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 2 电极偏移情况下各方法的识别效果

Tab.2 Recognition effect of each method under electrode displacement

方法 Pc/% Po/% HC HO RF WE WF KNN+FE[30] 28.21±0.00 9.23±0.00 94.90±0.00 100±0.00 100±0.00 66.67±0.00 SVM+FE[31] 42.64±0.00 18.97±0.00 87.82±0.00 100±0.00 100±0.00 70.49±0.00 CNN[8] 90.72±6.97 13.42±7.82 23.21±10.15 96.79±2.58 96.75±5.29 64.18±2.73 CNN+FE 29.89±4.24 33.91±16.26 95.01±3.60 98.90±1.31 91.58±10.66 70.87±2.74 CNN+FE+ETD 85.56±4.44 40.33±13.50 75.20±8.05 98.85±1.18 86.18±5.24 78.24±2.26 RESNET+FE[32] 31.09±3.94 35.17±10.96 90.78±4.88 98.92±1.29 92.22±3.26 70.78±2.66 Bi-GRU[33] 84.10±8.19 5.41±5.09 72.67±10.26 100±0.00 99.34±1.09 72.04±2.35 LSTM[34] 73.64±11.84 4.33±1.99 71.13±18.33 100±0.00 99.40±1.02 70.88±3.55 TRANSFORMER[35] 80.64±5.84 20.15±3.53 66.17±8.62 98.55±1.38 89.14±2.19 66.14±3.25 LCNN[36] 93.13±2.98 12.51±3.53 81.03±6.54 98.26±0.52 91.08±4.90 76.38±0.94 TDACAPS[14] 97.59±2.62 74.87±15.71 95.18±4.42 98.56±1.82 55.39±16.87 84.77±0.82 ETDTBN 99.80±0.41 66.36±6.90 69.85±7.44 100±0.00 99.85±0.36 86.95±1.64