Piano playing work memory testing and training integrated quantitative evaluation method

Abstract

The invention discloses a work memory measurement and training integrated quantitative evaluation method for piano playing, which realizes the integration of work memory measurement and training by quickly memorizing a music score and reproducing the piano playing of playing music and combining human body physiological monitoring and piano. The method comprises the following steps: collecting pulse waves during individual static sitting and piano playing training, and recording playing performance characteristics such as piano playing accuracy and the like, and response time and accuracy of executing a work memory task normal form N-back after training; extracting physiological features such as heart rate based on the preprocessed pulse wave signal; constructing a distribution model based on N-back reaction time, and obtaining a score for model training reference by combining accuracy; and establishing a random forest evaluation model in combination with the physiological features and the playing performance features, and quantitatively evaluating the work memory of the individual and the work memory training effect of piano playing. According to the invention, music training and physiological signal acquisition are combined, and an objective and generalizable scheme is provided for work memory evaluation and training.

Description

Technical Field

[0001] The invention belongs to the field of cognitive ability assessment and information technology, and specifically relates to an integrated quantitative assessment method for testing and training working memory of silent piano playing. Background Art

[0002] As an important research field in psychology and neuroscience, memory involves the diversity of the brain in encoding, storing and retrieving information. According to the differences in the brain's encoding, storage and retrieval of information, as well as the different patterns of neuronal activity, memory can be divided into short-term memory and long-term memory. Working memory (WM) is a special form of short-term memory, which refers specifically to an individual's ability to store and manipulate information within a limited time. It is an important component of cognitive function, involving the storage, processing and manipulation of instantaneous information, and constitutes the core function of complex cognitive activities (such as reasoning, decision-making, problem solving and language comprehension). Therefore, working memory plays an extremely important role in daily life.

[0003] Existing working memory training methods mainly include computerized training, auditory task training, and visual task training. Although these methods can improve an individual's working memory ability to a certain extent, they often have some shortcomings. For example, computerized training may lack application in real situations, resulting in low transferability of training effects in daily life. In addition, traditional training methods mostly focus on specific types of tasks and fail to comprehensively improve an individual's working memory function.

[0004] In terms of working memory assessment, existing methods mainly rely on behavioral tests, such as digital sequence recall, letter sequence recall, etc. Although these methods are more commonly used, they also have certain limitations. First, these assessment methods mainly rely on the subjective response of individuals and lack interest. Individuals may be more easily affected by factors such as emotions and attention, resulting in unstable results. Second, traditional assessment methods usually cannot fully reflect the physiological basis of working memory and lack objective measurement of individuals.

[0005] Studies have shown that piano playing is more effective than traditional computerized training in training working memory. Among them, silent piano playing is not only a form of artistic expression, but also a complex cognitive activity that requires individuals to quickly memorize music scores and reproduce music. Silent piano playing requires individuals to simultaneously process the visual information of the music score and the memory of the music score. It also includes motor control during the playing process and continuous memory refresh in the music score. It involves the encoding, storage and retrieval of information, and has more dimensional training effects than traditional training. Through silent piano training, the individual's working memory can be effectively activated and the working memory can be trained in multiple dimensions. Summary of the invention

[0006] The objective of the present invention is to improve the limitations of traditional working memory training methods in terms of effectiveness and transferability, as well as the deficiencies of existing working memory assessment methods in terms of subjectivity and reliability, and to provide a quantitative assessment method for integrated training and measurement of working memory in piano silent performance. The present invention integrates piano playing, working memory training, and working memory measurement, provides a more interesting way to train and evaluate working memory, and uses objective measurement of the physiological state of photoplethysmogram (PPG) signals and piano silent performance to evaluate an individual's working memory, thereby objectively quantifying the training effect of piano silent performance on working memory and achieving the integration of working memory training and measurement.

[0007] To achieve the above objective, the present invention adopts the following technical solutions:

[0008] A quantitative assessment method for integrated training and measurement of working memory in piano silent performance aims to improve the effectiveness and transfer effect of training by introducing piano silent performance as a novel working memory training method, and at the same time uses photoplethysmogram (PPG) signals for objective assessment to overcome the subjectivity and instability of traditional assessment methods and achieve the integration of measurement and training. The identification method includes the following steps:

[0009] S1. Construct an integrated test and training scenario, collect the pulse wave signals of the subject during sitting still and during piano silent performance, and record the performance characteristics such as the playing accuracy in the individual's piano silent performance training. After one round of piano silent performance training, conduct one round of the working memory task paradigm N-back to record the reaction time and accuracy rate; among them, the piano silent performance training refers to repeatedly displaying a random note sequence of a certain length on the screen with a certain time limit; in the first round, the subject needs to remember the sound note sequence of the current round, and starting from the second round of testing, the subject needs to play the music score of the previous round and remember the note sequence of the current round; the working memory paradigm task is the working memory task paradigm N-back, the reaction time refers to the time between the appearance of the stimulus and the subject's response in the working memory task, the correct rate refers to the correct rate of judging whether the existing stimulus is consistent with the stimulus content of the previous N rounds after remembering N stimuli, the stimulus is an arbitrary Arabic numeral (0-9), the subject needs to remember the Arabic numerals in the latest N stimuli, compare the Arabic numeral of the current stimulus with the Arabic numerals in the previous N stimuli, and make a judgment of consistency or inconsistency;

[0010] Place an intelligent piano and a photoplethysmogram (PPG) acquisition device in the test scenario. The subject places the middle finger in the PPG acquisition device to obtain the pulse wave signal; the intelligent piano refers to a modified piano that can record performance characteristics such as playing accuracy and playing time during an individual's piano silent performance;

[0011] S2. Denoise the collected pulse wave signals; use an optimized algorithm combining genetic optimization algorithm and variational mode decomposition to decompose the original pulse wave signals into several intrinsic mode functions; on this basis, calculate the central frequencies of each IMF signal, and apply wavelet threshold filtering technology according to these central frequencies to obtain the filtered pulse wave signals, and then further remove noise and baseline by combining with a Butterworth low-pass filter, remove the periods affected by motion artifacts through adaptive period screening to obtain the preprocessed pulse wave signals, and finally perform feature point detection and extraction; this composite filtering method can better filter the noise in the original pulse wave signals and improve the signal-to-noise ratio of the pulse wave signals.

