Lower limb motion recognition method and system based on surface electromyogram signals

Abstract

The invention relates to a lower limb movement recognition method and system based on surface electromyogram signals, and the method comprises the following steps: 1, collecting the surface electromyogram signals in different movement modes, carrying out the preprocessing, and extracting the time domain, frequency domain and nonlinear features; 2, performing hybrid optimization on penalty parameters and kernel parameters of the kernel extreme learning machine by using an artificial bee colony algorithm-sparrow search algorithm to obtain an optimal kernel extreme learning machine classifier; and step 3, carrying out identification by using the optimized classifier. The method and the system are beneficial to improving the accuracy of lower limb motion mode recognition.

Description

technical field [0001] The invention belongs to the technical field of motion recognition, and in particular relates to a lower limb motion recognition method and system based on surface electromyography signals. Background technique [0002] China is currently the only country in the world with an elderly population of over 200 million. With the continuous growth of age, the functions of all aspects of the body of the elderly will undergo obvious changes, resulting in lower limb movement disorders. Osteoarthritis of the knee is the main cause of lower extremity motor dysfunction in the elderly, greatly affecting their daily life. [0003] When a patient is diagnosed with knee osteoarthritis, the physical therapist is trained to design muscle exercises related to daily activities for the purpose of rehabilitation. To help clinicians monitor rehabilitation progress remotely while assessing and treating patients, more efficient and intelligent frameworks are needed to decode...

AI Technical Summary

Problems solved by technology

Xi Xugang and others used support vector machines to classify the four movements of fist clenching, fist stretching, wrist extension, and wrist flexion, and achieved good results. However, support vector machines have long calculation times and are prone to local optimal solutions.

The backpropagation neural network (BP) trains the neural network through forward propagation of experimental data and backpropagation of errors. When performing motion classification, there will be disadvantages such as cumbersome weight parameter settings and slow training speed.

Huang Guangbin proposed a single hidden layer feedforward neural network algorithm, that is, the extreme learning machine (ELM), the learning speed is faster than the BP neural network, and the accuracy is higher, but the weight parameters of ELM and the number of neurons in the hidden layer will have a greater impact on the classification results

Kernel extreme learning machine (KELM) introduces regularization coefficient and kernel function into ELM, which can solve the problem of poor stability of classification results caused by random selection of weight parameters and number of hidden layer neurons in ELM.

Liu Ao et al. used particle swarm optimization (PSO) to optimize the penalty parameters and kernel functions of the kernel extreme learning machine, which can achieve better classification performance, but PSO is prone to fall into local optimum during the optimization process, resulting in reduced convergence accuracy, thus making Classification performance drops

Gao Xiangming et al. used the artificial bee colony algorithm (ABC) to optimize the penalty parameters and kernel functions of the kernel extreme learning machine, and the classification accuracy is high. Although it has the advantages of strong global optimization ability and few parameters, it is easy to fall into local optimum.

Du Ye et al. used the sparrow search algorithm (SSA) to optimize the penalty parameters and kernel functions of the kernel extreme learning machine. The local optimization ability is strong, but the global optimization ability needs to be strengthened, and the classification accuracy needs to be improved.

Images