A sensor-based lower limb training effect evaluation system

Abstract

The present invention belongs to the field of lower limb training effect evaluation, and specifically discloses a sensor-based lower limb training effect evaluation system, the system includes: an electromyographic data acquisition module, an electromyographic signal inspection module and a muscle strength evaluation module. The present invention constructs an evaluation model to analyze electromyographic signals, uses root mean square to perform cross-channel normalization processing on electromyographic signals, and extracts time domain features and correlation coefficients from the normalized signals to ensure that electromyographic signals of different channels have similar magnitudes, combines random forest and support vector machine regression to construct an evaluation model, classifies the state of electromyographic signals and then evaluates muscle strength, combines the advantages of both, improves the accuracy of the evaluation model, and uses genetic algorithms to optimize the parameters of the support vector machine regression model, effectively reduces the risk of overfitting, and improves the generalization ability of the evaluation model.

Description

Technical Field

[0001] The present invention relates to the field of lower limb training effect evaluation, and in particular to a sensor-based lower limb training effect evaluation system. Background Art

[0002] The evaluation of the effect of lower limb training specifically refers to a series of evaluations and tests for lower limb exercises. The evaluation of the effect of lower limb training is crucial to determine the effectiveness of the training plan, monitor the progress of training, prevent sports injuries, and optimize sports performance. However, there are individual differences in the evaluation data. It is unreliable to rely on experience to analyze the effect of lower limb training. It will be affected by subjective judgment, knowledge limitations, and personal preferences. Analyzing muscle strength based on electromyographic signals can more accurately evaluate the training effect. The change of electromyographic signals is a dynamic process. The feature extraction method must be able to adapt to this dynamic process and vary due to individual physiological differences. The relationship between muscle strength and electromyographic signals is usually a more complex nonlinear relationship. The characteristics of electromyographic signals are different at different levels of muscle activity. Traditional analysis methods cannot automatically adjust to adapt to changes in electromyographic signals. Summary of the invention

[0003] In view of the above situation, in order to overcome the defects of the prior art, the present invention provides a sensor-based lower limb training effect evaluation system. In view of the technical problems that there are individual differences in evaluation data, relying on experience to analyze the lower limb training effect is unreliable and will be affected by subjective judgment, knowledge limitations and personal preferences, the present invention constructs an evaluation model to analyze the electromyographic signal, which can provide consistent evaluation results, reduce the influence of subjective judgment, and improve the accuracy of the evaluation; the change of the electromyographic signal is a dynamic process, the feature extraction method must be able to adapt to this dynamic process, and there are different technical problems due to individual physiological differences. The present invention uses the root mean square to perform cross-channel normalization on the electromyographic signal, and extracts time domain features and Correlation coefficient ensures that the electromyographic signals of different channels have similar magnitudes to facilitate subsequent data analysis and reduce the computational complexity of data analysis; the relationship between muscle strength and electromyographic signals is usually a complex nonlinear relationship, and the characteristics of electromyographic signals are different when the muscles are at different activity levels. The traditional analysis method cannot automatically adjust to adapt to the changes in electromyographic signals. The present invention combines random forest and support vector machine regression to construct an evaluation model, classifies the state of the electromyographic signal and then evaluates the muscle strength, combining the respective advantages of the two to improve the accuracy of the evaluation model, and uses a genetic algorithm to optimize the parameters of the support vector machine regression model, effectively reducing the risk of overfitting and improving the generalization ability of the evaluation model.

[0004] The technical solution adopted by the present invention is as follows: The present invention provides a sensor-based lower limb training effect evaluation system, the sensor-based lower limb training effect evaluation system comprises an electromyographic data acquisition module, an electromyographic signal inspection module and a muscle strength evaluation module;

[0005] The electromyographic data acquisition module attaches surface electrodes to the skin along the direction of muscle fibers, connects the surface electrodes to an electromyograph via wires, and records electromyographic signals;

[0006] The electromyographic signal inspection module monitors the electromyographic signal in real time, and if there is obvious noise and interference, the electromyographic signal is recollected, otherwise, the electromyographic signal is transmitted to the muscle strength assessment module;

[0007] The muscle strength assessment module uses a muscle strength assessment method to assess muscle strength, and the muscle strength assessment method specifically includes the following steps:

[0008] Step S1: training data collection, recording the training status, using an electromyograph and a sensor to collect electromyographic signals and muscle strength during lower limb training, recorded as training signals and training strength, the training signals, training status and training strength constituting a training data set;

[0009] Step S2: data preprocessing, including data cleaning, data normalization and feature extraction;

[0010] Step S3: Classification of lower limb training content and strength assessment, using random forest and support vector machine regression to build an assessment model, training the assessment model, and outputting predicted strength;

[0011] Step S4: Input the electromyographic signal into the training completed evaluation model for analysis to obtain the strength evaluation result of the lower limb training.

