基于双分支网络的表面肌电信号识别方法

王万良,潘杰,王铮,潘家宇

Recognition method of surface electromyographic signal based on two-branch network

Wanliang WANG,Jie PAN,Zheng WANG,Jiayu PAN

表 5 所有方法的计算资源分析

Tab.5 Computational resource analysis for all methods

方法 Np/106 C/MB te/s tp/s tr/ms KNN+FE[30] — — 0.78±0.01 86.86±1.22 0.020±0.001 SVM+FE[31] — — 0.48±0.01 77.06±1.62 0.030±0.001 CNN[8] 4.98 162.73 22.65±0.98 148.61±3.38 0.078±0.003 CNN+FE 0.76 117.30 14.13±0.46 101.32±2.11 0.066±0.002 CNN+FE+ETD 1.15 184.48 20.19±0.65 153.51±3.65 0.086±0.002 RESNET+FE[32] 0.55 336.88 21.96±0.52 147.53±4.06 0.071±0.002 Bi-GRU[33] 1.41 118.10 16.76±0.50 160.49±2.98 0.078±0.002 LSTM[34] 1.52 146.35 17.14±0.59 168.25±3.16 0.082±0.002 TRANSFORMER[35] 2.49 335.68 24.16±0.66 187.23±4.56 0.079±0.002 LCNN[36] 1.77 533.50 21.06±0.62 180.85±4.29 0.082±0.003 TDACAPS[14] 0.87 741.38 216.45±6.16 1978.12±14.18 1.047±0.030 ETDTBN 3.73 321.41 17.85±0.48 156.28±3.99 0.076±0.002