Piano playing posture dynamic correction method and system based on multi-dimensional data

Abstract

The invention discloses a piano playing posture dynamic correction method and system based on multi-dimensional data, and relates to the technical field of posture correction, and the method comprises the steps: collecting the multi-dimensional data of a player in real time through vision, pressure and electromyographic signal sensors, and generating an original data set; performing cleaning, noise reduction and feature extraction on the original data set, generating a preprocessed data set, and extracting biomechanical feature parameters; based on the extracted biomechanical characteristic parameters, a biomechanical model reflecting the posture state and the potential deviation of the player is constructed, and through multi-dimensional data acquisition and fusion, deep learning model analysis and a closed-loop optimization system, the posture state and the potential deviation of the player can be analyzed. The problems that an existing piano playing posture correction method cannot dynamically monitor in real time, accurately recognize posture deviation, provide personalized correction suggestions and continuously optimize the correction effect are solved.

Description

Technical Field

[0001] The present invention relates to the technical field of posture correction, and in particular to a method and system for dynamic correction of piano playing posture based on multi-dimensional data. Background Art

[0002] Piano posture correction technology has continued to advance with the development of sensors and artificial intelligence. Early on, it relied primarily on manual observation and guidance from teachers. Later, systems based on single video image analysis emerged. More recently, these technologies have gradually integrated multiple sensors and simple machine learning algorithms. However, these technologies are still relatively nascent and have many limitations.

[0003] Existing technologies have significant deficiencies in multidimensional data fusion, real-time performance, personalized correction, and intelligence. They often rely on single visual data, lacking multimodal data fusion such as pressure and electromyographic signals, making it impossible to comprehensively and accurately assess posture. Data processing is subject to delays, making it impossible to meet the needs of real-time correction for piano playing. Furthermore, based on fixed rules or simple algorithms, they are unable to adapt to individual differences among players, nor can they dynamically adjust and optimize, making it difficult to achieve precise and personalized posture correction. Furthermore, they cannot achieve the precise, real-time, personalized dynamic correction effect achieved by the present invention. Summary of the Invention

[0004] In view of the above existing problems, the present invention is proposed.

[0005] Therefore, the present invention provides a dynamic correction method for piano playing posture based on multi-dimensional data to solve the problems of being unable to perform real-time dynamic monitoring and comprehensive analysis of playing posture, being unable to accurately identify the type and degree of posture deviation, and how to provide performers with personalized correction suggestions synchronized with the playing rhythm.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0007] In a first aspect, the present invention provides a method for dynamic correction of piano playing posture based on multi-dimensional data, which is characterized by comprising:

[0008] The multi-dimensional data of the performer is collected in real time through visual, pressure and electromyographic signal sensors to generate the original data set;

[0009] Clean, reduce noise and extract features from the original data set to generate a preprocessed data set and extract biomechanical feature parameters;

[0010] Based on the extracted biomechanical characteristic parameters, a biomechanical model reflecting the player's posture state and potential deviations is constructed;

[0011] Use deep learning models to analyze biomechanical models in real time to generate assessment results that identify the type and degree of posture deviation;

[0012] Combining the evaluation results with the performance rhythm information, it generates correction suggestions that are synchronized with the performance rhythm and reminds the performer to adjust his posture through multimodal feedback.

[0013] Feedback data and correction effect data are fed back to the deep learning model to optimize model parameters and form a closed-loop optimization system.

[0014] As a preferred solution of the method for dynamic correction of piano playing posture based on multi-dimensional data of the present invention, wherein: the multi-dimensional data of the player is collected in real time by visual, pressure and electromyographic signal sensors to generate an original data set, the specific steps are as follows:

[0015] Installing miniature visual sensors, pressure sensors, and myoelectric signal sensors on key parts of the pianist's arms, fingers, spine, etc.

[0016] Connect each sensor to the data processing terminal via Wi-Fi, Bluetooth or 4G for calibration and synchronization;

[0017] Start the sensor and collect visual, pressure and electromyographic signal data in real time at the set sampling rate and frame rate, store them in the specified location of the terminal, and form the original data set.

[0018] As a preferred solution of the method for dynamic correction of piano playing posture based on multi-dimensional data of the present invention, the steps of cleaning, denoising and feature extraction of the original data set, generating a preprocessed data set and extracting biomechanical characteristic parameters are as follows:

[0019] Check the original data set and remove missing values, erroneous values ​​and outliers;

[0020] Apply filtering algorithms to reduce noise on visual, pressure and electromyographic signal data respectively;

[0021] Feature extraction algorithms are used to extract key features from the denoised data, such as body posture features, force change features, and muscle activity features, to generate a preprocessed data set and extract biomechanical feature parameters.

[0022] As a preferred solution of the method for dynamic correction of piano playing posture based on multi-dimensional data of the present invention, wherein: based on the extracted biomechanical characteristic parameters, a biomechanical model reflecting the player's posture state and potential deviation is constructed, the specific steps are:

[0023] Initialize the biomechanical model according to the principles of human biomechanics and define the model structure and parameters;

[0024] Mapping biomechanical characteristic parameters to the model, adjusting and optimizing the model parameters to accurately reflect the player's posture;

[0025] Analyze posture data, identify potential deviations, build deviation models in the model, and describe and quantify the type and degree of posture deviation.