[0012] Furthermore, step S2, data preprocessing, specifically includes the following steps:

[0013] Step S21: data cleaning, removing the DC component of the electromyographic signal, filtering with a narrow-band notch filter and half-wave rectification in sequence to obtain a filtered signal;

[0014] Step S22: Use the root mean square to perform cross-channel normalization processing on the filtered signal to obtain a normalized signal. The formula used is as follows:

[0015] ;

[0016] In the formula, It is for The function to normalize the filtered signal in each EMG channel, is the size of the window for filtering the signal, is the time interval of the window for filtering the signal, is the myoelectric channel in the window, It's time. is the first The signal of the EMG channel, yes Moment Discrete EMG signals of EMG channels;

[0017] Step S23: extracting time domain features and correlation coefficients from the normalized signal, calculating the mean absolute value of the normalized signal to reflect the signal strength of the normalized signal, and calculating the zero-order power spectrum, second-order power spectrum, fourth-order power spectrum and sixth-order power spectrum of the normalized signal by using the differential characteristics of the discrete electromyographic signal. The formula used is as follows:

[0018] ;

[0019] In the formula, yes The power spectrum of the order, , is the frequency index, The frequency index is Discrete electromyographic signal at ;

[0020] Step S24: Calculate the first-order average amplitude change and the second-order average amplitude change to represent indirect frequency information. The formula used is as follows:

[0021] ;

[0022] In the formula, yes The function of the change in the order average amplitude, , is the change value of the discrete EMG signal;

[0023] Step S25: Use a logarithmic function to perform nonlinear transformation on the above-mentioned mean absolute value, zero-order power spectrum, second-order power spectrum, fourth-order power spectrum, sixth-order power spectrum, first-order average amplitude change and second-order average amplitude change, calculate the similarity strength of each pair of myoelectric channels, and remove the myoelectric channels with a similarity higher than 70%. The formula used is as follows:

[0024] ;

[0025] In the formula, and There are two different myoelectric channels. is the correlation coefficient function, and They are and The average value of and They are The EMG channel value at the moment.

[0026] Further, step S3, classification of lower limb training content and strength assessment, specifically includes the following steps:

[0027] Step S31: Establish and initialize a random forest model, input the training signal and training state into the random forest model for training, and the random forest model outputs the training state;

[0028] Step S32: Establish and initialize the support vector machine regression model, group the training signals and training forces by training status and input them into the support vector machine regression model for training. The support vector machine regression model outputs the predicted force, sets the Gaussian radius basis function as the kernel function of the support vector machine regression model, and maps the training force to the high-dimensional space. The formula used is as follows:

[0029] ;

[0030] In the formula, is a constant, is the kernel function, is the input training signal, It is A training signal, It is training status;

[0031] Step S33: Calculate the optimal regression function, the formula used is as follows:

[0032] ;

[0033] In the formula, is the regression function, is the optimal regression function, is the weight vector, is a predetermined constant parameter, and They are the upper and lower constraints of the output of the support vector machine regression model;

[0034] Step S34: define the loss function of the support vector machine regression model, the formula used is as follows:

[0035] ;

[0036] In the formula, is the loss function of the support vector machine regression model, is the input training signal, is the input training force, It is the power of prediction, It is the training state;

[0037] Step S35: Optimize the support vector machine regression model by using a genetic algorithm to optimize the hyperplane parameters in the support vector machine regression model. The objective function formula of the genetic algorithm is as follows:

[0038] ;

[0039] In the formula, is the objective function of the genetic algorithm, The value range is [0, 1], which is the weight that controls the relationship between the number of features and the evaluation accuracy. is the evaluation accuracy, is the total number of samples in the training dataset.