[0012] S3. Extract the heart rate, heart rate variability, blood oxygen saturation and morphological features of the subject from the preprocessed pulse wave signals; among them, the heart rate and heart rate variability can be further calculated by extracting the peak-to-peak time intervals of the pulse waves, and the blood oxygen saturation is obtained by formula calculation and fitting of the pulse waves detected by the infrared light and red light channels collected by the PPG device, and the morphological features are extracted by extracting the characteristic points of each pulse wave cycle; construct a working memory physiological feature dataset through the physiological features obtained above; heart rate variability, as an important feature reflecting the cardiac autonomic regulation ability, reflects an individual's cognitive ability and has a close relationship with working memory; the change of blood oxygen saturation is also related to the oxygen supply of the brain, which in turn affects the performance of working memory; morphological features reflect blood pressure, peripheral resistance, etc. and are related to working memory.

[0013] S4. Conduct a mathematical statistical distribution analysis on the individual working memory task response time recorded in step S1, construct a distribution model of the response time and use the Kolmogorov-Smirnov test to establish a distribution function of the response time, and combine the correct rate of the working memory task to quantify the reference score after the S1 piano silent performance training through the distribution model.

[0014] S5. Conduct the N-back error step of the working memory task paradigm on several subjects, and then enter the piano silent performance working memory training cycle, repeat steps S1 - S3 to obtain a physiological feature dataset characterized by heart rate, heart rate variability, blood oxygen saturation and morphological features, the performance features in the piano silent performance working memory training and the distribution model fitted by the N-back response time of the working memory task paradigm, standardize the physiological feature dataset, and use the training reference score as the target variable to construct a working memory evaluation model through a random forest model in combination with the working memory physiological feature dataset and the performance features in the individual piano silent performance working memory training.

[0015] S6. Input the piano behavior data and physiological data of individual piano silent practice training into the working memory evaluation model in step S5, so as to evaluate the training effect of working memory.

[0016] Furthermore, the process of step S1 is as follows:

[0017] S101. Under the external conditions of sufficient light and no interference, collect the pulse wave signals of the subject during sitting still and during piano silent practice training through the PPG collection device placed on the side of the subject's single hand; based on the characteristics of the continuity and variability of working memory, the collected pulse wave signals respectively correspond to the sitting still baseline pulse wave signal and the pulse wave signal during piano silent practice training; the pulse wave signals collected by the PPG collection device in this step are used to extract physiological features later, construct a physiological feature dataset, and objectively evaluate the working memory and the training effect of piano silent practice.

[0018] S102. Repeatedly display a random musical note sequence of a certain length on the screen, with a certain time limit for the display; in the first round, the subject needs to remember the musical note sequence of the current round, and starting from the second round of testing, the subject needs to play the music score of the previous round and remember the musical note sequence of the current round. Record the playing results of each musical note of the subject during piano silent practice training, so as to calculate the playing accuracy of the subject; this step combines the working memory task paradigm N-back with the characteristics of piano playing, integrating piano playing, working memory training, and working memory testing, and providing a more interesting working memory training and testing.

[0019] Among them, the piano playing accuracy is:

[0020] In the formula, Accurancy piano represents the piano playing accuracy, Num match represents the number of keys pressed by the subject that match the note positions and pitches in the corresponding musical note sequence, and Sum note represents the total number of musical notes;

[0021] The piano playing time is: T piano = T last - T start

[0022] In the formula, T piano represents the playing time, T last represents the time when the subject presses the last key, and T start represents the start time of the display of the musical note sequence;

[0023] S103. Under the external condition of quietness, the subject performs the working memory task paradigm N-back;

[0024] S104. Display the stimulus content of the N-back working memory task paradigm on the screen. The subject needs to compare the currently displayed stimulus with the previous N stimuli in memory and determine whether the content of the currently displayed stimulus is consistent with that of the previous N stimuli in memory. The stimulus is an arbitrary Arabic numeral (0-9). This step of displaying the numerals for the subject to remember and judge on the screen ensures that the subject performs the working memory task within a continuous period of time, stimulating the maintenance, refreshing, and scheduling of the working memory part of the subject; this step examines the working memory of the subject based on the finiteness of working memory capacity.

[0025] S105. The first N rounds of stimuli are reference stimuli. Starting from the (N + 1)-th round, the subject needs to compare the currently displayed numeral with the previous N numerals in memory, determine whether the content of the currently displayed stimulus is consistent with that of the previous N numerals in memory, and record the reaction time of the subject and whether the judgment is accurate to calculate the accuracy rate. The reaction time and accuracy rate of the subject performing the working memory recorded in this step are used to establish a distribution model later for quantitatively scoring the working memory.

[0026] Among them, the reaction time calculation formula is: RT = T response -T show

[0027] In the formula, RT is the reaction time, T response and T show respectively represent the time when the subject presses the button and the time when the current numeral starts to be displayed.

[0028] The accuracy calculation formula is:

[0029] In the formula, Accurancy represents the accuracy rate, Num correct represents the number of correct judgments of the subject, and Num total represents the total number of stimuli that the subject needs to judge.

[0030] Furthermore, the process of step S2 is as follows:

[0031] S201. First, define the objective function and initialize the population of the genetic optimization algorithm. Use the mean square error in the signal reconstruction error as the objective function to evaluate its fitness, and optimize the key parameters in the variational mode decomposition optimized by the genetic optimization algorithm for the collected original pulse wave signal - the number of modes K and the penalty factor α. This step globally optimizes the parameters of the variational mode decomposition through the genetic optimization algorithm, avoiding the possible local optimal problem faced by the traditional variational mode decomposition method, so as to better adapt to the signal.

[0032] S202. Repeat step S201 to obtain the optimal parameter combination in variational mode decomposition, and perform the variational mode decomposition process on the collected original pulse wave signal to obtain several IMF components;

[0033] S203. Perform a fast Fourier transform on the several IMF components obtained in step S202 to calculate their central frequencies, and select M high-frequency IMF components with higher central frequencies according to the central frequencies. This step is to determine the central frequency for the subsequent refined processing of the pulse wave signal;

[0034] S204. Perform wavelet threshold denoising on the M high-frequency IMF components selected in step S203. This step performs refined processing on the pulse wave signal; wavelet threshold denoising has a good filtering and denoising effect on noises such as environmental noise, motion artifacts, and power frequency noise, while retaining some useful information in the pulse wave signal;

[0035] S205. Linearly superimpose the M high-frequency IMF components processed in step S204 to reconstruct the preliminarily preprocessed pulse wave signal;

[0036] S206. Filter the preliminarily preprocessed pulse wave signal obtained in step S205 using a Butterworth low-pass filter. This step aims to further remove various noises in the pulse wave signal;

[0037] S207. Detect and extract the feature points of the pulse wave signal filtered in step S206 to obtain the starting point. This step aims to extract the starting point of the pulse wave to prepare for removing the baseline drift in the subsequent process;