[0026] As a preferred solution of the piano playing posture dynamic correction method based on multi-dimensional data of the present invention, wherein: the use of a deep learning model to perform real-time analysis on the biomechanical model to generate an identification result of the posture deviation type and degree assessment, the specific steps are:

[0027] Collect a large amount of posture data, including normal posture and various deviation posture data, to train deep learning models;

[0028] The data of the biomechanical model is fed into the trained deep learning model for real-time analysis;

[0029] Based on the model output, an assessment result of the type and degree of the performer's posture deviation is generated to provide a basis for correction recommendations.

[0030] As a preferred embodiment of the method for dynamic correction of piano playing posture based on multi-dimensional data of the present invention, wherein: the method combines the evaluation results and the playing rhythm information to generate correction suggestions synchronized with the playing rhythm and reminds the player to adjust the posture through multimodal feedback, the specific steps are as follows:

[0031] Analyze piano playing rhythm in real time to extract rhythm features and changing patterns;

[0032] Generate correction suggestions that match the playing rhythm based on the posture deviation assessment results and the playing rhythm information;

[0033] Corrective suggestions are conveyed to the performer through multimodal feedback such as visual cues, auditory warnings, and tactile vibrations.

[0034] As a preferred solution of the piano playing posture dynamic correction method based on multi-dimensional data described in the present invention, wherein: the feedback data and correction effect data are fed back to the deep learning model, the model parameters are optimized, and a closed-loop optimization system is formed. The specific steps are:

[0035] Collect the performers' feedback on the correction suggestions and the effect of the corrected posture, organize and mark them;

[0036] Feed the data back to the deep learning model, and update and adjust the model parameters through the model's self-learning optimization mechanism;

[0037] Evaluate the performance of the closed-loop optimization system and further optimize each part of the system based on the results to improve the effectiveness and efficiency of posture correction.

[0038] In a second aspect, the present invention provides a piano playing posture dynamic correction system based on multi-dimensional data, comprising:

[0039] Data acquisition module: collects multi-dimensional data of the performer in real time through visual, pressure and electromyographic signal sensors to generate original data sets;

[0040] Data preprocessing module: cleans, reduces noise and extracts features from the original data set, generates a preprocessed data set and extracts biomechanical characteristic parameters;

[0041] Model building module: Based on the extracted biomechanical characteristic parameters, a biomechanical model reflecting the player's posture state and potential deviations is constructed;

[0042] Analysis and evaluation module: Uses deep learning models to analyze biomechanical models in real time and generates assessment results to identify the type and degree of posture deviation;

[0043] Feedback Correction Module: This module combines the evaluation results with the performance rhythm information to generate correction suggestions that are synchronized with the performance rhythm and uses multimodal feedback to remind the performer to adjust their posture.

[0044] System optimization module: Feedback data and correction effect data are fed back to the deep learning model to optimize model parameters and form a closed-loop optimization system.

[0045] In a third aspect, the present invention provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program is executed by the processor, any step of the method for dynamic correction of piano playing posture based on multi-dimensional data as described in the first aspect of the present invention is implemented.

[0046] In a fourth aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any step of the method for dynamic correction of piano playing posture based on multi-dimensional data as described in the first aspect of the present invention.

[0047] The beneficial effects of the present invention are as follows: through multi-dimensional data collection and fusion, comprehensive and accurate monitoring and analysis of piano playing posture is achieved, the type and degree of posture deviation can be identified in real time, and correction suggestions synchronized with the playing rhythm are generated. Multimodal feedback is used to promptly remind the player to adjust their posture, effectively improving the real-time, accuracy, and pertinence of piano playing posture correction. At the same time, the closed-loop optimization system can continuously optimize the parameters of the deep learning model based on feedback data and correction effects, allowing the system to continuously adapt to the individual differences of different players and achieve personalized correction. The correction effect becomes more accurate and efficient as the usage time increases, which is of significant significance for improving the standardization of players' posture and preventing performance-related injuries, bringing innovative solutions to the field of piano playing posture correction. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0049] Figure 1 This is a flow chart of the method for dynamic correction of piano playing posture based on multi-dimensional data in Example 1. DETAILED DESCRIPTION

[0050] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0051] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0052] Secondly, the term "one embodiment" or "embodiment" herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in various places throughout this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive of other embodiments.

[0053] Example 1, with reference to Figure 1 , which is the first embodiment of the present invention, provides a method for dynamic correction of piano playing posture based on multi-dimensional data, comprising the following steps:

[0054] S1. Collect multi-dimensional data of the performer in real time through visual, pressure and electromyographic signal sensors to generate the original data set.

[0055] Specifically, micro visual sensors, pressure sensors and electromyographic signal sensors are installed on key parts of the pianist, such as the arms, fingers, and spine.

[0056] It should be noted that micro-visual sensors, pressure sensors, and myoelectric signal sensors are installed in key areas of the pianist, such as the arms, fingers, and spine. Specifically, the instrument first precisely measures and locates key areas of the pianist's body to determine the optimal sensor locations, ensuring accurate capture of posture changes and muscle activity during performance. Medical-grade non-woven tape or elastic fabric bands are then used to secure the micro-visual sensors, pressure sensors, and myoelectric signal sensors to the forearm and upper arm, the fingertips and bases, and the cervical, thoracic, and lumbar spine. These sensors are compact and lightweight, causing no restriction or discomfort to the pianist's movements. They also offer excellent stability and reliability, ensuring continuous and reliable operation during extended performances. By installing these micro-visual sensors, pressure sensors, and myoelectric signal sensors in key areas of the pianist's body, comprehensive and multi-dimensional monitoring of playing posture is achieved, covering multiple aspects such as body position, muscle strength, and neural electrical activity, providing a rich and comprehensive data foundation for subsequent posture analysis and correction.