[0040] The beneficial effects achieved by the present invention using the above scheme are as follows:

[0041] (1) In view of the technical problem that there are individual differences in evaluation data, the effect of lower limb training based on experience is unreliable and will be affected by subjective judgment, knowledge limitations and personal preferences, the present invention constructs an evaluation model to analyze electromyographic signals, which can provide consistent evaluation results, reduce the influence of subjective judgment, and improve the accuracy of evaluation;

[0042] (2) The change of electromyographic signals is a dynamic process. The feature extraction method must be able to adapt to this dynamic process. Different technical problems exist due to individual physiological differences. The present invention uses the root mean square to perform cross-channel normalization on the electromyographic signals and extracts time domain features and correlation coefficients from the normalized signals to ensure that the electromyographic signals of different channels have similar magnitudes, so as to facilitate subsequent data analysis and reduce the computational complexity of data analysis.

[0043] (3) In view of the fact that the relationship between muscle strength and electromyographic signals is usually a complex nonlinear relationship, the characteristics of electromyographic signals are different when the muscles are at different activity levels, and the traditional analysis methods cannot automatically adjust to adapt to the changes in electromyographic signals, the present invention combines random forest and support vector machine regression to construct an evaluation model, classifies the state of electromyographic signals and then evaluates muscle strength, combining the advantages of both methods to improve the accuracy of the evaluation model, and uses genetic algorithms to optimize the parameters of the support vector machine regression model, effectively reducing the risk of overfitting and improving the generalization ability of the evaluation model. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 A module connection diagram of a sensor-based lower limb training effect evaluation system provided by the present invention;

[0045] Figure 2 This is a connection diagram of the data preprocessing steps in step S2 of the present invention.

[0046] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention and do not constitute a limitation of the present invention. DETAILED DESCRIPTION

[0047] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments; based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0048] Example 1: See Figure 1 , this embodiment provides a sensor-based lower limb training effect evaluation system, the sensor-based lower limb training effect evaluation system includes an electromyographic data acquisition module, an electromyographic signal inspection module and a muscle strength evaluation module;

[0049] The electromyographic data acquisition module attaches surface electrodes to the skin along the direction of muscle fibers, connects the surface electrodes to an electromyograph via wires, and records electromyographic signals;

[0050] The electromyographic signal checking module monitors the electromyographic signal in real time. If there is obvious noise and interference, the electromyographic signal is recollected; otherwise, the electromyographic signal is transmitted to the muscle strength evaluation module.

[0051] Example 2: See Figure 1 This embodiment is based on the above embodiment, and the muscle strength assessment module uses a muscle strength assessment method to assess muscle strength, and the muscle strength assessment method specifically includes the following steps:

[0052] Step S1: training data collection, recording the training status, using an electromyograph and a sensor to collect electromyographic signals and muscle strength during lower limb training, recorded as training signals and training strength, the training signals, training status and training strength constituting a training data set;

[0053] Step S2: data preprocessing, including data cleaning, data normalization and feature extraction;

[0054] Step S3: Classification of lower limb training content and strength assessment, using random forest and support vector machine regression to build an assessment model, training the assessment model, and outputting predicted strength;

[0055] Step S4: Input the electromyographic signal into the training completed evaluation model for analysis to obtain the strength evaluation result of the lower limb training.

[0056] Through the above operations, in view of the technical problem that there are individual differences in evaluation data, relying on experience to analyze the effect of lower limb training is unreliable and will be affected by subjective judgment, knowledge limitations and personal preferences, the present invention constructs an evaluation model to analyze electromyographic signals, which can provide consistent evaluation results, reduce the influence of subjective judgment, and improve the accuracy of evaluation.

[0057] Example 3: See Figure 1 and Figure 2 This embodiment is based on the above embodiment, step S2, data preprocessing, specifically includes the following steps:

[0058] Step S21: data cleaning, removing the DC component of the electromyographic signal, filtering with a narrow-band notch filter and half-wave rectification in sequence to obtain a filtered signal;

[0059] Step S22: Use the root mean square to perform cross-channel normalization processing on the filtered signal to obtain a normalized signal. The formula used is as follows:

[0060] ;

[0061] In the formula, It is for The function to normalize the filtered signal in each EMG channel, is the size of the window for filtering the signal, is the time interval of the window for filtering the signal, is the myoelectric channel in the window, It's time. is the first The signal of the EMG channel, yes Moment Discrete EMG signals of EMG channels;

[0062] Step S23: extracting time domain features and correlation coefficients from the normalized signal, calculating the mean absolute value of the normalized signal to reflect the signal strength of the normalized signal, and calculating the zero-order power spectrum, second-order power spectrum, fourth-order power spectrum and sixth-order power spectrum of the normalized signal by using the differential characteristics of the discrete electromyographic signal. The formula used is as follows:

[0063] ;