[0038] S208. Use the successfully located starting point of the pulse wave to obtain the baseline drift curve through cubic spline interpolation, so as to achieve baseline drift removal and finally obtain the preprocessed pulse wave signal. This step aims to extract the pulse wave base point through feature point detection, and use the cubic spline interpolation method to fit the pulse wave base point to achieve baseline drift removal and obtain the baseline drift curve; further, by linearly subtracting the baseline drift curve from the pulse wave signal, the baseline drift noise of the pulse wave signal is removed;

[0039] S209. Segment each pulse cycle through the starting point of each pulse cycle obtained in step S207, use the Pearson correlation coefficient to judge the signal quality of the current pulse wave cycle, select the one with the highest Pearson correlation coefficient as the cycle template, and use the cycle template to screen out abnormal cycles with the remaining cycles through the Pearson correlation coefficient. Cycles with a Pearson correlation coefficient lower than a certain value are determined as abnormal cycles and excluded;

[0040] S210. Detect and extract the characteristic points from the pulse wave signal that has undergone the filtering process in step S209 to obtain the peak points and dicrotic wave peak points. This step aims to extract the characteristic points of the pulse wave to prepare for subsequent feature extraction.

[0041] Further, the process of step S3 is as follows:

[0042] S301. Obtain the peak points of the pre-pulse wave signal from step S2, extract the times corresponding to all the peak points in the pulse wave signal, calculate the average heart rate mean_hr, and use it for the extraction of heart rate variability. This step provides a data basis for subsequent heart rate variability feature extraction by calculating the times corresponding to all the peak points;

[0043] S302. According to the times corresponding to all the peak points in the pulse wave signal obtained in step S301, perform a difference calculation on the time series corresponding to the peak points to obtain the time interval series of the peak points of the pulse wave signal, that is, the heart beat time interval series. Calculate the physiological features of heart rate variability through the heart beat time interval series of the pulse wave signal, including time domain features and frequency domain features. The time domain features include the root mean square of the differences between adjacent heart beat intervals RMSSD, the standard deviation of the differences between adjacent heart beat intervals SDSD, the standard deviation of normal heart beat intervals SDNN, the proportion pNN50 of the differences between adjacent heart beat intervals exceeding 50 ms, the proportion pNN20 of the differences between adjacent heart beat intervals exceeding 20 ms, and the triangular index of heart rate variability triangular_index; the frequency domain features include the total power of the signal TP, the very low frequency power VLF, the low frequency power LF, the high frequency power HF, the normalized low frequency power LFnu, and the normalized high frequency power HFnu. This step constructs a physiological feature dataset that reflects the autonomic nerve regulation ability in multiple dimensions through the analysis and extraction of multi-dimensional physiological features of the pulse wave signal;

[0044] S303. Combine the data of the red light channel and the infrared light channel of the collected pulse wave signal with the empirical formula of blood oxygen saturation containing unknown parameters, and perform linear fitting with the measured data of a standard finger clip oximeter to estimate the unknown parameters in the empirical formula of blood oxygen saturation to obtain a complete empirical formula of blood oxygen saturation, and obtain the blood oxygen saturation SpO 2 . This step provides a theoretical calculation basis for the extraction of blood oxygen saturation by constructing an empirical formula of blood oxygen saturation, so as to extract the blood oxygen saturation features that reflect the working memory load situation and the working memory performance situation;

[0045] S304. Extract morphological features from each waveform using the feature points extracted in step S2. These include the rising branch time T1, the falling branch time T2, the reflection time T3, the systolic peak amplitude HA, and the dicrotic wave peak amplitude HB. The rising branch time refers to the time interval between the starting point of the pulse wave and the systolic peak, the falling branch time is the time interval from the systolic peak to the end point, and the reflection time is the time interval from the systolic peak to the dicrotic wave peak. This step constructs a direct reflection of hemodynamic characteristics through pulse wave morphological feature extraction, thereby indirectly reflecting the working memory situation.

[0046] Further, the process of step S4 is as follows:

[0047] S401. Statistically analyze all the reaction time data recorded in step S1 and calculate the parameters of the theoretical distribution of the reaction time data. This step aims to statistically analyze the recorded reaction time data to prepare for subsequent distribution fitting.

[0048] S402. Calculate the cumulative distribution function of the theoretical distribution and use the Kolmogorov - Smirnov test to check the goodness of fit of the theoretical distribution of the data. This step aims to check the goodness of fit of the theoretical distribution of the data, screen out the optimal distribution model that conforms to the characteristics of the actual data, and reduce the interference of outliers.

[0049] S403. Using the cumulative distribution function of the theoretical distribution obtained in step S402, combine the accuracy rate and reaction time in the N - back trials of the working memory task paradigm in step S1 for the subjects. Input the reaction time in step S1 into the cumulative distribution function to obtain the reaction time score, weight the reaction time and combine it with the accuracy rate to obtain the training reference score, and the total score ranges from 0 to 100 points.

[0050] Among them, the calculation formula for the training reference score is as follows: Score = (1 - CDF(RT)) × 100 × Accurancy

[0051] In the formula, CDF is the fitted cumulative distribution function, and RT is the average reaction time for answering 2 - back questions in this round.

[0052] This step aims to construct a multi - dimensional and objective working memory evaluation benchmark by combining the accuracy rate and reaction time dimensions with the optimal distribution model, which is used as the target variable for the subsequent model, aiming to improve the model accuracy.

[0053] Further, the process of step S5 is as follows:

[0054] S501. Conduct piano silent playing working memory training and the working memory task paradigm N - back for several subjects, as in step S1. This step aims to obtain a dataset for constructing a random forest regression model through experiments on several subjects.

[0055] S502. Repeat steps S2 - S3 to obtain a physiological feature dataset characterized by heart rate, heart rate variability, blood oxygen saturation, and morphological features, as well as performance features in piano silent memory training. This step aims to extract features from the collected subject data to obtain a feature dataset for constructing a random forest regression model;

[0056] S503. Standardize the physiological feature dataset and construct a model dataset that corresponds one - to - one with the corresponding training reference scores in combination with the performance features in individual piano silent memory training. This step aligns the feature variables and target variables by standardizing the physiological feature dataset, providing a standardized training dataset for subsequent random forest modeling, thereby improving the model performance;

[0057] S504. Perform feature screening on the model dataset, remove redundant features, and construct an optimal dataset; the purpose of this step is to reduce model overfitting and improve the model learning efficiency and accuracy;

[0058] S505. Through a random forest regression model, construct a working memory evaluation model, combine cross - validation, verify the regression model error, and realize the mapping from physiological features and performance features in piano silent memory training to training reference scores, thereby evaluating individual working memory and working memory training effects. This step improves the model training effect through cross - validation, realizes the construction of the working memory evaluation model, and subsequently individuals can directly evaluate their own working memory and working memory training effects through pulse wave signal acquisition and piano silent training.

