Quantitative evaluation method for spatial cognitive ability training effect of piano visual playing

Abstract

The invention discloses a quantitative evaluation method for the training effect of piano visual playing on spatial cognitive competence, and the method comprises the following steps: building an objective spatial cognitive scoring system which is used for measuring the reference data of the spatial cognitive competence of a user; based on a wireless pulse wave acquisition platform, pulse wave data of a user is measured, preprocessed and denoised, and multiple physiological indexes are extracted; the method comprises the following steps: synchronously recording user pulse wave physiological data and spatial cognitive ability scores evaluated by a spatial cognitive scoring system before and after piano visual playing training of a plurality of users, recording visual playing accuracy in a visual playing process, and constructing a data set; and training a regression model based on the data set, fitting a spatial cognitive competence score difference value, and constructing a formula for evaluating a training effect according to the feature weight. According to the invention, an objective and comfortable method can be provided for evaluating the training effect of piano visual playing on spatial cognitive competence.

Description

Technical Field

[0001] The present invention belongs to the technical fields of information processing and intelligent health, and particularly relates to a quantitative evaluation method for the training effect of piano sight-reading on spatial cognitive ability. Background Art

[0002] Spatial cognitive ability is one of the indispensable cognitive abilities for humans to carry out daily activities, solve problems, and perform complex tasks. It involves how to perceive spatial layout, object position, spatial relationship, and the changes of objects in space. This ability plays a crucial role in daily life and various cognitive activities. Piano sight-reading training, as a highly concentrated cognitive activity, requires individuals to process a large amount of visual, auditory, and motor information in a short time.

[0003] In existing research, the evaluation methods for the training effect of piano sight-reading on spatial cognitive ability can be divided into two categories, namely, behavioral task assessment method and physiological index assessment method. Behavioral task assessment methods such as mental rotation task, spatial navigation task, etc., judge the spatial cognitive ability and its changes of the subjects by testing their performance under specific tasks. These methods are intuitive and easy to operate, but there are still certain subjectivity and limitations. Physiological index assessment method, because it is difficult for people to subjectively control physiological reactions, and there is a close correlation between physiological indexes and spatial cognitive ability, so the physiological index assessment method has good objectivity. At present, physiological index assessment methods such as electroencephalogram (EEG), electrocardiogram (ECG), skin conductance, etc., the measured data and images are relatively objective, but these methods are costly and rely on professional equipment, and it is difficult to be widely applied. Existing research has shown that the pulse wave signal has a large amount of cardiovascular information, and sufficient information can be extracted to evaluate spatial cognitive ability and its changes. The heart rate variability extracted based on the pulse wave signal can be used as a substitute for heart rate variability, and the morphological characteristics of the pulse wave signal can reflect physiological indexes such as vascular elasticity and hardness. At the same time, the pulse wave signal can be collected non-invasively, and the collection site is only one finger, which is simple and convenient, and has strong practicability.

[0004] Currently, most of the research on the training effect of piano sight-reading on spatial cognitive ability focuses on the significant analysis of training intensity and behavioral task performance, ignoring the importance of piano sight-reading data (accuracy, rhythm mastery, etc.), and there are subjective limitations in behavioral task methods. There are also studies using methods combining heart rate variability, electroencephalogram signals and task performance data to evaluate the training effect, but there are various limitations in the sensitivity, reliability and usability of ECG and EEG measurement methods. Summary of the Invention

[0005] The object of the present invention is to improve the subjectivity problem of evaluating the training effect of piano sight-reading on spatial cognitive ability relying on cognitive task data. By combining piano sight-reading performance and pulse wave physiological data, a regression model for predicting the difference in spatial cognitive ability scores before and after is constructed, and a quantitative evaluation method for the training effect of piano sight-reading on spatial cognitive ability is provided.

[0006] To achieve the above object, the present invention adopts the following technical solutions: A quantitative evaluation method for the training effect of piano sight-reading on spatial cognitive ability, the quantitative evaluation method comprising the following steps: S1. Establish a spatial cognitive scoring system for measuring the baseline data of the user's spatial cognitive ability and calculating the user's spatial cognitive score; wherein, the spatial cognitive scoring system gives questions with different rotation angles and respectively records the user's answering time t and correctness d; the user operates the spatial cognitive scoring system, and obtains the user's spatial cognitive score according to the rotation angle, the user's answering time and correctness. S2. Through a pulse wave acquisition device, when the user operates the spatial cognitive scoring system in step S1, synchronously measure the user's pulse wave for 5 minutes to obtain the user's pulse wave signal; preprocess and denoise the pulse wave signal; perform per-cycle feature point recognition on the processed pulse wave signal, and on this basis, extract the feature point sequence of each cycle of the pulse wave signal, wherein the feature points include the pulse start point and the main wave peak point of the pulse wave signal; calculate the morphological features of each cycle according to the morphological feature points, including the main wave amplitude A, the rising branch time T, the waveform area S, and the rising branch area S1; calculate the time domain indexes of each morphological feature according to the morphological feature sequence respectively; through peak point recognition, obtain the peak point interval RR1 sequence and calculate the heart rate variability index; finally obtain the pulse wave morphological features and heart rate variability features. S3. Before and after several users' piano sight-reading, by executing steps S1 to S2, synchronously record the user's pulse wave signal and the spatial cognitive ability score based on the scoring system, obtain the difference in the spatial cognitive ability scores before and after the user's piano sight-reading, the pulse wave morphological features and the heart rate variability physiological indexes; record the sight-reading accuracy during the piano sight-reading process; construct a data set for training the XGBoost regression model. S4. Train the XGBoost regression model based on the data set, fit the difference in spatial cognitive ability scores, obtain the regression coefficients of each feature, and construct an evaluation formula for evaluating the training effect according to the regression coefficients of each feature, and obtain the quantitative evaluation score of the user through the evaluation formula.