[0057] Specifically, each sensor is connected to a data processing terminal via Wi-Fi, Bluetooth or 4G for calibration and synchronization.

[0058] It should be noted that each sensor is connected to the data processing terminal via Wi-Fi, Bluetooth, or 4G for calibration and synchronization. Select the appropriate wireless transmission technology based on the actual environment and needs. Wi-Fi is suitable for short-distance, high-data-volume transmission scenarios with a stable router; Bluetooth is suitable for medium- and short-distance, low-power consumption, and good device compatibility; 4G is suitable for long-distance or wide-area coverage scenarios. Install the corresponding wireless communication modules into the sensors and data processing terminal, and perform pairing and connection settings. After a successful connection, use professional calibration software to calibrate each sensor, including color and brightness calibration of the visual sensor, pressure range calibration and zero point calibration of the pressure sensor, sensitivity calibration and noise suppression calibration of the electromyography sensor, etc., to eliminate the impact of individual sensor differences and environmental factors on the data and improve data accuracy and consistency. At the same time, all sensors are synchronized through a high-precision clock synchronization algorithm to ensure the consistency of the collected data on the timeline, providing a reliable time reference for subsequent data fusion and analysis. Sensors are connected to the data processing terminal using wireless technologies such as Wi-Fi, Bluetooth or 4G, and calibrated and synchronized to ensure the stability and real-time nature of data transmission. At the same time, the constraints of cables are eliminated, allowing performers to play freely without interference, thereby improving the accuracy and reliability of data collection.

[0059] Specifically, the sensor is started to collect visual, pressure and electromyographic signal data in real time according to the set sampling rate and frame rate, and stored in the designated location of the terminal to form an original data set.

[0060] It should be noted that the sensors are activated to collect visual, pressure, and electromyographic signal data in real time at the set sampling and frame rates, and stored in a designated location on the terminal to form a raw data set. The sampling and frame rate parameters are set on the data processing terminal. The sampling rate is optimized based on the different sensors and measurement objects. For example, the electromyographic signal sensor sampling rate is set to 1000-2000Hz to capture high-frequency changes in muscle activity; the pressure sensor sampling rate is set to 100-500Hz to meet the measurement requirements of key pressure changes; and the visual sensor frame rate is set to 30-60fps to ensure that the details of the performer's movements can be clearly captured. Then, the written data acquisition program sends a start command to each sensor, and the sensor begins collecting data in real time. The collected data is transmitted to the data processing terminal in real time via a wireless transmission link and stored according to the preset file format and storage path to form a complete original data set, providing original material for subsequent data processing and analysis. By reasonably setting the sampling rate and frame rate to start sensor data collection, it can accurately capture subtle movements and physiological changes during the performance, forming a high-quality original data set, which provides strong support for subsequent in-depth analysis and personalized correction suggestions, thereby realizing accurate monitoring and effective correction of piano playing posture.

[0061] S2. Clean, reduce noise and extract features of the original data set to generate a preprocessed data set and extract biomechanical feature parameters.

[0062] Specifically, the original data set was checked to remove missing values, erroneous values, and outliers.

[0063] It should be noted that the original data set is checked and missing values, erroneous values, and outliers are removed. This requires a comprehensive scan and analysis of the original data set to identify missing values, erroneous values, and outliers. Missing values ​​can be handled by data interpolation or deletion of relevant data points, erroneous values ​​can be corrected by calibration or re-collecting data, and outliers can be detected and eliminated through statistical methods or machine learning algorithms. In this way, the integrity and accuracy of the data can be ensured, providing a reliable foundation for subsequent data processing. By checking the original data set and eliminating missing values, erroneous values, and outliers, strict control of data quality is achieved, ensuring the integrity and accuracy of the data. The role of this step is to provide a reliable original data basis for subsequent data processing, avoid analysis deviations and erroneous conclusions caused by data quality issues, and thus improve the reliability of the entire system for playing posture analysis and correction.

[0064] Specifically, filtering algorithms are used to reduce noise on visual, pressure and electromyographic signal data respectively.

[0065] It should be noted that filtering algorithms are used to reduce noise in the visual, pressure, and electromyographic signal data. Specifically, this involves selecting appropriate filters, such as low-pass filters, high-pass filters, band-pass filters, or band-stop filters, and designing and applying them based on the characteristics of the data and the frequency range of the noise. For example, for visual data, a Gaussian filter can be used to smooth the image and remove high-frequency noise; for pressure data and electromyographic signal data, a wavelet transform or other adaptive filtering algorithms can be used to remove low-frequency or high-frequency noise. In this way, the quality and clarity of the data can be improved, making subsequent feature extraction more accurate and effective. By using filtering algorithms to reduce noise in the visual, pressure, and electromyographic signal data, further purification and optimization of the data are achieved. This step not only improves the signal-to-noise ratio of the data, making the data clearer and easier to analyze, but also provides high-quality data input for subsequent feature extraction, thereby improving the accuracy and efficiency of feature extraction, and thus enhancing the overall system's monitoring accuracy and analysis capabilities for playing posture.

[0066] Specifically, a feature extraction algorithm is used to extract key features from the denoised data, such as body posture features, force change features, and muscle activity features, to generate a preprocessed data set and extract biomechanical feature parameters.