[0059] Further, the process of step S6 is as follows:

[0060] S601. An individual conducts piano silent training to obtain corresponding pulse wave signals and piano silent performance features, without performing the working memory task paradigm N - back. This step, based on step S5, realizes the evaluation of individual working memory and working memory training effects only through pulse wave signals and piano silent performance features;

[0061] S602. Obtain physiological features of heart rate, heart rate variability, blood oxygen saturation, and pulse wave morphological features through steps S2 - S3. This step extracts the physiological features for using the working memory evaluation model;

[0062] S603. Input the physiological features and piano silent performance features into the working memory evaluation model to obtain a score, thereby visually evaluating individual working memory and working memory training effects through score changes. This feature directly evaluates individual working memory and working memory training effects by inputting physiological features and piano silent performance into the working memory evaluation model.

[0063] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0064] 1) The present invention uses piano silent playing as a method for working memory training. By constructing a test scenario, it collects the pulse wave signals of an individual during sitting still and during piano silent playing training, and records the playing accuracy rate of the individual during piano silent playing training, providing an integrated solution for working memory training and measurement, which solves the singleness and subjectivity of traditional working memory training methods and measurement methods. At the same time, using a photoplethysmogram (PPG) acquisition device, the subject places the middle finger in the PPG acquisition device to obtain the pulse wave signal.

[0065] 2) The present invention proposes a method for quantitatively scoring working memory based on statistical and regression models. Using performance characteristics such as heart rate, heart rate variability, blood oxygen saturation, morphological features extracted from the preprocessed pulse wave signal, and playing accuracy recorded by an intelligent piano as inputs; conducting a mathematical statistical distribution analysis on the working memory task paradigm N-back, constructing a distribution model of reaction time and using the Kolmogorov-Smirnov test to establish a distribution function of reaction time, and combining with the correct rate of the working memory task to obtain a reference score of the model, and using the reference score as the output to train a random forest regression model. This method can objectively reflect the working memory level and the training effect of piano silent playing working memory, avoiding the one-sidedness caused by only measuring through physiological characteristics.

[0066] 3) The present invention uses an improved signal processing method. Using an optimization algorithm that combines a genetic optimization algorithm and variational mode decomposition to decompose the pulse wave signal into several intrinsic mode function (IMF) signals, calculating the central frequency of each IMF signal, and performing wavelet threshold filtering according to the central frequency to obtain the filtered pulse wave signal, and combining with adaptive period screening to adaptively extract the pulse wave signal period template to detect and eliminate abnormal periods disturbed by motion artifacts, which can better extract useful signals and further improve the accuracy of working memory and the evaluation of working memory training effect. Description of the Drawings

[0067] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are only some embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0068] Figure 1 It is a flowchart of an integrated quantitative evaluation method for working memory measurement and training of piano silent playing disclosed in the present invention;

[0069] Figure 2 It is a flowchart for denoising the original pulse wave in Embodiment 1 of the present invention;

[0070] Figure 3 is the flowchart of the extraction process of heart rate, heart rate variability, blood oxygen saturation, and morphological features in Embodiment 1 of the present invention;

[0071] Figure 4 is the scatter plot of model prediction in Embodiment 1 of the present invention;

[0072] Figure 5 is the schematic diagram of IMF components in Embodiment 2 of the present invention;

[0073] Figure 6 is the schematic diagram of IMF components, their spectrograms, and central frequencies in Embodiment 2 of the present invention;

[0074] Figure 7 is the schematic diagram of the reconstructed pulse wave signal in Embodiment 2 of the present invention;

[0075] Figure 8 is the schematic diagram of the signal after Butterworth filtering in Embodiment 2 of the present invention;

[0076] Figure 9 is the schematic diagram of the signal after baseline drift removal in Embodiment 2 of the present invention. Detailed implementation manners

[0077] To enable those skilled in the art of the present technology to better understand the solution of this application, the technical solutions in the embodiments of this application will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of this application. Obviously, the described embodiments are only a part of the embodiments of this application, rather than all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative efforts fall within the scope of protection of this application.

[0078] Referring to "embodiment" in this application means that the specific features, structures, or characteristics described in conjunction with the embodiment may be included in at least one embodiment of this application. The phrase appears in various positions in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment mutually exclusive with other embodiments. Those skilled in the art understand explicitly and implicitly that the embodiments described in this application can be combined with other embodiments.

[0079] Embodiment 1

[0080] Figure 1It is a flowchart of an integrated quantification evaluation method for working memory measurement and training in piano silent performance provided by an embodiment of the present invention. This embodiment is implemented in the scenario of a piano silent performance and a working memory task paradigm N-back, and a random forest evaluation model is established, so as to be able to evaluate the working memory level and the training effect of working memory, and provide a model for directly evaluating working memory and the training effect of working memory in Embodiment 2. The specific steps are as follows:

[0081] S1. Construct an integrated test and training scenario, collect the pulse wave signals of the subject during sitting still and during piano silent performance, and record performance characteristics such as the playing accuracy in the individual piano silent performance training. After one round of piano silent performance training, conduct a 2-back task test in the working memory task paradigm N-back experimental paradigm to record the reaction time and accuracy rate; in the test scenario, place a photoplethysmogram acquisition device on the non-dominant hand side of the subject, and the subject places the middle finger of the non-dominant hand in the PPG acquisition device to obtain the pulse wave signal;

[0082] S101. Under the external conditions of sufficient light and no interference, collect the pulse wave signals of the subject during sitting still and during piano silent performance training through the PPG acquisition device placed on the single hand side of the subject; based on the characteristics of the continuity and variability of working memory, the collected pulse wave signals respectively correspond to the sitting baseline pulse wave signal and the pulse wave signal during piano silent performance training;

[0083] S102. Repeat the display of a random note sequence with a length of 7 on the screen, with a display time limit of 15 seconds, for a total of 8 rounds; in the first round, the subject needs to memorize the note sequence of the current round. Starting from the second round of testing, the subject needs to play the score of the previous round and memorize the note sequence of the current round. Record the playing result of each note of the subject during piano silent performance training, so as to calculate the playing accuracy of the subject;

[0084] S103. Under quiet external conditions, the subject performs the working memory task paradigm N-back on the computer, and the measurement duration of performing the working memory task is approximately 140 seconds;

[0085] S104. In this embodiment, the 2-back working memory paradigm task in the working memory task paradigm N-back is adopted. The "+" sign for maintaining memory retention and enabling the subject to concentrate attention appears on the screen for 0.7 seconds, and then the stimulus content of the 2-back working memory paradigm, that is, a randomly generated Arabic numeral (0-9), is displayed for 2 seconds. Subsequently, the above cycle is repeated 52 rounds in total; the subject needs to compare the currently displayed stimulus with the previous 2 stimuli in memory and determine whether the content of the currently displayed stimulus is consistent with that of the previous 2 stimuli in memory; for the 2-back working memory paradigm task, the subject needs to compare the currently displayed stimulus with the previous 2 stimuli in memory and determine whether the content of the currently displayed stimulus is consistent with that of the previous 2 stimuli in memory. If the subject judges them to be consistent, the dominant hand presses the button '1', and if the subject judges them to be inconsistent, the dominant hand presses the button '2'.