[0007] Further, the process of step S1 is as follows: S101. Under sufficient lighting conditions, the user sits in front of the computer screen and uses the right hand to control the keyboard to input answers and answer the test questions of the spatial cognitive ability scoring system; S102. Display a number of test questions on the screen, randomly show two different characters. The displayed characters are obtained by rotating the original character R, or by rotating the original character R after mirror flipping. The rotation angles are 0 degrees, 60 degrees, 120 degrees, 180 degrees or 240 degrees. The rotation angles of the two characters are both random, reducing the repetition rate of the test questions; S103. After the test questions are given, the user answers and discriminates whether the two different characters are in a mirror relationship or a rotation relationship; if character 1 can be obtained by rotating a certain angle to get character 2, then the two characters are in a rotation relationship; if character 1 needs to be mirror-operated in addition to rotation to get character 2, then it is a mirror relationship. If both the mirror relationship and the rotation relationship exist, it is determined as the mirror relationship; S104. Record the time taken by the user to answer each test question , whether each test question is answered correctly and the rotation angle difference between the two characters of each test question , , this rotation angle difference defines the minimum rotation amount required from the rotation angle of character 1 to the rotation angle of character 2; S105. According to the user's answer data, fit the distribution law of the answer data and calculate the user's spatial cognitive ability score.

[0008] The specific process of step S105 is as follows: S1051. Fit the data distribution of the answering times of a number of users , select multiple distribution types, and then evaluate the goodness of fit through the K-S test, and select the one with the highest goodness of fit as the data distribution of the answering time; S1052. According to the data distribution type of the answering time , perform Z-score standardization or log standardization on to eliminate the bias and scale difference in the data. If is a right-skewed distribution, perform log standardization on it, otherwise perform Z-score standardization; S1053. Since the score is assigned on a 100-point scale, perform "Max-min" normalization on the standardized data to scale the data to the specified interval [0, 100] for subsequent score calculation; S1054. Add the rotation angle as a difficulty factor to the score calculation formula. Considering that there is a linear relationship between the answering time and the rotation angle, use the principal component analysis method to reduce the dimensions of the two features and merge them into a comprehensive feature , the steps are as follows: Construct a matrix containing the answering time and the rotation angle of the data matrix : ; Calculate the response time and the rotation angle of the covariance matrix. The covariance matrix reflects the linear relationship between two variables: ; Solve for the eigenvalues and eigenvectors . The eigenvalues reflect the variance of the data in the corresponding direction, and the eigenvectors represent the direction of data variation, which is also the direction of the principal components: ; After calculating all the eigenvalues, sort them in descending order. The data direction corresponding to the largest eigenvalue is the direction with the greatest variability and importance in the data. Select the first principal components, that is, the first eigenvectors and eigenvalues, where

[0009] is the number of principal components to be retained. The eigenvalues of the retained principal components need to exceed the mean of all eigenvalues. Finally, calculate the reduced data matrix: where is the original data matrix, is the matrix composed of the first eigenvectors; Finally, based on the comprehensive features of the response time and the rotation angle and the response correctness , construct a formula for calculating the spatial cognitive ability score

[0010] where represents the number of response questions, is the comprehensive feature of the rotation angle and the response time for the th question, is the response correctness for the th question, 100 for correct and 0 for wrong, and and are the weights of the comprehensive feature and the correctness in the spatial cognitive ability score respectively, and both are greater than 0 and add up to 1, and are set through experimental results.

[0011] Furthermore, the process of step S2 is as follows: S201. When the user operates the spatial cognition scoring system and answers questions for spatial cognition ability scoring, synchronously collect the pulse wave signals of the left index finger to ensure data synchronization; S202. Perform wavelet threshold denoising on the collected pulse waves to remove high-frequency noise while retaining the main features of the signals; Considering the support length, regularity, and similarity to the pulse wave signal waveform, the selected wavelet basis function is "db8"; S203. Use a finite impulse response filter to perform band-pass filtering on the signals to further improve the signal quality; S204. Remove the baseline drift from the filtered signals to filter out the baseline drift noise; The process is as follows: Use the cubic spline interpolation method to fit the starting position of the pulse wave signal to obtain the baseline drift curve, and subtract the baseline drift curve from the filtered signal to obtain the pulse wave signal with baseline drift noise removed; S205. Identify the characteristic points of the preprocessed and denoised signals, including the starting point of the pulse wave cycle and the pulse wave peak, to obtain the characteristic point sequence of the entire signal segment; Specifically, use a sliding window combined with the first-order difference method to identify the pulse wave peak and the starting point; S206. Traverse the characteristic point sequence and calculate the area enclosed by each cycle waveform by integration. Denote the number of signal cycles as , and obtain 4 morphological feature sequences with a length of ; These 4 morphological feature sequences respectively correspond to the main wave amplitude , , , of each cycle; , the rising branch time , the waveform area under , and the rising branch area ; S207. Respectively extract the respective time-domain indexes of the 4 morphological feature sequences through statistical mathematical operations to obtain the time-domain index vectors, including the mean of the feature sequence , the standard deviation of the feature sequence , the root mean square of the first-order difference of the feature sequence , and the standard deviation of the first-order difference of the feature sequence . Combine the 4 time-domain index vectors to obtain the time-domain feature vector ; Among them, each time-domain index vector is:

[0012]

[0013]

[0014]