[0067] It should be noted that feature extraction algorithms are used to extract key features from the denoised data, such as body posture features, force change features, and muscle activity features, to generate a preprocessed dataset and extract biomechanical feature parameters. This requires selecting an appropriate feature extraction algorithm based on the different data types and feature requirements. For example, for visual data, Hough transforms, FAST corner detection, or other advanced feature extraction methods can be used to identify key features of body posture; for pressure data, methods such as windowed averaging and peak detection can be used to extract force change features; and for electromyographic signal data, time domain features such as root mean square value and frequency domain features such as power spectral density can be used to extract muscle activity features. These algorithms can transform complex multidimensional data into representative and discriminative feature vectors, providing key input data for subsequent biomechanical model construction. By using feature extraction algorithms to extract key features from the denoised data and generate a preprocessed dataset, effective fusion and compression of multidimensional data is achieved. This step extracts key information from the original data to form a highly representative and discriminative feature vector, which provides key input data for the subsequent construction of the biomechanical model. This greatly reduces the amount of data without losing important information and improves the system's processing efficiency and analysis performance.

[0068] S3. Based on the extracted biomechanical characteristic parameters, a biomechanical model is constructed to reflect the player's posture state and potential deviations.

[0069] Specifically, the biomechanical model is initialized according to the principles of human biomechanics, and the model structure and parameters are defined.

[0070] It should be noted that the biomechanical model can be expressed as: M = {J, S, P, F}, where J represents the set of joints in the human body, S represents the set of bones, P represents the set of muscles, and F represents the set of muscle forces. Each joint, bone, muscle, and muscle force has corresponding parameters, such as joint angle, bone length, muscle attachment point, etc. The biomechanical characteristic parameters are mapped to the model, and the model parameters are adjusted and optimized to accurately reflect the performer's posture state. In order to optimize the model parameters, the least squares method can be used to minimize the error between the model output and the actual measurement data. The error function can be expressed as: where y i is the actual measurement data, is the model output, and n is the number of data points. By adjusting the model parameters to minimize E, the optimal model parameters can be obtained. The posture data is analyzed to identify potential deviations. A deviation model is established in the model to describe and quantify the type and degree of posture deviation. The deviation model can be expressed as: D = {d1, d2, ..., d k}, where each d represents a posture deviation type, and its quantization degree can be expressed as: Among them, p i is the actual posture parameter, is the standard posture parameter, represents the Euclidean norm. This formula calculates the percentage deviation between the actual posture and the standard posture.

[0071] Initializing the biomechanical model is a critical starting step in the multi-dimensional data-based dynamic correction method for piano playing posture. This process is carried out strictly in accordance with the principles of human biomechanics, utilizing a professional biomechanical software platform to meticulously define the model's architecture and parameters. Specifically, the model covers detailed parameters such as joint structure, muscle attachment points, and bone lengths and angles for key parts of the human body, particularly those closely related to piano playing, such as the arms, fingers, and spine. By referencing extensive human anatomy and biomechanics literature and incorporating characteristics of different genders, ages, and body types, the model is personalized to ensure a close match with the performer's anatomy. By initializing the biomechanical model and defining its structure and parameters based on the principles of human biomechanics, a precise digital modeling of the performer's anatomy is achieved, providing a foundational platform that aligns with human physiology for subsequent posture analysis and correction. This ensures the scientific and personalized nature of the entire correction system, allowing subsequent analysis results to be precisely tailored to the unique physical characteristics of each performer.

[0072] Specifically, the biomechanical characteristic parameters are mapped into the model, and the model parameters are adjusted and optimized to accurately reflect the performer's posture state.

[0073] It should be noted that mapping biomechanical characteristic parameters to the model and optimizing them is a precise and complex process. First, characteristic parameters corresponding to the model are extracted from previously collected multidimensional data, such as arm movement angle, finger pressure, spinal curvature, and muscle electrical activity. Then, a data mapping algorithm is used to correlate these characteristic parameters with the corresponding parts in the model to ensure accurate data correspondence. Furthermore, a nonlinear optimization algorithm is used to repeatedly adjust and optimize the model parameters to minimize the error between the model output and the actual measured data. By calculating the gradient of the error function, the model parameters are iteratively adjusted until the model accurately simulates the player's actual posture state, providing a reliable basis for subsequent posture analysis. Through the process of mapping biomechanical characteristic parameters to the model and optimizing the model parameters, the biomechanical model accurately reflects the player's actual posture state during piano playing, significantly improving the accuracy and reliability of posture monitoring. This not only helps to promptly detect subtle posture deviations but also provides key support for in-depth analysis of the root causes of posture problems, laying a solid foundation for the subsequent development of effective corrective strategies.

[0074] Specifically, the posture data is analyzed, potential deviations are identified, a deviation model is established in the model, and the type and degree of posture deviation are described and quantified.

[0075] It should be noted that analyzing posture data, identifying potential deviations, and modeling these deviations within the model are key steps in achieving effective correction. Professional data analysis software is used to deeply mine and analyze the posture data mapped to the model. By comparing this data with a standard piano playing posture database and integrating advanced pattern recognition algorithms, various potential deviations in playing posture are accurately identified. For each deviation, its severity is further quantified, and a corresponding deviation model is constructed. These deviation models, through mathematical formulas and parameter settings, describe key information such as the magnitude, direction, and impact of the deviation in detail. For example, for insufficient finger flexion angle, the model accurately represents the difference between the actual angle and the standard angle, and analyzes its potential impact on playing quality and hand health, providing a scientific basis for subsequent generation of targeted correction recommendations. By analyzing posture data, identifying potential deviations, and modeling these deviations within the model, a comprehensive and quantitative assessment of playing posture issues is achieved, detailing the type and severity of the deviation, and providing clear and specific feedback to the player. This helps to achieve precise personalized correction, improve the standardization and health of playing posture, prevent playing injuries caused by bad posture, and ultimately achieve the goal of improving piano playing quality and sustainable development.

