Improved High-density Myoelectric Pattern Recognition Control Against Electrode Shift Using Data Augmentation and Dilated Convolutional Neural NetworkT

Improved High-density Myoelectric Pattern Recognition Control Against Electrode Shift Using Data Augmentation and Dilated Convolutional Neural NetworkT 553 217 Transactions on Neural Systems and Rehabilitation Engineering (TNSRE)

The objective of this work is to develop a robust method for myoelectric control towards alleviating the in-terference of electrode shift. Methods: In the proposed method, a preprocessing approach was first performed to convert high-den-sity surface electromyogram (HD-sEMG) signals into a series of images, and the electrode shift appeared as pixel shift in these im-ages. Next, a data augmentation approach was applied to the train-ing data from just one position (no shift), so as to simulate HD-sEMG images derived from fictitious shift positions. The dilated convolutional neural network (DCNN) was subsequently adopted for classification. Compared to common convolutional neural net-work, DCNN always contained a larger receptive field that was supposed to be adept at mining wider spatial contextual infor-mation in images. This property was further confirmed to facili-tate the classification of myoelectric patterns using HD-sEMG. The performance of the proposed method was evaluated with HD-sEMG data recorded by a 10 10 electrode array placed over forearm extensors of ten subjects during their performance of six wrist and finger extension tasks. Results: Under a variety of actual electrode shift conditions, the proposed method achieved a mean classification accuracy of 95.34%, and it outperformed other com-mon methods. Conclusion: This work demonstrated feasibility and usability of combining data augmentation and DCNN in predict-ing myoelectric patterns in the context of electrode shifts. Signifi-cance: The proposed method is a practical solution for robust my-oelectric control against electrode array shifts.