[0015] The above avg() represents the average value calculation; std() represents the standard deviation calculation; drms() represents the calculation of the root mean square of the first-order difference; dstd() represents the calculation of the root mean square of the first-order difference; And the time-domain feature vector is ; S208. Differentiate the time between adjacent peaks of the pulse wave to obtain the pulse beat interval sequence, and use the cubic spline interpolation method to interpolate this beat interval sequence to obtain the pulse variability signal; the original signal obtained by differentiation in this step belongs to a non-uniformly sampled signal, and extracting the frequency-domain information of pulse variability requires the signal to be uniformly sampled. Therefore, the cubic spline interpolation method is used to interpolate the signal to obtain a uniformly sampled signal; S209. Calculate the heart rate variability index based on the pulse wave variability signal; among them, the heart rate variability index is divided into time-domain index, frequency-domain index, and non-linear index; the time-domain index is calculated by mathematical statistics, and the indexes include standard deviation of normal-to-normal intervals SDNN, root mean square of differences between adjacent normal-to-normal intervals RMSSD, standard deviation of differences between adjacent normal-to-normal intervals SDSD, mean of adjacent normal-to-normal intervals Mean, minimum heart rate Min_HR, average heart rate Mean_HR, standard deviation of heart rate STD_HR; the frequency-domain index is calculated by power spectral density, and the indexes include low-frequency band power LF, high-frequency band power HF, ratio of low-frequency band power to high-frequency band power LF / HF; the non-linear index is calculated by mathematical statistics, including short-term variability SD1, long-term variability SD2, and Hurst index.

[0016] Furthermore, the process of step S3 is as follows: S301. Conduct piano sight-reading training for several users. Before and after the training, execute steps S1 to S2, and synchronously record the physiological data of the users' pulse waves and the spatial cognitive ability scores based on the scoring system to obtain the difference in the users' spatial cognitive ability scores, the morphological characteristics of the pulse wave, and the physiological indexes of heart rate variability; during the piano sight-reading process, calculate and record the sight-reading accuracy according to the distance between the required fingering and the actual fingering; initially construct the first data set for training the XGBoost regression model; Here, fingering refers to pressing a certain key with a certain finger. The actual key number pressed is output through the key scanning module integrated in the piano, and the actual finger pressed is output through the photographic equipment and the finger recognition algorithm. The sight-reading accuracy refers to the fitting degree of the playing distance , and calculate the fitting degree of the playing distance according to the distance between the required fingering and the actual fingering , and the calculation formula is as follows:

[0017] Among them, is the key number of the th note required by this round of fingering, is the key number actually pressed for the th time, is the finger number corresponding to the th note required by this round of fingering, is the finger number actually pressed for the th time; S302. Use the Pearson correlation coefficient and the Spearman rank correlation coefficient to perform a correlation analysis on the spatial cognition score and physiological characteristics, screen out the characteristics with a strong relationship with the spatial cognition score from the first dataset, and construct a second dataset.

[0018] Further, the process of step S4 is as follows: S401. Standardize the physiological characteristics and spatial cognition score values in the second dataset, and divide the second dataset into a training set and a test set; among them, the training set is used to construct a regression model for predicting the difference in spatial cognition ability scores, and the test set is used to verify the model performance indicators; S402. Train the XGBoost regression model with the training set, and use the test set to evaluate the performance indicators of the XGBoost regression model. The performance indicators used for evaluation include the mean square error MSE, the root mean square error RMSE, and the coefficient of determination; S403. The XGBoost regression model is a machine learning model based on gradients for classification or regression purposes. It predicts the difference in spatial cognition scores according to the input features, and at the same time outputs the regression coefficients of each feature, that is, the feature importance. The regression coefficients can be used as the source of variable weights in constructing the scoring formula. Based on the regression coefficients of each feature, a evaluation formula for evaluating the training effect is constructed, and the quantitative evaluation score of the user is obtained through the evaluation formula.

[0019] The expression of the evaluation formula is as follows:

[0020] Among them, is the fitted difference in spatial cognition ability scores, is used as the quantitative evaluation score, Mean is the mean of adjacent heart rate intervals, SD1 is the short-term variability, STD_HR is the standard deviation of heart rate, is the sight-reading accuracy, LF / HF is the low-frequency to high-frequency power ratio, Min_HR is the minimum heart rate, and the weights of each index are derived from the regression coefficients of each feature output by the XGBoost regression model.

[0021] Compared with the prior art, the present invention has the following advantages and beneficial effects: (1) Based on photoplethysmogram and piano sight-reading performance, this invention evaluates the training effect of piano sight-reading on spatial cognitive ability, achieving millisecond-level synchronous sampling of physiological signals and behavioral data. Compared with the traditional ECG / EEG evaluation method that requires complex operations such as patch electrodes or EEG caps, photoplethysmogram only needs single-point contact to complete signal acquisition. At the same time, the acquisition of sight-reading data only requires the contact output of the piano keyboard. The data collected by this method has real-time and strong reliability.

[0022] (2) This invention proposes an objective spatial cognitive ability scoring system for constructing actual value data of spatial cognitive ability. Aiming at the non-linear interference in the traditional rotation angle-time scoring, distribution fitting is innovatively introduced to probabilistically model the response time, effectively eliminating the interference of outliers. At the same time, principal component analysis is used to orthogonalize the original features to eliminate the feature multicollinearity. Finally, an evaluation model is constructed through XGBoost regression, and an evaluation formula is constructed based on its regression coefficients. The quantitative evaluation score of the user is obtained through the evaluation formula.