[0086] S105. In this embodiment, a computer-automated 2-back working memory paradigm task is adopted. The first 2 rounds of stimuli are reference stimuli. Starting from the 3rd round, the subject needs to compare the currently displayed number with the previous 2 numbers in memory, determine whether the content of the currently displayed stimulus is consistent with that of the previous 2 numbers in memory, and record the reaction time of the subject and whether the judgment is accurate so as to calculate the correct rate. The computer automatically records the display time of the Arabic numeral and the time when the subject presses the button, calculates the reaction time of the subject each time, and at the same time records whether the subject's judgment is correct in each round so as to calculate the correct rate, and saves it as a txt file.

[0087] S2. Denoise each original pulse wave signal, and the specific steps are as follows:

[0088] As Figure 2 is the flowchart for denoising the original pulse wave signal.

[0089] S201. First, define the objective function and initialize the population of the genetic optimization algorithm. Use the mean square error in the signal reconstruction error as the objective function to evaluate its fitness, and optimize the key parameters in the variational mode decomposition - the number of modes K and the penalty factor α for the collected original pulse wave signal by the genetic optimization algorithm;

[0090] Among them, the definition of the mean square error MSE is as follows:

[0091] In the formula, N is the signal length; s(t i ) is the value of the original signal at the i-th sampling point; is the value of the reconstructed signal at the i-th sampling point;

[0092] Among them, the population size of the initialized genetic optimization algorithm is 20, the crossover probability is 0.8, the mutation probability is 0.2, the number of iterations is 25, and the number of optimizations is 2, corresponding to two optimization parameters, the modal number K and the penalty factor α;

[0093] S202. Repeat step S201 to obtain the optimal parameter combination in variational mode decomposition, and perform the variational mode decomposition process on the collected original pulse wave signal to obtain several IMF components;

[0094] S203. Perform fast Fourier transform on the K IMF components obtained in step S202 to calculate their central frequencies, and screen out M high-frequency IMF components with higher central frequencies according to the central frequencies; Since the modal number K is optimized by the genetic optimization algorithm for each pulse wave signal, different pulse wave signals may be divided into different K IMF components, and the fast Fourier transform calculates the central frequencies of the K IMF components. Because the normal heart rate is usually above 0.66 Hz, a frequency threshold of 0.66 Hz is set to screen out M high-frequency IMF components greater than the frequency threshold;

[0095] S204. Perform wavelet threshold denoising on the M high-frequency IMF components selected in step S203; Among them, in the wavelet threshold denoising algorithm, the wavelet basis function uses "db9", the number of wavelet decomposition layers uses 4 layers, the threshold function uses the soft threshold function, and the threshold uses the global threshold; Among them, the global threshold formula is the general threshold formula:

[0096] In the formula, λ represents the threshold, σ represents the standard deviation of the noise, and N represents the signal length;

[0097] S205. Linearly superimpose the M high-frequency IMF components processed in step S204 to reconstruct the preliminarily preprocessed pulse wave signal;

[0098] S206. Filter the preliminarily preprocessed pulse wave signal obtained in step S205 using a Butterworth low-pass filter. Under normal circumstances, the heart rate of normal people is between 60 and 100 beats per minute. Considering the error, the heart rate range is expanded to 40 to 180 beats, corresponding to a frequency band of 0.66 to 3 Hz. In order to retain the morphological information in the pulse wave, the working frequency band of the Butterworth low-pass filter is set to 0 to 7 Hz, the filter order is set to 3, and filtering is performed to obtain the pulse wave signal;

[0099] S207. Detect and extract the starting point from the pulse wave signal that has undergone the filtering process in step S206 to obtain the starting point. Calculate the difference between two adjacent elements of the filtered pulse wave signal to obtain the difference sequence of the pulse wave signal, which is similar to the first derivative function. Use peak detection on the difference sequence to find the peak points. The peak points of the difference sequence correspond to the maximum rising points of the pulse wave signal. Search forward from the maximum rising points to find the starting point, and identify the starting point through the multi-scale automatic peak detection method;

[0100] S208. For the successfully located starting point, obtain the baseline drift curve through cubic spline interpolation, thereby achieving baseline drift removal, and finally obtaining the preprocessed pulse wave signal. This step aims to extract the pulse wave base point through feature point detection, and use the cubic spline interpolation method to fit the pulse wave base point to achieve baseline drift removal and obtain the baseline drift curve; further, linearly subtract the baseline drift curve from the pulse wave signal to remove the baseline drift noise of the pulse wave signal;

[0101] S209. Segment each pulse cycle through the starting point of each pulse cycle obtained in step S207. Use the Pearson correlation coefficient to judge the signal quality of the current pulse wave cycle, select the one with the highest Pearson correlation coefficient as the cycle template, calculate the Pearson correlation coefficient between the cycle template and the remaining cycles, set the Pearson correlation coefficient threshold to 0.8, and the cycles with a coefficient lower than 0.8 are determined to be abnormal cycles and are excluded;

[0102] S210. Detect and extract feature points from the pulse wave signal that has undergone the filtering process in step S308 to obtain peak points and dicrotic wave points.

[0103] S3. For the preprocessed pulse wave signal, extract the heart rate, heart rate variability, blood oxygen saturation, and pulse wave morphological features of the subject. The specific steps are as follows:

[0104] As Figure 3 is the flowchart of the pulse wave feature extraction process.

[0105] S301. From the peak points of the pulse wave signal obtained in step S2, extract the times corresponding to all the peak points in the pulse wave signal, calculate the average heart rate mean_hr, and use it for the extraction of heart rate variability; among them, the average heart rate calculation formula:

[0106]

[0107] Among them, represents the average value of the time intervals between all adjacent peak points of the pulse wave signal;

[0108] S302. Based on the times corresponding to all the peak points in the pulse wave signal obtained in step S301, perform a difference calculation on the time series corresponding to the peak points to obtain the time interval series of the peak points of the pulse wave signal, that is, the heart rate interval series. Calculate the physiological characteristics of heart rate variability through the heart rate interval series of the pulse wave signal, including time domain characteristics and frequency domain characteristics. The time domain characteristics include the root mean square of the differences between adjacent heart rate intervals RMSSD, the standard deviation of the differences between adjacent heart rate intervals SDSD, the standard deviation of normal heart rate intervals SDNN, the proportion pNN50 of the differences between adjacent heart rate intervals exceeding 50 ms, the proportion pNN20 of the differences between adjacent heart rate intervals exceeding 20 ms, and the triangular index of heart rate variability triangular_index; the frequency domain characteristics include the total power of the signal TP, the very low frequency band power VLF, the low frequency power LF, the high frequency power HF, the normalized low frequency power LFnu, and the normalized high frequency power HFnu; among them, the Lomb-Scargle method is used to calculate the power spectral density (PSD) estimate of the heart rate interval series, and the frequency domain characteristics are calculated through the PSD.

