基于多通道表面肌电信号的手势识别方法研究

Study on Gesture Recognition Method Based on Multi-channel sEMG Signal

为验证基于表面肌电信号控制智能假手的可靠性,构建一种基于网格划分(Grid Search)优化支持向量机(SVM)的手势动作识别模型。从Ninapro数据集中截取部分动作信号数据,通过提取5种时域特征和基于小波包的时频域特征,利用SVM和Grid Search-SVM对不同种手势动作进行分类,并对比分类模型的可靠性。实验结果表明,在训练集数据量不同时,分类效果不同;且在训练程度相同时,优化后的分类模型较未优化模型分类效果更好,平均准确率相比提高了21.37%。该模型的建立为以后实时控制智能假肢奠定了基础。

A gesture recognition model based on Grid Search optimization support vector machine(SVM)was constructed to verify the reliability of the intelligent artificial hand control based on sEMG signals.Part of the motion signal data was intercepted from the Ninapro data set.By extracting five time-domain features and time-frequency domain features based on wavelet packets,SVM and Grid Search-SVM were used to classify different gestures and compare the reliability of the classification model.The experimental results show that the classification effect is different when the data amount of training set is different.At the same training level,the optimized classification model had better classification effect than the unoptimized model,and the average accuracy was improved by 21.37%.The model lays a foundation for the future real-time control of intelligent prosthesis.