[0023] (3) This invention extracts a variety of physiological features of the pulse wave, including heart rate variability features and several pulse wave morphology features, aiming to comprehensively capture the physiological signal information related to spatial cognitive ability. In the heart rate variability analysis, cubic spline interpolation is innovatively used to fit the baseline to eliminate baseline drift, and a variety of time-frequency domain features are extracted. At the same time, aiming at the limitation of only analyzing the heart rate variability of the pulse wave in the traditional method, this invention introduces the pulse wave morphology features and uses them together with the heart rate variability features as physiological indicators. By extracting a variety of pulse wave features simultaneously, the deviation that may be brought by a single physiological index is avoided, thus improving the accuracy and robustness of the training effect evaluation. Description of the Drawings

[0024] 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.

[0025] Figure 1 is a flowchart of a method for quantitatively evaluating the training effect of piano sight-reading on spatial cognitive ability disclosed in this invention; Figure 2 is a flowchart of the acquisition and calculation of the spatial cognitive ability score in Embodiment 1 of this invention; Figure 3 is a flowchart of preprocessing and denoising the pulse wave and extracting physiological indicators in Embodiment 1 of this invention; Figure 4 is a flowchart of collecting sight-reading performance data and constructing a dataset in Embodiment 1 of this invention; Figure 5 It is the flowchart for establishing the evaluation training effect formula in Embodiment 1 of the present invention; Figure 6 It is the reference diagram of the spatial cognitive ability test questions in Embodiment 2 of the present invention; Figure 7 It is the effect diagram of pulse wave denoising in Embodiment 2 of the present invention; Figure 8 It is the schematic diagram of the pulse wave feature extraction points in Embodiment 2 of the present invention; Figure 9 It is the schematic diagram of finger numbering in Embodiment 2 of the present invention; Figure 10 It is the reference diagram of the fingering spectrum in Embodiment 2 of the present invention. Detailed implementation manners

[0026] In order 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.

[0027] Referring to "embodiment" in this application means that the specific features, structures or characteristics described in combination with the embodiment can be included in at least one embodiment of this application. The appearance of this phrase in various positions in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. Those skilled in the art explicitly and implicitly understand that the embodiments described in this application can be combined with other embodiments.

[0028] Embodiment 1 Figure 1 It is the flowchart of a method for quantitatively evaluating the training effect of piano sight-reading on spatial cognitive ability provided by the embodiment of the present invention. The judgment benchmark data for spatial cognitive ability in this embodiment is obtained by establishing an objective spatial cognitive ability scoring system, and it is applicable to the quantitative evaluation of the training effect of piano sight-reading on spatial cognitive ability. The specific steps are as follows: S1. Establish a spatial cognitive scoring system for measuring the benchmark data of the user's spatial cognitive ability and calculating the user's spatial cognitive score; among them, the spatial cognitive scoring system gives test questions with different rotation angles and respectively records the user's answering time and correctness ; the user operates the spatial cognitive scoring system, and obtains the user's spatial cognitive score based on the rotation angle, the user's answering time and correctness. The specific steps are as follows: Such asFigure 2 Flow chart for collecting and calculating spatial cognitive ability scores.

[0029] S101. Under the environmental condition of sufficient light, the user sits in front of the computer screen and uses the right hand to control the keyboard to input answers to answer the test questions of the spatial cognitive ability scoring system. S102. Give a number of test questions on the screen, randomly display two different characters, and the displayed characters are obtained by rotating the original characters or the original characters are obtained by rotating after mirror flipping, and the rotation angles are 0 degrees, 60 degrees, 120 degrees, 180 degrees or 240 degrees. The rotation angles of the two characters are both random, reducing the repetition rate of the test questions. S103. After the test questions are given, the user answers and distinguishes whether the two different characters are in a mirror relationship or a rotation relationship; if character 1 can be obtained by rotating a certain angle to get character 2, then the two characters are in a rotation relationship; if character 1 needs to be mirror-operated in addition to the rotation operation to get character 2, then it is a mirror relationship. S104. Record the answering time of each test question by the user , whether each test question is answered correctly and the rotation angle difference between the two characters of each test question , , and this rotation angle difference defines the minimum rotation amount required from the rotation angle of character 1 to the rotation angle of character 2. The calculation formula is as follows:

[0030] where , are the rotation angles of the two characters respectively.

[0031] S105. Calculate the user's spatial cognitive ability score according to the user's answering data.

[0032] The specific process of step S105 is as follows: S1051. Perform data distribution fitting on the answering times of a number of users , select multiple distribution types, and then evaluate the goodness of fit through the K-S test, and select the one with the highest goodness of fit as the data distribution of the answering time. S1052. According to the data distribution type of the answering time , perform Z-score standardization processing or log standardization processing on to eliminate the bias and scale difference in the data. If is a right-skewed distribution, then perform log standardization processing on it, otherwise perform Z-score standardization processing. The formulas for the two processing methods are as follows:

[0033]

[0034] Among them, is the original data The value after standardization.

[0035] S1053. Since the final score is out of 100, continue to perform "Max-min" normalization on the data to scale the data to the specified interval [0, 100]. The formula is as follows: ; Similarly, perform "Max-min" normalization on the rotation angle to scale the data to the specified interval [0, 100].