[0109] S303. Through the data of the red light channel and the infrared light channel of the preprocessed pulse wave signal, combine the empirical formula of blood oxygen saturation containing unknown parameters, and perform a linear fit with the measured data of the standard finger clip oximeter, so as to estimate the unknown parameters in the empirical formula of blood oxygen saturation to obtain a complete empirical formula of blood oxygen saturation, and obtain the blood oxygen saturation SpO through the empirical formula of blood oxygen saturation. 2 。

[0110] First, calculate the AC / DC ratio of the signals of the red light channel and the infrared light channel. AC / DC is the ratio of the alternating current component (AC) to the direct current component (DC):

[0111] In the formula, AC red and DC red respectively represent the alternating current component and the direct current component of the red light; AC ir and DC ir represent the alternating current component and the direct current component of the infrared light.

[0112] Empirical formula of blood oxygen saturation: SpO 2 = A·R + B

[0113] The specific empirical formula of blood oxygen saturation in this embodiment is: SpO 2 = -3.47301·R + 101.18149

[0114] The constants A and B are obtained by performing a linear fit with the measured data of the standard finger clip oximeter. Subsequently, the blood oxygen saturation can be directly calculated through the empirical formula of blood oxygen saturation;

[0115] S304. Extract morphological features from each waveform using the feature points extracted in step S2, including the rising time T1, falling time T2, reflection time T3, systolic peak amplitude HA, and dicrotic wave peak amplitude HB. The rising time refers to the time interval between the starting point of the pulse wave and the systolic peak, the falling time is the time interval from the systolic peak to the end point, and the reflection time is the time interval from the systolic peak to the dicrotic wave peak.

[0116] S4. Analyze the distribution of the individual working memory task response times recorded in step S1, construct a distribution model of the response times, and use the Kolmogorov-Smirnov test to evaluate the goodness of fit of the established model to the actual data, thereby establishing a distribution function of the response times; combine the distribution characteristics of the response times with the accuracy rate of the working memory task to quantitatively score the subsequent individual working memory level. The specific steps are as follows:

[0117] S401. Statistically analyze all the response time data recorded in step S1 and calculate the parameters of the theoretical distribution of the response time data. Among them, use the maximum likelihood estimation method to calculate the parameters of various probability distribution models of the response time, such as normal distribution, gamma distribution, exponential distribution, Weibull distribution, lognormal distribution, etc.

[0118] S402. Calculate the cumulative distribution function of the theoretical distribution and use the Kolmogorov-Smirnov test to evaluate the goodness of fit of the theoretical distribution of the data; evaluate the goodness of fit of the constructed response time distribution model to the actual data through the Kolmogorov-Smirnov test. At the same time, use the Akaike information criterion (AIC) and Bayesian information criterion (BIC) to further evaluate the adaptability of the distribution model.

[0119] Among them, in this embodiment 1, the response time data follows a Weibull distribution, and the probability density distribution function is:

[0120]

[0121] where PDF is the Weibull distribution probability density distribution function;

[0122] The cumulative distribution function is:

[0123] where CDF is the Weibull distribution cumulative distribution function;

[0124] S403. Through the cumulative distribution function of the theoretical distribution obtained in step S402, combine the accuracy rate and response time in the N-back trials of the working memory task paradigm in step S1 of the subject, input the response time in step S1 into the cumulative distribution function to obtain a response time score, and weight the response time and combine it with the accuracy rate to obtain a training reference score.

[0125] S5. Conduct a piano silent-play working memory training cycle on a number of subjects, repeat steps S1 - S3 to obtain a physiological feature dataset characterized by heart rate, heart rate variability, blood oxygen saturation, and morphological features, performance features in piano silent-play working memory training, and a distribution model fitted by the N-back response time of the working memory task paradigm. Standardize the physiological feature dataset. Using the training reference score as the target variable, through a random forest model, combine the working memory physiological feature dataset and the performance features in individual piano silent-play working memory training to construct a working memory evaluation model to evaluate individual working memory and the effect of working memory training. The specific steps are as follows:

[0126] S501. Conduct piano silent-play working memory training and the working memory task paradigm N-back on a number of subjects, as in step S1;

[0127] S502. Repeat steps S2 - S3 to obtain a physiological feature dataset characterized by heart rate, heart rate variability, blood oxygen saturation, and morphological features, and performance features in piano silent-play working memory training;

[0128] S503. Standardize the physiological feature dataset and combine it with the performance features in individual piano silent-play working memory training to construct a model dataset that corresponds one-to-one with the corresponding training reference score;

[0129] S504. Perform feature screening on the model dataset, remove redundant features, and construct an optimal dataset; Feature screening uses recursive feature elimination;

[0130] S505. Through a random forest regression model, construct a working memory evaluation model, combine cross-validation to verify the regression model error, and achieve the mapping of the training reference score from physiological features and performance features in piano silent-play working memory training, so as to evaluate individual working memory and the effect of working memory training;

[0131] Among them, the error is the root mean square error RMSE, and the calculation formula is as follows:

[0132] In the formula, n is the number of samples, y i is the actual value of the i-th sample, is the predicted value of the model;

[0133] The root mean square error RMSE in Example 1 is 3.5184;

[0134] Figure 4 This is the scatter plot predicted by the working memory evaluation model in the embodiment of the present invention.

[0135] S6. Input the piano behavior data and physiological data of individual piano silent practice training into the working memory evaluation model in step S5 to obtain a score, thereby evaluating the training effect of working memory. The specific steps are as follows:

[0136] S601. The individual conducts piano silent practice training to obtain corresponding pulse wave signals and piano silent performance characteristics, without the need to perform the working memory task paradigm N-back;

[0137] S602. Obtain physiological characteristics such as heart rate, heart rate variability, blood oxygen saturation, and pulse wave morphology characteristics through steps S2 - S3;

[0138] S603. Input the physiological characteristics and piano silent performance characteristics into the working memory evaluation model to obtain a score, thereby visually evaluating the individual's working memory and the training effect of working memory through the change in the score.

