Performances of surface EMG and Ultrasound signals in recognizing finger motion

Abstract

This paper compared the performances of the surface electromyography (sEMG) with Ultrasound(US) signals in finger motion recognition. Compared to traditional sEMG-based human machine interface (HMI), the ultrasound can provide some information about the morphological changes of muscles and has higher resolution. It is possible for the US-based HMI to recognize some more dexterous hand gesture especially for some finger motions. In the experiment, the subjects were instructed to perform 14 different finger motions. The sEMG signals and the ultrasound images were collected simultaneously. The 2-fold cross-validation with LDA classifier was used to analyze the accuracy. The mean accuracy of US-based HMI was nearly 96.37% while the sEMG was 92.41%. One-way ANOVA analysis showed that the US-based HMI had a significantly higher classification accuracy than sEMG-based HMI. What's more, the difference between individuals and finger motions of ultrasound was significantly smaller than that of sEMG. The US-based HMI performed better and more stable than sEMG, which may imply that ultrasound has a great potential to be an alternative method to sEMG regarding precise control, or it can corporate with sEMG to achieve a better HMI.