[0036] S1054. Add the rotation angle as a difficulty factor to the score calculation formula. Considering that there is a linear relationship between the answering time and the rotation angle, use the principal component analysis method to reduce the dimensionality of the two features and merge them into a comprehensive feature , the steps are as follows: Construct a data matrix containing the answering time and the rotation angle : : ; Calculate the covariance matrix of the answering time and the rotation angle . The covariance matrix reflects the linear relationship between the two variables: ; Solve for the eigenvalues and eigenvectors . The eigenvalue reflects the variance of the data in the corresponding direction, and the eigenvector represents the direction of data variation, which is also the direction of the principal component: ; After calculating all the eigenvalues, sort them in descending order. The data direction corresponding to the largest eigenvalue is the direction with the greatest variability and importance in the data. Select the first principal components, that is, the first eigenvectors and eigenvalues, is the number of principal components to be retained. The eigenvalues of the retained principal components need to exceed the mean of all eigenvalues. Finally, calculate the data matrix after dimensionality reduction:

[0037] Among them, is the original data matrix, is the matrix composed of the first eigenvectors; Finally, based on the comprehensive features of response time and rotation angle and response correctness , a spatial cognitive ability score calculation formula is constructed as follows:

[0038] where the number of representative responses, is the comprehensive feature of the rotation angle and response time for the th question, is the response correctness for the th question, 100 for correct and 0 for incorrect, and and are the weights of the comprehensive feature and correctness in the spatial cognitive ability score respectively, and both are greater than 0 and add up to 1, and are set through experimental results.

[0039] S2. When the user operates the spatial cognitive score system in step S1, a 5-minute pulse wave measurement is synchronously performed on the user through a pulse wave acquisition device to obtain the user's pulse wave signal; the pulse wave signal is preprocessed and denoised; feature point recognition is performed on the processed pulse wave signal cycle by cycle, and on this basis, the feature point sequence of each cycle of the pulse wave signal is extracted, where the feature points include the pulse start point and the main wave peak point of the pulse wave signal; the morphological features of each cycle are calculated according to the morphological feature points, including the main wave amplitude , rising branch time , waveform area under the curve , rising branch area ; the time domain indexes of each morphological feature are calculated according to the morphological feature sequence; through peak point recognition, the peak point interval sequence is obtained, and the heart rate variability index is calculated; finally, the pulse wave morphological features and heart rate variability features are obtained; the specific steps are as follows: As Figure 3 is the flowchart for preprocessing and denoising the pulse wave and extracting physiological indexes.

[0040] S201. When the user operates the spatial cognitive score system to answer questions for spatial cognitive ability score, the pulse wave signal of the left index finger is synchronously collected to ensure data synchronization; S202. The collected pulse wave is processed by wavelet threshold denoising to remove high-frequency noise. Considering the support length, regularity, and similarity with the pulse wave signal waveform, the selected wavelet basis function is "db8"; Among them, the steps of wavelet threshold denoising are as follows: 1) Decomposition: Convolve the signal with the scaling function and the mother wavelet function to obtain the approximation coefficients (low-frequency part) and detail coefficients (high-frequency part) of the signal. The convolution formula for the decomposition process is as follows:

[0041]

[0042] Among them, and are the low-frequency part and high-frequency part of the first layer respectively. is the version of the pulse wave signal after downsampling, while and are the coefficients of the scaling function and the coefficients of the mother wavelet function respectively. The relationship between the wavelet function and the scaling function is as follows:

[0043] Among them, is the high-pass filter coefficient, which is related to the scaling function.

[0044] 2) Downsampling: In each decomposition step, the filtered signal is downsampled to reduce the time resolution of the signal and extract the low-frequency part. Downsampling is performed on both the approximation coefficients and the high-frequency coefficients.

[0045] 3) Recursive decomposition: After obtaining the approximation coefficients of the first layer, continue to decompose the approximation coefficients to obtain the approximation coefficients and detail coefficients of the second layer, and so on. The sampling rate of the pulse wave acquisition device used is 100 Hz, so the maximum frequency of the extracted pulse wave signal is 50 Hz. Considering that the frequency range of the human heart rate signal is 0.7 - 3 Hz, the number of wavelet decomposition layers is set to 4 layers. Through wavelet decomposition, the pulse wave signal is decomposed into a series of detail coefficients and approximation coefficients :

[0046] 4) Threshold processing: At each scale , perform threshold processing on the high-frequency part of the signal, that is, the detail coefficients . The selected threshold function is the hard threshold function:

[0047] Among them, is the threshold, and the selection method of is the threshold selection VisuShrink based on noise estimation. This method is a global unified threshold :

[0048] Among them, is the standard deviation of the noise, is the length of the signal, which can be calculated by the following formula:

[0049] where is the median of the absolute values of the detail coefficients of the first-level wavelet decomposition.

[0050] 5) Perform inverse wavelet transform (IDWT) on the processed wavelet coefficients to reconstruct the denoised signal , and the reconstruction process is the opposite of the decomposition process. The reconstruction process includes upsampling and filtering. Upsampling performs zero interpolation on the signal, and filtering selects the corresponding reconstruction filter. The formula for the reconstruction process is as follows:

[0051] where and respectively represent the tap coefficient sequences of the low-pass and high-pass filters corresponding to the selected wavelet function.