[0076] S4. Use deep learning models to perform real-time analysis on the biomechanical model to generate assessment results for the type and degree of posture deviation.

[0077] Specifically, a large amount of posture data is collected, including normal posture and various deviation posture data, to train the deep learning model.

[0078] It should be noted that a large amount of posture data, including normal posture and various deviation posture data, is collected to train the deep learning model. The training process of the deep learning model can be expressed as: Among them, θ is the model parameter, f(x i ; θ) is the model's response to input x i The prediction of y i is the actual label, L is the loss function, N is the number of training data, and the evaluation results of the type and degree of the performer's posture deviation are generated according to the model output to provide a basis for correction suggestions. The evaluation results can be expressed as: R = {r1, r2, ..., r m}, where each r represents the evaluation result of a posture deviation, including the deviation type and degree. The deviation degree can be expressed as: r j =softmax(z j ), where z jThe softmax function converts the raw scores output by the model into a probability distribution representing the likelihood of a deviation type. The detailed process is as follows. First, through collaboration with multiple professional music schools, a large number of pianists of varying skill levels, ranging from beginners to professionals, were recruited to ensure data diversity and representativeness. During each piano performance, miniature visual sensors, pressure sensors, and myoelectric sensors, previously installed on key areas of the players' arms, fingers, spine, and other key locations, were activated simultaneously to collect multi-dimensional data, including body posture, key pressure, and myoelectric signals. This data was then carefully annotated to clearly identify normal postures, deviations, and the type and extent of the deviations. Subsequently, the massive amount of collected data was preprocessed, including data cleaning and normalization, to eliminate noise and unify the data format to meet the input requirements of the deep learning model. Next, an appropriate deep learning model architecture was selected, such as a convolutional neural network (CNN) for visual data processing or a long short-term memory network (LSTM) for time series sensor data. This preprocessed data was then fed into the model for training. During the training process, model hyperparameters such as the learning rate and batch size were continuously adjusted. Through repeated iterations, the model gradually learned the characteristic differences between normal and deviant postures, until it achieved high recognition accuracy and recall on the validation set. This ultimately completed the training of the deep learning model. By collecting a large amount of posture data and training the deep learning model, it achieved in-depth mining and learning of the characteristics of normal and deviant postures, building a high-precision and highly generalizable posture recognition model. This provides a solid theoretical and data foundation for subsequent posture deviation detection and the generation of correction suggestions. This model can accurately identify various complex piano playing posture issues.

[0079] Specifically, the data of the biomechanical model is input into the trained deep learning model for real-time analysis.

[0080] It should be noted that the biomechanical model data is input into a trained deep learning model for real-time analysis. The specific process is as follows. During piano playing, the data acquisition module continuously acquires real-time data from various sensors and rapidly preprocesses this data according to a predetermined data format and preprocessing process to ensure data quality before input into the deep learning model. This preprocessed data is then transmitted to the trained deep learning model in real time via an efficient communication interface. The deep learning model leverages the feature mapping and pattern recognition capabilities learned during the training phase to rapidly analyze the input biomechanical model data, calculating the degree of deviation between the player's current posture and the normal posture and determining the type of deviation. This process is completed in a remarkably short time, enabling real-time monitoring and analysis of piano playing posture, providing timely data support for further corrective action recommendations. Inputting biomechanical model data into the trained deep learning model for real-time analysis enables real-time and rapid monitoring of playing posture, addressing the lag and inefficiency issues of traditional methods. This ensures that players receive timely feedback during performance, improving the timeliness and effectiveness of corrections.

[0081] Specifically, the model output generates evaluation results of the type and degree of the performer's posture deviation, providing a basis for correction recommendations.

[0082] It should be noted that the model output generates an assessment result of the type and degree of the performer's posture deviation, providing a basis for corrective recommendations. The specific process is as follows. After completing the analysis of the input data, the deep learning model will output a series of result data related to posture deviation, including the probability distribution vector of the deviation type and the quantitative index of the degree of deviation. These result data are further converted and interpreted through a specially designed post-processing algorithm. For example, the probability distribution vector of the deviation type will be converted into a clear posture deviation type label, such as "excessive wrist flexion" and "scoliosis", etc.; the quantitative index of the degree of deviation will be converted into an easy-to-understand degree description, such as "mild deviation", "moderate deviation", "severe deviation", etc. At the same time, combined with the performer's basic information (such as height, playing level, etc.) and the difficulty of the performance repertoire, the evaluation results are personalized to make it more in line with the performer's actual situation. Finally, a detailed posture deviation assessment report is generated, which covers key information such as the posture problems, specific deviation types and degrees of the performer during the performance. This assessment report will serve as an important basis for the subsequent generation of targeted correction suggestions, helping performers to accurately understand their own posture problems and provide a clear direction for subsequent posture correction. Based on the model output, the assessment results of the performer's posture deviation type and degree are generated, providing a scientific and accurate basis for correction suggestions, making the correction measures more targeted and personalized, meeting the actual needs of different performers, effectively improving the quality and effect of piano playing posture correction, and promoting the improvement of performers' technical level and performance expression.

[0083] S5. Combining the evaluation results and the performance rhythm information, generating correction suggestions synchronized with the performance rhythm and reminding the performer to adjust his posture through multimodal feedback.

[0084] Specifically, the piano playing rhythm is analyzed in real time to extract the rhythm characteristics and changing rules.