[0064] In the formula, yes The power spectrum of the order, , is the frequency index, The frequency index is Discrete electromyographic signal at ;

[0065] Step S24: Calculate the first-order average amplitude change and the second-order average amplitude change to represent indirect frequency information. The formula used is as follows:

[0066] ;

[0067] In the formula, yes The function of the change in the order average amplitude, , is the change value of the discrete EMG signal;

[0068] Step S25: Use a logarithmic function to perform nonlinear transformation on the above-mentioned mean absolute value, zero-order power spectrum, second-order power spectrum, fourth-order power spectrum, sixth-order power spectrum, first-order average amplitude change and second-order average amplitude change, calculate the similarity strength of each pair of myoelectric channels, and remove the myoelectric channels with a similarity higher than 70%. The formula used is as follows:

[0069] ;

[0070] In the formula, and There are two different myoelectric channels. is the correlation coefficient function, and They are and The average value of and They are The EMG channel value at the moment.

[0071] Through the above operations, the change of electromyographic signals is a dynamic process, and the feature extraction method must be able to adapt to this dynamic process. In addition, there are different technical problems due to individual physiological differences. The present invention uses the root mean square to perform cross-channel normalization on the electromyographic signals, and extracts time domain features and correlation coefficients from the normalized signals to ensure that the electromyographic signals of different channels have similar magnitudes, so as to facilitate subsequent data analysis and reduce the computational complexity of data analysis.

[0072] Example 4: See Figure 1 This embodiment is based on the above embodiment, step S3, lower limb training content classification and strength evaluation, specifically includes the following steps:

[0073] Step S31: Establish and initialize a random forest model, input the training signal and training state into the random forest model for training, and the random forest model outputs the training state;

[0074] Step S32: Establish and initialize the support vector machine regression model, group the training signals and training forces by training status and input them into the support vector machine regression model for training. The support vector machine regression model outputs the predicted force, sets the Gaussian radius basis function as the kernel function of the support vector machine regression model, and maps the training force to the high-dimensional space. The formula used is as follows:

[0075] ;

[0076] In the formula, is a constant, is the kernel function, is the input training signal, It is A training signal, It is training status;

[0077] Step S33: Calculate the optimal regression function, the formula used is as follows:

[0078] ;

[0079] In the formula, is the regression function, is the optimal regression function, is the weight vector, is a predetermined constant parameter, and They are the upper and lower constraints of the output of the support vector machine regression model;

[0080] Step S34: define the loss function of the support vector machine regression model, the formula used is as follows:

[0081] ;

[0082] In the formula, is the loss function of the support vector machine regression model, is the input training signal, is the input training force, It is the power of prediction, It is the training state;

[0083] Step S35: Optimize the support vector machine regression model by using a genetic algorithm to optimize the hyperplane parameters in the support vector machine regression model. The objective function formula of the genetic algorithm is as follows:

[0084] ;

[0085] In the formula, is the objective function of the genetic algorithm, The value range is [0, 1], which is the weight that controls the relationship between the number of features and the evaluation accuracy. is the evaluation accuracy, is the total number of samples in the training dataset.

[0086] Through the above operations, in view of the technical problem that the relationship between muscle strength and electromyographic signals is usually a complex nonlinear relationship, the characteristics of electromyographic signals are different when the muscles are at different activity levels, and the traditional analysis methods cannot automatically adjust to adapt to the changes in electromyographic signals, the present invention combines random forest and support vector machine regression to construct an evaluation model, classifies the state of the electromyographic signals and then evaluates the muscle strength, combines the respective advantages of the two, improves the accuracy of the evaluation model, and uses a genetic algorithm to optimize the parameters of the support vector machine regression model, effectively reduces the risk of overfitting, and improves the generalization ability of the evaluation model.

[0087] Embodiment 5. This embodiment is based on the above embodiment. In Embodiment 1 and Embodiment 3, an electromyograph is used to obtain electromyographic signals. Surface electrodes are attached to the skin, mainly to the gluteus maximus, quadriceps femoris, hamstrings, triceps surae, biceps femoris and adductor muscles. The sampling frequency is 1 khz. The sensor described in Embodiment 3 includes a pressure sensor, an acceleration sensor and a strain sensor. The pressure sensor is fixed on the sole of the foot to measure the pressure distribution applied to the sole of the foot to obtain the force distribution during training. The strain sensor is attached to the key structure of the leg training equipment to monitor its strain to indirectly reflect the situation.