[0139] In summary, using the pulse wave physiological characteristics and piano silent performance characteristics as feature variables, and the training reference score as the target variable, a data set is established and a random forest regression model is trained, thereby constructing a working memory evaluation model for subsequent evaluation of the working memory and the training effect of working memory.

[0140] Embodiment 2

[0141] Figure 1 It is a flowchart of a method for integrated quantification evaluation of working memory measurement and training in piano silent practice provided by an embodiment of the present invention. This embodiment is based on Embodiment 1. Through the working memory evaluation model of Embodiment 1, it is possible to realize the measurement of working memory and the intuitive display of the training effect only through the PPG signal and piano silent practice, without the traditional working memory paradigm N-back. The individual conducts 4 times of piano silent practice training to obtain corresponding pulse wave signals and piano silent performance characteristics, and evaluates the working memory and the training effect of working memory. The specific steps are as follows:

[0142] S1. Collect the pulse wave signals of the subject during sitting still and during piano silent practice and record performance characteristics such as the playing accuracy in the individual piano silent practice training, without the need to perform the working memory task paradigm N-back;

[0143] S2. Refer to the corresponding steps in Embodiment 1, which will not be elaborated here. Among them, for one segment of the pulse wave signal, the optimal parameter combination of the genetic optimization algorithm is the number of modes 7 and the penalty factor 512.0868417 as shown in Table 1; The IMF component diagram is as Figure 5 ; The IMF components and their spectrograms and central frequencies are as Figure 6 , The reconstructed pulse wave signal is as Figure 7 ; The signal after Butterworth filtering is as Figure 8 ; The signal after removing baseline drift is as Figure 9 ;

[0144] Table 1. Optimal parameter combination table of the genetic optimization algorithm in Example 2

[0145]

[0146] S3. Refer to the corresponding steps in Example 1, which will not be elaborated here;

[0147] S4. Step S4 does not need to be carried out;

[0148] S5. Step S5 does not need to be carried out;

[0149] S6. Refer to the corresponding steps in Example 1, which will not be elaborated here; among them, the working memory and the training effect of the working memory are shown in Table 1, and the training effect of the individual working memory is visually evaluated through the change of the score;

[0150] Table 2. Score table of the training effect of piano silent performance in Example 2

[0151]

[0152] In summary, according to the random forest working memory evaluation model, the score of the training effect of each piano silent performance is obtained, so as to realize the quantitative evaluation of the intuitive working memory and the training effect of the working memory

[0153] The technical features of the above embodiments can be combined arbitrarily. For the sake of brief description, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered as the scope described in this specification.

[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 other changes, modifications, substitutions, combinations, and simplifications made without departing from the spirit and principle of the present invention shall be equivalent replacement methods and are all included in the protection scope of the present invention.

Claims

1. A method for quantitatively evaluating the integrated testing and training of working memory for silent piano playing, characterized in that: The identification method comprises the following steps: S1. Construct a test-training integrated scenario, collect the pulse wave signals of the subjects during meditation and piano silent playing, and record the performance characteristics of the playing accuracy in the individual piano silent playing training. After completing a round of piano silent playing training, conduct a round of working memory task paradigm N-back to record reaction time and accuracy; A smart piano and a photoelectric volumetric pulse wave (PPG) acquisition device are placed in the test scene, and the subject places the middle finger in the PPG acquisition device to obtain the pulse wave signal; the smart piano refers to a modified piano that can record the performance characteristics of the individual's silent piano playing, including the accuracy of playing and the playing time; the reaction time refers to the time between the appearance of the stimulus and the subject's response in the working memory task, and the accuracy refers to the accuracy of judging whether the current stimulus is consistent with the stimulus content of the previous N rounds after memorizing N stimuli, where N≤3; S2. Denoise each original pulse wave signal; use an optimization algorithm combining a genetic optimization algorithm and a variational mode decomposition to decompose the pulse wave signal into several intrinsic mode function signals, hereinafter referred to as IMF, calculate the center frequency of each IMF signal, perform wavelet threshold filtering according to the center frequency, thereby obtaining a filtered pulse wave signal, and then combine a Butterworth low-pass filter to remove noise and baseline, and remove the period interfered by motion artifacts through adaptive period screening to obtain a pre-processed pulse wave signal, and finally perform feature point detection and extraction; S3. Extract the heart rate, heart rate variability, blood oxygen saturation and morphological features of the subject from the preprocessed pulse wave signal; wherein the heart rate and heart rate variability can be further calculated by extracting the peak-to-peak time interval of the pulse wave, the blood oxygen saturation is obtained by formula calculation and fitting of the pulse wave detected by the infrared light and red light channels collected by the PPG acquisition device, and the morphological features are obtained by extracting features from the characteristic points of each pulse wave cycle; construct a working memory physiological feature data set based on the physiological features obtained above; S4, performing mathematical statistical distribution analysis on the individual working memory task reaction time recorded in step S1, constructing a distribution model of the reaction time and using the Kolmogorov-Smirnov test to establish a distribution function of the reaction time, and combining the accuracy of the working memory task to quantify the reference score after the piano silent playing training in step S1 through the distribution model; S5. Perform a piano silent working memory training cycle on several subjects, repeat steps S1 to S3, obtain a physiological characteristic data set characterized by heart rate, heart rate variability, blood oxygen saturation and morphological characteristics, and a distribution model fitted by the performance characteristics in the piano silent working memory training and the N-back reaction time of the working memory task paradigm, standardize the physiological characteristic data set, use the training reference score as the target variable, and use a random forest model to combine the working memory physiological characteristic data set and the performance characteristics in the individual piano silent working memory training to construct a working memory evaluation model to evaluate the individual working memory and working memory training effect; S6. Input the piano behavior data and physiological data of the individual piano silent playing training into the working memory evaluation model of step S5 to obtain a score, thereby evaluating the working memory and the working memory training effect.

2. The method for quantitatively evaluating the working memory of piano playing according to claim 1, characterized in that: The process of step S1 is as follows: S101, under the condition of sufficient light and no interference, the pulse wave signal of the subject when sitting quietly and performing silent piano playing training is collected by a PPG collection device placed on the side of one hand of the subject; S102, displaying a random note sequence of a certain length on the screen for multiple times, with a certain time limit; in the first round, the subject needs to memorize the note sequence of the current round, and starting from the second round of testing, the subject needs to play the music score of the previous round and memorize the note sequence of the current round, and the result of each note played by the subject is recorded during the silent piano training, so as to calculate the playing accuracy of the subject; S103, under quiet external conditions, the subjects performed the working memory task paradigm N-back; S104, displaying the stimulus content of the working memory task paradigm N-back on the screen, and the subject needs to compare the stimulus currently displayed with the previous N stimuli in memory to determine whether the content of the stimulus currently displayed is consistent with the previous N stimuli in memory; S105. The first N rounds of stimulation are reference stimulation. Starting from the N+1th round, the subject needs to compare the stimulation currently displayed with the first N stimulations in memory to determine whether the content of the stimulation currently displayed is consistent with that of the first N stimulations in memory, and record the subject's reaction time and whether the judgment is accurate to calculate the accuracy.