[0085] It should be noted that the sound signal of the piano performance is collected using an audio interface and converted into a digital signal. The digital signal is then analyzed in the frequency domain using the Fast Fourier Transform (FFT) algorithm to extract characteristic parameters such as the frequency, amplitude, and time domain position of the notes. Next, by calculating the time intervals and frequency change rates between adjacent notes, the tempo, dynamic variations, and rhythmic pattern transitions of the performance are analyzed. Based on this, a rhythm feature database is established to store and classify the extracted rhythm feature parameters for subsequent correlation and matching with posture correction recommendations. This series of sophisticated signal processing and feature extraction steps enables precise grasp of the piano performance rhythm, providing a key basis for the subsequent generation of correction recommendations that match the performance rhythm. By analyzing the piano performance rhythm in real time and extracting its characteristics and changing patterns, a precise grasp of the performance process is achieved, enabling correction recommendations to be closely linked to the actual performance. This not only improves the relevance and effectiveness of correction recommendations but also enhances the performer's acceptance and implementation of these recommendations.

[0086] Specifically, based on the posture deviation evaluation results and the playing rhythm information, correction suggestions matching the playing rhythm are generated.

[0087] It should be noted that the generation of correction suggestions that match the performance rhythm is based on posture deviation assessment results and performance rhythm information. The specific operation is as follows: a correction suggestion generation model is constructed, which uses posture deviation assessment results and performance rhythm information as input variables. By analyzing a large amount of performance data, a library of correction strategies corresponding to different deviation types and degrees is preset. For example, for excessive finger bending deviation, gentle key presses are recommended during fast rhythm sections, while finger stretching exercises are recommended during slow rhythm sections. Furthermore, the frequency and intensity of correction suggestions are adjusted based on the speed and dynamic changes of the performance rhythm, ensuring that correction suggestions are given at appropriate rhythm points, without affecting the continuity of the performance, while effectively guiding the performer to adjust their posture. Using this model, the input posture deviation assessment results and performance rhythm information are processed in real time to generate personalized correction suggestions, including detailed information such as the specific adjustment action, adjustment timing, and adjustment amplitude. Based on the posture deviation assessment results and performance rhythm information, correction suggestions that match the performance rhythm are generated, achieving personalized and dynamic adjustment of correction suggestions, allowing performers to adjust their posture without affecting the continuity of the performance. This correction method, which is adapted to the playing rhythm, helps performers better balance the relationship between posture norms and musical expression.

[0088] Specifically, correction suggestions are conveyed to the performer through multimodal feedback methods such as visual cues, auditory warnings, and tactile vibrations.

[0089] It should be noted that a display screen is placed within the pianist's field of vision to display visual prompts, such as diagrams of body posture, identification of deviations, and arrows indicating the corrective action. An audio player is also provided to play auditory warning signals, such as a specific frequency tone or voice prompts. Furthermore, a tactile vibration device is installed on the pianist's clothing or chair. Depending on the urgency and importance of the corrective action, different vibration patterns and intensities are set to provide tactile reminders to the pianist. The generated corrective action content undergoes modal conversion, converting it into corresponding visual, auditory, and tactile signals, which are then output synchronously via the appropriate feedback devices. For example, for severe postural deviations, a visual flashing prompt, a high-pitched auditory warning, and a strong tactile vibration are triggered simultaneously; for minor deviations, a gentle visual prompt and soft tactile vibration are used. In this way, multimodal feedback can comprehensively attract the pianist's attention, ensuring that they receive the corrective action in a timely manner and make adjustments. By conveying corrective action through multimodal feedback methods such as visual cues, auditory warnings, and tactile vibrations, the player's multiple senses are fully engaged, improving the efficiency and timeliness of corrective action delivery. Multimodal feedback can provide graded reminders based on the severity and urgency of the deviation, ensuring that performers can effectively receive and respond to correction suggestions in different performance scenarios, thereby significantly improving the effectiveness of piano playing posture correction and user experience.

[0090] S6. Feedback the feedback data and correction effect data to the deep learning model to optimize the model parameters and form a closed-loop optimization system.

[0091] Specifically, the player's feedback data on the correction suggestions and the effect data of the corrected posture are collected and marked.

[0092] It should be noted that in actual operation, a set of standardized data collection forms is first designed to record the performer's specific feedback on each correction suggestion, including but not limited to the feedback time, suggestion content, and the performer's subjective feelings (such as whether it is easy to understand, whether it conforms to current playing habits, etc.). At the same time, a dedicated software module in the data processing terminal automatically reads and records the corrected posture data. This data is collected in real time by the micro-visual sensor, pressure sensor, and electromyographic signal sensor mentioned above. The collected feedback data and posture effect data are classified and organized according to key information such as the performer number, correction suggestion sequence number, and timestamp, and marked to ensure data traceability and relevance. For example, a separate folder is created for each performer, and subfolders are generated within the folder according to date and correction suggestion content. The corresponding feedback form and posture data file are stored therein, and corresponding tag information is added to the file name, such as "Performer_A_20250422_Suggestion 1_Feedback and Effect Data". By collecting the performer's feedback data on the correction suggestion and the corrected posture effect data, and organizing and marking them, a comprehensive tracking and quantitative evaluation of the correction process is achieved. This not only provides a rich data resource for subsequent model optimization, but also enables the system to make personalized adjustments and optimizations based on the individual differences of each performer and the actual correction effect. Specifically, the collection of feedback data can directly reflect the performer's acceptance and adaptability to the correction suggestions, providing a subjective dimension reference for model optimization; and the collation and marking of the posture effect data after correction provides a basis for objectively evaluating the correction effect. In this way, the system can more accurately grasp the specific needs and correction difficulties of each performer, thereby achieving truly personalized correction.

