal processing technology and machine learning algorithm, and can customize exclusive music intervention schemes for individuals so as to promote floor implementation of a personalized health management system. In order to achieve the above purpose, the present invention adopts the following technical scheme: The invention provides an evaluation recommendation method of cardiovascular health intervention effects by music score playing training, which comprises the following steps: S1, allowing a subject to sit still in a quiet environment until a physiological state is