[0052] S203. Perform band-pass filtering on the signal using a finite impulse response filter with a lower cut-off frequency of 0.4 Hz and an upper cut-off frequency of 5 Hz; S204. Use cubic spline interpolation to fit the starting position of the pulse wave signal to obtain the baseline drift curve, and subtract the baseline drift curve from the filtered signal to obtain a further denoised pulse wave signal; S205. Identify the feature points of the preprocessed and denoised signal, and use a sliding window combined with the first-order difference method to identify the pulse wave peaks and starting points; S206. Traverse the feature point sequence and integrate to calculate the area enclosed by each cycle waveform. Denote the number of signal cycles as , and obtain 4 morphological feature sequences with a length of ; these 4 morphological feature sequences respectively correspond to the main wave amplitude , , , ; the rising branch time , the waveform area under , the rising branch area ; ; S207. Respectively extract the respective time-domain indicators of the 4 morphological feature sequences through statistical mathematical operations to obtain a time-domain indicator vector, including the mean of the feature sequence , Standard deviation of the feature sequence , Root mean square of the first-order difference of the feature sequence , Standard deviation of the first-order difference of the feature sequence , Combine the four time-domain index vectors to obtain a time-domain feature vector ; S208. Differentiate the time between adjacent peaks of the pulse wave to obtain the pulse beat interval sequence, and use the cubic spline interpolation method to interpolate this beat interval sequence to obtain the pulse variability signal; the original signal obtained by differentiation in this step belongs to a non-uniformly sampled signal, and extracting the frequency-domain information of pulse variability requires the signal to be uniformly sampled. Therefore, use the cubic spline interpolation method to interpolate the signal to obtain a uniformly sampled signal; S209. Calculate the heart rate variability index according to the pulse wave variability signal; among them, the heart rate variability index is divided into time-domain index, frequency-domain index and non-linear index; the time-domain index is calculated by mathematical statistics, and the indexes include standard deviation of the R-R interval SDNN, root mean square of the difference between adjacent R-R intervals RMSSD, standard deviation of the difference between adjacent R-R intervals SDSD, mean of adjacent R-R intervals Mean, minimum heart rate Min_HR, average heart rate Mean_HR, standard deviation of heart rate STD_HR; the frequency-domain index is calculated by power spectral density, and the indexes include low-frequency band power LF, high-frequency band power HF, low-frequency band to high-frequency band power ratio LF / HF; the non-linear index is calculated by mathematical statistics, including short-term variability SD1, long-term variability SD2, Hurst index.

[0053] S3. Before and after the piano sight-reading of several users, by performing steps S1 - S2, synchronously record the user's pulse wave signal and the spatial cognitive ability score based on the scoring system, and obtain the difference in the spatial cognitive ability score, the pulse wave morphological characteristics and the heart rate variability physiological indexes before and after the user's piano sight-reading; record the sight-reading accuracy during the piano sight-reading process; construct a dataset for XGBoost regression model training, and the specific steps are as follows: Such as Figure 4 is the flowchart for collecting sight-reading performance data and constructing a dataset.

[0054] S301. Conduct piano sight-reading training for several users. Before and after the training, perform steps S1 - S2, synchronously record the user's pulse wave physiological data and the spatial cognitive ability score based on the scoring system, and obtain the difference in the user's spatial cognitive ability score, the pulse wave morphological characteristics and the heart rate variability physiological indexes; during the piano sight-reading process, calculate and record the sight-reading accuracy according to the distance between the required fingering and the actual fingering; initially construct the first dataset for XGBoost regression model training; S302. Use the Pearson correlation coefficient and the Spearman rank correlation coefficient to perform a correlation analysis on the spatial cognition score and physiological characteristics, screen out the characteristics with a strong relationship with the spatial cognition score from the first dataset, and construct a second dataset.

[0055] S4. Train an XGBoost regression model based on the dataset, fit the difference in spatial cognition ability scores, obtain the regression coefficients of each feature, and construct an evaluation formula for evaluating the training effect based on the regression coefficients of each feature. Obtain the quantitative evaluation score of the user through the evaluation formula. The specific steps are as follows: Such as Figure 5 Is the flowchart for establishing the evaluation training effect formula.

[0056] S401. Standardize the physiological characteristics and spatial cognition score values in the second dataset, and divide the second dataset into a training set and a test set. Among them, the training set is used to construct a regression model for predicting the difference in spatial cognition ability scores, and the test set is used to verify the model performance indicators. S402. Train the XGBoost regression model with the training set, and use the test set to evaluate the performance indicators of the XGBoost regression model. The performance indicators used for evaluation include the mean squared error MSE, the root mean squared error RMSE, and Coefficient of determination; The XGBoost regression model is a machine learning model based on gradients for classification or regression purposes. It predicts the difference in spatial cognition scores according to the input features, and at the same time outputs the regression coefficients of each feature, that is, the feature importance. The regression coefficients can be used as the source of variable weights in the construction of the scoring formula. Construct an evaluation formula for evaluating the training effect based on the regression coefficients of each feature, and obtain the quantitative evaluation score of the user through the evaluation formula.

[0057] In summary, construct a dataset based on the physiological characteristics of the pulse wave, the accuracy of piano sight-reading, and the spatial cognition score, and train an XGBoost regression model to fit the difference in spatial cognition ability scores. Finally, use the regression coefficients of the regression model as the weights of the variables in the evaluation formula to construct an evaluation formula for evaluating the training effect.

[0058] Embodiment 2 Figure 1 Is the flowchart of a method for quantitatively evaluating the training effect of piano sight-reading on spatial cognition ability provided by an embodiment of the present invention. Taking the accuracy of piano sight-reading and the physiological feature vector extracted from the pulse wave before and after sight-reading as reference indicators, the specific steps are as follows: S1. Establish a spatial cognition score system for measuring the benchmark data of the user's spatial cognition ability and calculating the user's spatial cognition score. Among them, the spatial cognition score system gives questions with different rotation angles and record the user's answering time respectively and correctness ; A user operation spatial cognition scoring system calculates the user's spatial cognition score based on the rotation angle, the user's answering time, and correctness. The specific steps are as follows: S101. Under the environmental condition of sufficient light, the user sits in front of the computer screen and uses the right hand to control two keys on the keyboard to answer the given test questions. The two keys respectively represent yes or no for the answer. Among them, the given test questions are as Figure 6 shown; S102. Refer to the corresponding steps in Embodiment 1, which will not be elaborated here; S103. Refer to the corresponding steps in Embodiment 1, which will not be elaborated here; S104. Refer to the corresponding steps in Embodiment 1, which will not be elaborated here; S2. Refer to the corresponding steps in Embodiment 1, which will not be elaborated here. Among them, the left hand measures the pulse wave, and the right hand is used to answer the spatial cognition ability test questions in step S1. The pulse wave filtering effect is as Figure 7 shown, and the schematic diagram for extracting the morphological feature points of the pulse wave is as Figure 8 shown. On this basis, the morphological features of each cycle of the user's pulse wave signal are extracted.