[0093] Specifically, the data is fed back to the deep learning model, and the model parameters are updated and adjusted through the model's self-learning optimization mechanism.

[0094] It should be noted that the collated and labeled data undergoes preprocessing, including data cleaning (removing erroneous records and duplicate data) and format conversion (unifying the data format for model access). The processed data is then fed into the deep learning model through a data interface. The model's internal self-learning optimization mechanism primarily involves two aspects: first, a gradient descent-based parameter update, which automatically adjusts model parameters to minimize prediction error by calculating a loss function (e.g., mean squared error) between the model's predictions and actual posture performance data. Second, a reinforcement learning mechanism uses performer feedback as a reward signal to optimize the model's correction suggestion generation strategy, aligning it with performer acceptance and actual correction results. Throughout the entire process, the model undergoes continuous iterative training, retaining the optimized parameter version with each iteration. Key training metrics (e.g., loss and accuracy) are recorded for subsequent evaluation. This data is fed back to the deep learning model, and the model's self-learning optimization mechanism updates and adjusts the model parameters, enabling continuous model evolution and performance improvement. After receiving new feedback and posture effect data, the deep learning model can automatically identify its own shortcomings and continuously optimize the strategy for generating corrective suggestions through parameter adjustment. The parameter update mechanism based on gradient descent can effectively reduce the model's prediction error and improve the accuracy of posture classification and deviation identification. The reinforcement learning mechanism fully considers the performer's subjective feedback, making the corrective suggestions generated by the model more consistent with the performer's actual needs and acceptance. This self-learning optimization mechanism enables the model to continuously accumulate experience during use, gradually adapting to the individual characteristics and playing styles of different performers, thereby significantly improving the effectiveness and pertinence of the corrective suggestions.

[0095] Specifically, the performance of the closed-loop optimization system is evaluated, and based on the results, each part of the system is further optimized to improve the effectiveness and efficiency of posture correction.

[0096] It should be noted that to comprehensively evaluate the performance of the closed-loop optimization system, a multi-dimensional evaluation metric system was established, covering accuracy (such as the accuracy of posture classification and the precision of deviation identification), real-time performance (the time delay from data acquisition to the generation of correction suggestions), adaptability (the ability to adapt to individual differences among performers), and user satisfaction (regular questionnaires collecting performers' evaluations of the overall system experience). Professional performance evaluation tools and scripts were used to test and analyze each component of the system. For example, the stability and speed of the data acquisition module were tested by simulating different playing posture scenarios; the prediction results of the deep learning model were verified using historical datasets, and its accuracy and recall were calculated; and user satisfaction surveys were conducted to collect performers' subjective opinions. Based on the evaluation results, targeted optimizations were implemented for each component of the system, such as optimizing the data transmission protocol to reduce latency, adjusting the model structure to improve accuracy and adaptability, and refining the feedback interface design to enhance the user experience. The entire evaluation and optimization process forms a continuous, iterative closed loop, ensuring continuous improvement in system performance. By optimizing system performance through the evaluation loop and further optimizing various aspects of the system based on the evaluation results, the system is comprehensively upgraded and optimized, ultimately enhancing the overall effectiveness and efficiency of posture correction. The multi-dimensional evaluation metric system comprehensively and objectively reflects the system's performance in terms of accuracy, real-time performance, adaptability, and user satisfaction, providing a clear direction and quantitative basis for system optimization. Addressing issues identified during the evaluation, targeted improvements to various aspects, such as data collection, model prediction, and feedback mechanisms, effectively enhance the system's overall performance. For example, optimizing the data transmission protocol reduces latency, enabling more timely delivery of correction recommendations to performers; adjusting the model structure improves accuracy and adaptability, enhancing the system's ability to adapt to diverse performers and complex performance scenarios; and improving the feedback interface design enhances the user experience and increases performers' enthusiasm and motivation for correction. Through this continuous evaluation and optimization loop, the system can continuously adapt to new needs and challenges, maintaining optimal performance and providing performers with efficient and accurate posture correction services.

[0097] This embodiment also provides a computer device, which is suitable for the case of a dynamic correction method for piano playing posture based on multi-dimensional data, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute computer-executable instructions to implement the dynamic correction method for piano playing posture based on multi-dimensional data proposed in the above embodiment.

[0098] The computer device may be a terminal, comprising a processor, a memory, a communication interface, a display screen and an input device connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device comprises a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner may be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. The display screen of the computer device may be a liquid crystal display or an electronic ink display screen, and the input device of the computer device may be a touch layer covering the display screen, or a button, trackball or touchpad provided on the housing of the computer device, or an external keyboard, touchpad or mouse.

[0099] This embodiment also provides a storage medium having a computer program stored thereon, which, when executed by a processor, implements the method for dynamic correction of piano playing posture based on multi-dimensional data as proposed in the above embodiment; the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk.

[0100] In summary, the present invention achieves comprehensive and accurate monitoring and analysis of piano playing posture through multi-dimensional data collection and fusion, can identify the type and degree of posture deviation in real time, and generate correction suggestions synchronized with the playing rhythm, and timely remind the player to adjust the posture through multi-modal feedback, effectively improving the real-time, accuracy and pertinence of piano playing posture correction. At the same time, the closed-loop optimization system can continuously optimize the deep learning model parameters based on feedback data and correction effects, so that the system can continuously adapt to the individual differences of different players and achieve personalized correction. As the use time increases, the correction effect becomes more accurate and efficient, which is of significant significance for improving the standardization of players' posture and preventing performance-related injuries, and brings innovative solutions to the field of piano playing posture correction.

