The pattern-recognition-based control of myoelectric prostheses offers amputees a natural, intuitive approach to more finely control the prostheses. In this context, we recently developed a multichannel surface electromyography (sEMG) module with a low sampling rate and applied it for hand-motion research. In this study, we investigate the effects of the sEMG-signal sampling rate and feature extraction window length on the classification accuracy in hand-motion recognition. Ten normal subjects and one forearm amputee were made to wear an armband module consisting of eight EMG sensors, and seven and four hand movements of the normal subjects and amputee, respectively, were measured. The EMG signal was measured at 500 Hz and down-sampled to 250, 100, and 50 Hz. Four time-domain features (mean average value, waveform length, zero crossing, and slope sign change) were calculated as the sEMG features with six selected window lengths, which were increased in 50 ms intervals (50, 100, 150, 200, 250, and 300 ms). Hand-motion recognition was performed using artificial neural network, support vector machine, decision tree, and k-nearest neighbor classifiers. Our results showed that for all classifiers and all subjects, the hand-motion classification accuracy increases with an increase in the sampling rate and window length. We believe that our findings will aid in selecting the appropriate sampling rate and window length for prosthetics meant for daily use.

肌电假肢的基于模式识别的控制为截肢者提供了一种自然、直观的方法来更精细地控制假肢。在这种情况下，我们最近开发了一种低采样率的多通道表面肌电图 (sEMG) 模块，并将其应用于手部运动研究。在本研究中，我们研究了 sEMG 信号采样率和特征提取窗口长度对手部动作识别分类精度的影响。10 名正常受试者和 1 名前臂截肢者佩戴由 8 个 EMG 传感器组成的臂章模块，分别测量正常受试者和截肢者的 7 次和 4 次手部运动。EMG 信号在 500 Hz 下测量并下采样到 250、100 和 50 Hz。四个时域特征（平均值、波形长度、过零、和斜率符号变化）被计算为具有六个选定窗口长度的 sEMG 特征，它们以 50 毫秒的间隔（50、100、150、200、250 和 300 毫秒）增加。使用人工神经网络、支持向量机、决策树和k-最近邻分类器进行手部运动识别。我们的结果表明，对于所有分类器和所有受试者，手部运动分类精度随着采样率和窗口长度的增加而增加。我们相信我们的发现将有助于为日常使用的假肢选择合适的采样率和窗口长度。支持向量机、决策树和 k-近邻分类器。我们的结果表明，对于所有分类器和所有受试者，手部运动分类精度随着采样率和窗口长度的增加而增加。我们相信我们的发现将有助于为日常使用的假肢选择合适的采样率和窗口长度。支持向量机、决策树和 k-近邻分类器。我们的结果表明，对于所有分类器和所有受试者，手部运动分类精度随着采样率和窗口长度的增加而增加。我们相信我们的发现将有助于为日常使用的假肢选择合适的采样率和窗口长度。