[0059] S3. Before and after several users' piano sight-reading, by executing steps S1 to S2, synchronously record the user's pulse wave signal and the spatial cognition ability score based on the scoring system, and obtain the difference in the spatial cognition ability scores, the pulse wave morphological features, and the heart rate variability physiological indexes before and after the users' piano sight-reading; record the sight-reading accuracy during the piano sight-reading process; construct a dataset for XGBoost regression model training. The specific steps are as follows: S301. Conduct piano sight-reading training for several users. Before and after the training, execute steps S1 to S2, synchronously record the user's pulse wave physiological data and the spatial cognition ability score based on the scoring system, and obtain the difference in the user's spatial cognition ability scores, the pulse wave morphological features, and the heart rate variability physiological indexes; during the piano sight-reading process, calculate and record the piano sight-reading accuracy according to the distance between the required fingering and the actual fingering; initially construct the first dataset for XGBoost regression model training; In this embodiment, the schematic diagram for finger numbering is as Figure 9 shown. The required fingering is obtained based on the given fingering score sheet, and the fingering score sheet is as Figure 10 shown. The actual fingering is detected by a photographic device. The difference in the numbers of the required fingering and the actual fingering is the distance between the required fingering and the actual fingering.

[0060] S302. Refer to the corresponding steps in Embodiment 1, which will not be elaborated here.

[0061] S4. On the basis of Embodiment 1, the regression coefficients of the regression model are used as the weights of the variable values of the prediction formula, and the regression coefficients of each index are shown in Table 1.

[0062] Table 1. Regression Coefficient Table of Each Feature in Embodiment 2

[0063] In summary, according to the regression coefficients, an evaluation formula is constructed to obtain an evaluation formula for the training effect of piano sight-reading on spatial cognitive ability. This evaluation formula combines pulse wave physiological data and sight-reading behavior data to achieve a quantitative evaluation of the training effect.

[0064] The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity of 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 recorded in this specification.

[0065] 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 quantitative evaluation method for the training effect of piano sight-reading on spatial cognitive ability, characterized in that, The quantization evaluation method includes the following steps: S1. Establish a spatial cognition scoring system to measure the baseline data of the user's spatial cognition ability and calculate the user's spatial cognition score. Among them, the spatial cognition scoring system presents questions with different rotation angles and respectively records the user's answering time t and correctness d. The user operates the spatial cognition scoring system, and obtains the user's spatial cognition score based on the rotation angle, the user's answering time, and the correctness. S2. When the user operates the spatial cognitive ability scoring system in step S1, use a pulse wave acquisition device to synchronously measure the user's pulse wave for 5 minutes to obtain the user's pulse wave signal; preprocess and denoise the pulse wave signal; perform per-cycle feature point recognition on the processed pulse wave signal, and on this basis, extract the feature point sequence of each cycle of the pulse wave signal. Among them, the feature points include the pulse starting point and the main wave peak point of the pulse wave signal; calculate the morphological features of each cycle according to the morphological feature points, including the main wave amplitude A, the rising branch time T, the waveform area S, and the rising branch area S1; calculate the time-domain indexes of each morphological feature according to the morphological feature sequence; obtain the peak point interval RR1 sequence through peak point recognition and calculate the heart rate variability index; finally, obtain the pulse wave morphological features and heart rate variability features; S3. Before and after several users' piano sight-reading, by executing steps S1 to S2, synchronously record the users' pulse wave signals and the spatial cognitive ability scores based on the scoring system, obtain the difference in spatial cognitive ability scores, pulse wave morphological features, and heart rate variability physiological indexes before and after the users' piano sight-reading; record the sight-reading accuracy during the piano sight-reading process; construct a dataset for XGBoost regression model training; S4. Train the XGBoost regression model based on the dataset, fit the difference in spatial cognitive ability scores, obtain the regression coefficients of each feature, and construct an evaluation formula for evaluating the training effect based on the regression coefficients of each feature. Obtain the quantization evaluation score of the user through the evaluation formula.

2. The quantitative evaluation method for the training effect of spatial cognitive ability by sight-reading of piano according to claim 1, wherein The process of step S1 is as follows: S101. Under the environmental condition of sufficient light, the user sits in front of the computer screen and uses the right hand to control the keyboard to input answers to answer the test questions of the spatial cognitive ability scoring system; S102. Present a number of test questions on the screen, randomly display two different characters, and the displayed characters are obtained by rotating the original characters or obtained by rotating the original characters after mirror flipping and then rotation, and the rotation angles are 0 degrees, 60 degrees, 120 degrees, 180 degrees or 240 degrees; S103. After the test questions are given, the user answers and distinguishes whether two different characters are in a mirror image relationship or a rotation relationship; S104. Record the time taken by the user to answer each question , whether each question answer is correct and the difference in the rotation angles of two characters for each question , ; S105. Calculate the user's spatial cognitive ability score according to the user's answer data.

