摘要
针对表面肌电(SEMG)的非平稳特性,提出采用小波包变换方法对其进行分类。分析了特征提取方法并采用小波包变换各频段能量构造特征矢量,经过学习矢量量化神经网络训练能够有效地从伸肌和屈肌采集的两道肌电信号中识别伸拳,展拳,腕内旋,腕外旋4种运动模式,平均识别率为94.5%。与其它时频分析方法比较,该方法不仅识别率高,鲁棒性好,也为其他非平稳生理信号分析提供了新手段。
A surface electromyography (SEMG) signal classification method based on wavelet packet transformation (WPT) is presented in this paper. The feature extraction method is analyzed. The energies in different frequency bands selected as robust feature vectors, four types of forearm movement are identified through learning vector quantization neural network. Compared with other time-frequency analysis method, this method has a higher identification rate and great potential in analyzing other non-stationary physiological signals.
出处
《医疗卫生装备》
CAS
2003年第9期7-8,10,共3页
Chinese Medical Equipment Journal
基金
国家自然科学基金项目(编号:60171006)。
关键词
小波包变换
表面肌电信号
学习矢量量化
时频分析
神经网络训练
wavelet packet transformation
EMG
time-frequency analysis
learning vector quantization
neural network
pattern classification