[0101] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A method for dynamic correction of piano playing posture based on multi-dimensional data, characterized by: include, The multi-dimensional data of the performer is collected in real time through visual, pressure and electromyographic signal sensors to generate the original data set; Clean, reduce noise and extract features from the original data set to generate a preprocessed data set and extract biomechanical feature parameters; Based on the extracted biomechanical characteristic parameters, a biomechanical model reflecting the player's posture state and potential deviations is constructed; Use deep learning models to analyze biomechanical models in real time to generate assessment results that identify the type and degree of posture deviation; Combining the evaluation results with the performance rhythm information, it generates correction suggestions that are synchronized with the performance rhythm and reminds the performer to adjust his posture through multimodal feedback. Feedback data and correction effect data are fed back to the deep learning model to optimize model parameters and form a closed-loop optimization system.

2. The method for dynamic correction of piano playing posture based on multi-dimensional data according to claim 1, wherein: The method uses visual, pressure and electromyographic signal sensors to collect multi-dimensional data of the performer in real time to generate an original data set. The specific steps are as follows: Installing miniature visual sensors, pressure sensors, and myoelectric signal sensors on key parts of the pianist's arms, fingers, spine, etc. Connect each sensor to the data processing terminal via Wi-Fi, Bluetooth or 4G for calibration and synchronization; Start the sensor and collect visual, pressure and electromyographic signal data in real time at the set sampling rate and frame rate, store them in the specified location of the terminal, and form the original data set.

3. The method for dynamic correction of piano playing posture based on multi-dimensional data according to claim 2, wherein: The raw data set is cleaned, denoised and feature extracted to generate a preprocessed data set and extract biomechanical feature parameters. The specific steps are: Check the original data set and remove missing values, erroneous values ​​and outliers; Apply filtering algorithms to reduce noise on visual, pressure and electromyographic signal data respectively; Feature extraction algorithms are used to extract key features from the denoised data, such as body posture features, force change features, and muscle activity features, to generate a preprocessed data set and extract biomechanical feature parameters.

4. The method for dynamic correction of piano playing posture based on multi-dimensional data according to claim 3, wherein: The biomechanical model reflecting the player's posture state and potential deviation is constructed based on the extracted biomechanical characteristic parameters. The specific steps are: Initialize the biomechanical model according to the principles of human biomechanics and define the model structure and parameters; Mapping biomechanical characteristic parameters to the model, adjusting and optimizing the model parameters to accurately reflect the player's posture; Analyze posture data, identify potential deviations, build deviation models in the model, and describe and quantify the type and degree of posture deviation.

5. The method for dynamic correction of piano playing posture based on multi-dimensional data according to claim 4, characterized in that: The method uses a deep learning model to perform real-time analysis on the biomechanical model to generate an assessment result of the type and degree of posture deviation, specifically comprising the following steps: Collect a large amount of posture data, including normal posture and various deviation posture data, to train deep learning models; The data of the biomechanical model is fed into the trained deep learning model for real-time analysis; Based on the model output, an assessment result of the type and degree of the performer's posture deviation is generated to provide a basis for correction recommendations.

6. The method for dynamic correction of piano playing posture based on multi-dimensional data according to claim 5, characterized in that: The method combines the evaluation results and the playing rhythm information to generate correction suggestions synchronized with the playing rhythm and reminds the performer to adjust his posture through multimodal feedback. The specific steps are: Analyze piano playing rhythm in real time to extract rhythm features and changing patterns; Generate correction suggestions that match the playing rhythm based on the posture deviation assessment results and the playing rhythm information; Corrective suggestions are conveyed to the performer through multimodal feedback methods such as visual cues, auditory warnings, and tactile vibrations.

7. The method for dynamic correction of piano playing posture based on multi-dimensional data according to claim 6, characterized in that: The feedback data and correction effect data are fed back to the deep learning model to optimize the model parameters and form a closed-loop optimization system. The specific steps are: Collect the performers' feedback on the correction suggestions and the effect of the corrected posture, organize and mark them; Feed the data back to the deep learning model, and update and adjust the model parameters through the model's self-learning optimization mechanism; Evaluate the performance of the closed-loop optimization system and further optimize each part of the system based on the results to improve the effectiveness and efficiency of posture correction.

8. A piano playing posture dynamic correction system based on multi-dimensional data, based on the piano playing posture dynamic correction method based on multi-dimensional data according to any one of claims 1 to 7, characterized in that: include, Data acquisition module: collects multi-dimensional data of the performer in real time through visual, pressure and electromyographic signal sensors to generate original data sets; Data preprocessing module: cleans, reduces noise and extracts features from the original data set, generates a preprocessed data set and extracts biomechanical characteristic parameters; Model building module: Based on the extracted biomechanical characteristic parameters, a biomechanical model reflecting the player's posture state and potential deviations is constructed; Analysis and evaluation module: Uses deep learning models to analyze biomechanical models in real time and generates assessment results to identify the type and degree of posture deviation; Feedback Correction Module: This module combines the evaluation results with the performance rhythm information to generate correction suggestions that are synchronized with the performance rhythm and uses multimodal feedback to remind the performer to adjust their posture. System optimization module: Feedback data and correction effect data are fed back to the deep learning model to optimize model parameters and form a closed-loop optimization system.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method for dynamic correction of piano playing posture based on multi-dimensional data according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method for dynamic correction of piano playing posture based on multi-dimensional data according to any one of claims 1 to 7 are implemented.

Images