3. The quantitative evaluation method for the training effect of piano sight-reading on spatial cognitive ability according to claim 2, characterized in that, The process of step S105 is as follows: S1051. Fit the data distribution of the response times of several users Select multiple distribution types, then evaluate the goodness of fit through the K-S test, and select the one with the highest goodness of fit as the data distribution of the response time; S1052. According to the response time For the data distribution type, perform Z-score standardization or log standardization to eliminate the bias and scale differences in the data. If it is a right-skewed distribution, perform log standardization on it; otherwise, perform Z-score standardization. S1053. Perform "Max-min" normalization processing on the standardized data to scale the data to the specified interval [0, 100]; S1054. Considering that there is a linear relationship between the response time and the rotation angle, the two features are dimensionally reduced by the principal component analysis method and merged into a comprehensive feature , and the steps are as follows: The structure includes the response time and the rotation angle data matrix : ; Calculated response time and rotation angle of the covariance matrix, which reflects the linear relationship between two variables: ； Solve for eigenvalues and eigenvectors , eigenvalues reflect the variance of the data in the corresponding direction, and eigenvectors represent the direction of data variation, which is also the direction of the principal component: ; After calculating all the eigenvalues, arrange them in descending order. The data direction corresponding to the largest eigenvalue is the direction with the greatest variability and importance in the data. Select the first principal components, that is, the first eigenvectors and eigenvalues. is the number of principal components to be retained. Finally, calculate the data matrix after dimensionality reduction: Among them, is the original data matrix, is the matrix composed of the first eigenvectors; Finally, based on the comprehensive features of response time and rotation angle and response correctness , a spatial cognitive ability score calculation formula is constructed as follows: Among them, is the comprehensive feature of the rotation angle and the answering time of the question, is the answering correctness of the question, 100 for correct and 0 for wrong, while and are the weights of the comprehensive feature and the correctness in the spatial cognitive ability score respectively, and the weights are initially set through a large amount of experimental data.

4. The quantitative evaluation method for the training effect of piano sight-reading on spatial cognitive ability according to claim 1, characterized in that The process of step S2 is as follows: S201. When the user operates the spatial cognitive ability scoring system and answers the spatial cognitive ability score, synchronously collect the pulse wave signal of the left index finger; S202. Perform wavelet threshold denoising processing on the collected pulse wave to remove high-frequency noise while retaining the main features of the signal; S203. Use a finite impulse response filter to perform band-pass filtering on the signal; S204. Remove the baseline drift of the filtered signal to filter out the baseline drift noise; S205. Perform feature point recognition on the preprocessed and denoised signal, including the pulse wave cycle starting point and the pulse wave peak, to obtain the feature point sequence of the entire signal; S206. Extract each morphological feature from the pulse wave signal for each period according to the characteristic point sequence, and denote the number of signal periods as , obtaining four morphological feature sequences with a length of ; , , , ; These four morphological feature sequences respectively correspond to the main wave amplitude , the rising branch time , the waveform area under the curve , and the rising branch area ; S207. Respectively extract the respective time-domain indicators for the 4 morphological feature sequences to obtain a time-domain indicator vector, including the mean of the feature sequence , the standard deviation of the feature sequence , the root mean square of the first-order difference of the feature sequence , the standard deviation of the first-order difference of the feature sequence . Combine the 4 time-domain indicator vectors to obtain a time-domain feature vector ; S208. Differentiate the time between adjacent peaks of the pulse wave to obtain the pulse beat interval sequence, and use the cubic spline interpolation method to interpolate this beat interval sequence to obtain the pulse variability signal; S209. Calculate the heart rate variability index based on the pulse wave variability signal. The heart rate variability index includes the standard deviation of the R-R interval (SDNN), the root mean square of the differences between adjacent R-R intervals (RMSSD), the standard deviation of the differences between adjacent R-R intervals (SDSD), the mean of adjacent R-R intervals (Mean), the minimum heart rate (Min_HR), the average heart rate (Mean_HR), the standard deviation of the heart rate (STD_HR), the low-frequency power (LF), the high-frequency power (HF), the ratio of low-frequency power to high-frequency power (LF / HF), the short-term variability (SD1), the long-term variability (SD2), and the Hurst exponent.

5. The quantitative evaluation method for the training effect of spatial cognitive ability by sight-reading the piano according to claim 1, characterized in that, The process of step S3 is as follows: S301. Conduct piano sight-reading training for several users. Before and after the training, execute steps S1 - S2, synchronously record the users' pulse wave physiological data and the scores of spatial cognitive ability based on the scoring system, to obtain the difference in the scores of the users' spatial cognitive ability, the morphological characteristics of the pulse wave, and the physiological indicators of heart rate variability. During the piano sight-reading process, calculate and record the accuracy of piano sight-reading according to the distance between the required fingering and the actual fingering. Initially construct the first dataset for training the XGBoost regression model. S302. Use the Pearson correlation coefficient and the Spearman rank correlation coefficient to analyze the correlation between the spatial cognitive scores and the physiological characteristics, screen out the characteristics with a strong relationship with the spatial cognitive scores from the first dataset, and construct the second dataset.

6. The quantitative evaluation method for the training effect of piano sight-reading on spatial cognitive ability according to claim 1, characterized in that, The process of step S4 is as follows: S401. Standardize the physiological characteristics and the spatial cognitive score values in the second dataset, and divide the second dataset into a training set and a test set. The training set is used to construct a regression model for predicting the difference in the scores of spatial cognitive ability, and the test set is used to verify the performance indicators of the model. S402. Train the XGBoost regression model with the training set, and use the test set to evaluate the performance metrics of the XGBoost regression model. The performance metrics used for evaluation include the mean squared error MSE, the root mean squared error RMSE, and the 2 coefficient of determination; S403. According to the XGBoost regression model, output the regression coefficients of each feature, construct an evaluation formula for evaluating the training effect based on the regression coefficients of each feature, and obtain the quantitative evaluation score of the user through the evaluation formula.

Images