[0088] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0089] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

[0090] The present invention and its embodiments are described above, and such description is not restrictive. The drawings show only one embodiment of the present invention, and the actual structure is not limited thereto. In short, if ordinary technicians in the field are inspired by it, without departing from the purpose of the invention, they can design a structure and embodiment similar to the technical solution without creativity, which should belong to the protection scope of the present invention.

Claims

1. A sensor-based lower limb training effect evaluation system, characterized in that: It includes EMG data acquisition module, EMG signal inspection module and muscle strength assessment module; The electromyographic data acquisition module attaches surface electrodes to the skin along the direction of muscle fibers, connects the surface electrodes to an electromyograph via wires, and records electromyographic signals; The electromyographic signal inspection module monitors the electromyographic signal in real time, and if there is obvious noise and interference, the electromyographic signal is recollected, otherwise, the electromyographic signal is transmitted to the muscle strength assessment module; The muscle strength assessment module uses a muscle strength assessment method to assess muscle strength, and the muscle strength assessment method specifically includes the following steps: Step S1: training data collection, recording the training status, using an electromyograph and a sensor to collect electromyographic signals and muscle strength during lower limb training, recorded as training signals and training strength, the training signals, training status and training strength constituting a training data set; Step S2: data preprocessing, including data cleaning, data normalization and feature extraction; Step S3: Classification of lower limb training content and strength assessment, using random forest and support vector machine regression to build an assessment model, training the assessment model, and outputting predicted strength; Step S4: inputting the electromyographic signal into the training completed evaluation model for analysis to obtain the strength evaluation result of the lower limb training; Step S2, data preprocessing, specifically includes the following steps: Step S21: data cleaning, removing the DC component of the electromyographic signal, filtering with a narrow-band notch filter and half-wave rectification in sequence to obtain a filtered signal; Step S22: Use the root mean square to perform cross-channel normalization processing on the filtered signal to obtain a normalized signal. The formula used is as follows: ； In the formula, It is for The function to normalize the filtered signal in each EMG channel, is the size of the window for filtering the signal, is the time interval of the window for filtering the signal, is the myoelectric channel in the window, It's time. is the first The signal of the EMG channel, yes Moment Discrete EMG signals of EMG channels; Step S23: extracting time domain features and correlation coefficients from the normalized signal, calculating the mean absolute value of the normalized signal to reflect the signal strength of the normalized signal, and calculating the zero-order power spectrum, second-order power spectrum, fourth-order power spectrum and sixth-order power spectrum of the normalized signal by using the differential characteristics of the discrete electromyographic signal. The formula used is as follows: ； In the formula, yes The power spectrum of the order, , is the frequency index, The frequency index is Discrete electromyographic signal at ; Step S24: Calculate the first-order average amplitude change and the second-order average amplitude change to represent indirect frequency information. The formula used is as follows: ； In the formula, yes The function of the change in the order average amplitude, , is the change value of the discrete EMG signal; Step S25: using a logarithmic function to perform nonlinear transformation on the mean absolute value, the zero-order power spectrum, the second-order power spectrum, the fourth-order power spectrum, the sixth-order power spectrum, the first-order average amplitude change, and the second-order average amplitude change, calculate the similarity strength of each pair of myoelectric channels, and remove the myoelectric channels with a similarity higher than 70%; Step S3: Classification of lower limb training content and strength assessment, specifically including the following steps: Step S31: Establish and initialize a random forest model, input the training signal and training state into the random forest model for training, and the random forest model outputs the training state; Step S32: Establish and initialize the support vector machine regression model, group the training signals and training forces by training status and input them into the support vector machine regression model for training. The support vector machine regression model outputs the predicted force, sets the Gaussian radius basis function as the kernel function of the support vector machine regression model, and maps the training force to the high-dimensional space. The formula used is as follows: ； In the formula, is a constant, is the kernel function, is the input training signal, It is A training signal, It is training status; Step S33: Calculate the optimal regression function, the formula used is as follows: ； In the formula, is the regression function, is the optimal regression function, is the weight vector, is a predetermined constant parameter, and They are the upper and lower constraints of the output of the support vector machine regression model; Step S34: define the loss function of the support vector machine regression model, the formula used is as follows: ； In the formula, is the loss function of the support vector machine regression model, is the input training signal, is the input training force, It is the power of prediction, It is the training state; Step S35: Optimizing the support vector machine regression model, using a genetic algorithm to optimize the hyperplane parameters in the support vector machine regression model.

Images