3. The method for integrated quantitative evaluation of working memory testing and training for silent piano playing according to claim 1, characterized in that: The process of step S2 is as follows: S201, first define the objective function and initialize the population of the genetic optimization algorithm, use the objective function to evaluate the fitness of the population, and optimize the collected original pulse wave signal through the genetic optimization algorithm to obtain the mode number K and penalty factor α in the variational mode decomposition; S202, repeating step S201 to obtain the optimal parameter combination in variational modal decomposition, performing variational modal decomposition on the collected original pulse wave signal to obtain a plurality of IMF components; S203, performing fast Fourier transform on the several IMF components obtained in step S202 to calculate their center frequencies, and screening out M high-frequency IMF components with higher center frequencies according to the center frequencies, where M≤K; S204, performing wavelet threshold denoising processing on the M high-frequency IMF components selected in step S203; S205, linearly superimposing the M high-frequency IMF components processed in step S204 to reconstruct a pulse wave signal after preliminary preprocessing; S206, filtering the pulse wave signal obtained after preliminary preprocessing in step S205 using a Butterworth low-pass filter; S207, performing pulse wave starting point detection and extraction on the pulse wave signal filtered in step S206 to obtain the starting point of each pulse cycle; S208, obtaining a baseline drift curve by using cubic spline interpolation for the pulse wave starting point that has been successfully located, thereby eliminating the baseline drift; S209, segmenting each pulse cycle according to the starting point of each pulse cycle obtained in step S207, using the Pearson correlation coefficient to judge the signal quality of the current pulse wave cycle, selecting the cycle template with the highest Pearson correlation coefficient as the cycle template, and using the cycle template and the remaining cycles to perform Pearson correlation coefficient to screen out abnormal cycles and remove them; S210, performing feature point detection and extraction on the pulse wave signal after feature screening in step S209 to obtain peak points and dicrotic points.

4. The method for quantitatively evaluating the working memory of piano playing according to claim 3, characterized in that: The process of step S3 is as follows: S301, obtaining the peak point of the pulse wave signal from step S2, extracting the time corresponding to all the peak points in the pulse wave signal, calculating the average heart rate mean_hr, and using it for extracting heart rate variability; S302, according to the time corresponding to all the peak points in the pulse wave signal obtained in step S301, perform differential calculation on the time series corresponding to the peak points to obtain the time interval sequence of the peak points of the pulse wave signal, that is, the heartbeat time interval sequence; calculate the physiological characteristics of heart rate variability through the heartbeat time interval sequence of the pulse wave signal, including time domain characteristics and frequency domain characteristics, and the time domain characteristics include the root mean square RMSSD of the difference between adjacent heartbeat intervals, the standard deviation SDSD of the difference between adjacent heartbeat intervals, the standard deviation SDNN of normal heartbeat intervals, the proportion pNN50 of the difference between adjacent heartbeat intervals exceeding 50ms, the proportion pNN20 of the difference between adjacent heartbeat intervals exceeding 20ms, and the triangular index triangular_index of heart rate variability; Frequency domain features include total signal power TP, very low frequency power VLF, low frequency (0.04-0.15Hz) power LF, high frequency (0.15-0.40Hz) power HF, normalized low frequency power LFnu and normalized high frequency power HFnu. The very low frequency corresponds to a frequency range of 0.00-0.04Hz, and the low frequency corresponds to a frequency range of 0.04-0.15Hz. S303, using the collected red light channel and infrared light channel data of the pulse wave signal, combined with the blood oxygen saturation empirical formula containing unknown parameters, and performing linear fitting with the measured data of the standard finger-clip oximeter, thereby estimating the unknown parameters in the blood oxygen saturation empirical formula to obtain a complete blood oxygen saturation empirical formula, and obtaining the blood oxygen saturation SpO2 through the blood oxygen saturation empirical formula; S304, extracting morphological features of each waveform through the feature points extracted in step S2, including rising branch time T1, falling branch time T2, reflection time T3, contraction peak amplitude HA, and dicrotic wave peak amplitude HB.

5. The method for quantitatively evaluating the working memory of silent piano playing according to claim 1, characterized in that: The process of step S4 is as follows: S401, counting all the reaction time data recorded in step S1, and calculating the parameters of the theoretical distribution of the reaction time data; S402, calculating the cumulative distribution function of the theoretical distribution, and using the Kolmogorov-Smirnov test to test the degree of fit of the data to the theoretical distribution; S403. The cumulative distribution function of the theoretical distribution obtained in step S402 is combined with the accuracy and reaction time of the subject in the N-back trial of the working memory task paradigm in step S1, and the reaction time in step S1 is input into the cumulative distribution function to obtain the reaction time score. The reaction time is weighted and combined with the accuracy to obtain the training reference score.

6. The method for quantitatively evaluating the working memory of silent piano playing according to claim 1, characterized in that: The process of step S5 is as follows: S501, conducting piano silent working memory training and working memory task paradigm N-back on several subjects, as in step S1; S502, repeating steps S2 to S3, obtaining a physiological characteristic data set characterized by heart rate, heart rate variability, blood oxygen saturation and morphological characteristics, and performance characteristics in piano silent working memory training; S503, standardizing the physiological characteristic data set and combining it with the performance of the individual in the silent piano working memory training to construct a model data set that corresponds to the corresponding training reference score; S504, screening features from the model data set, removing redundant features, and constructing an optimal data set; S505. A working memory evaluation model is constructed through a random forest regression model. Combined with cross-validation, the regression model error is verified to achieve the mapping of the training reference score from the physiological characteristics and the performance characteristics in the piano silent working memory training, thereby evaluating the individual working memory and the working memory training effect.

7. The method for quantitatively evaluating the working memory of silent piano playing according to claim 1, characterized in that: The process of step S6 is as follows: S601, the individual performs silent piano training to obtain the corresponding pulse wave signal and silent piano performance characteristics, without the need to perform the working memory task paradigm N-back; S602, obtaining physiological characteristics of heart rate, heart rate variability, blood oxygen saturation and pulse wave morphology characteristics through steps S2 to S3; S603, inputting the physiological characteristics and the silent piano performance characteristics into the working memory evaluation model to obtain a score, thereby intuitively evaluating the individual working memory and the working memory training effect through the score changes.

Images