Surface Electromyography (sEMG) is a non-invasive, computer-based technique that records the electrical impulses.  

However, the present pattern-recognition-based control strategy can realize some myoelectric control but it is not as smooth as human hand.

Recently, researchers from the Shenzhen Institutes of Advanced Technology (SIAT) of the Chinese Academy of Sciences, proposed a continuous estimation method for 6 daily grasp movements by LSTM  (Long short-term memory network), and the comparison tests showed that LSTM held the best performance.  

According to a study published in Biomedical Signal Processing and Control, the team designed an experiments on six daily grasp movements, these six movements were selected in the light of different shapes and diameters of the objects, and 22 sensors were spaced around on a CyberGlove for recording sEMG signals.  

To estimate six grasp movements, researchers carried out the tests through 3 evaluation criterions, Pearson Correlation Coefficient (CC), Root Mean SquareError (RMSE) and Normalized Root Mean Square Error (NRMSE), and compared LSTM with other two algorithms, SPGP (Sparse Gaussian Processes using Pseudo-inputs) and RBF (Radial Basis Function Neural Network), the results exhibited that LSTM could performed better as well as faster in all 6 movements. 

Although, in some joints, SPGP or RBF has the better performance than LSTM, the statistical analysis showed that LSTM could perform better in continuous estimation of 20 finger joint angles than SPGP and RBF. 

"Our results shows a bright prospect of LSTM, which can be used in bioelectrical signals processing and human-machine-interaction.” Said Dr. LIN, “It should be noted that base on different applications, the method should be personalized and optimized.” 

Figure1. Summary of the NRMSE of LSTM, RBF and SPGP for 6 movements. (Image by LIN Chuang) 

Figure2. The chain structure with repetitive modules of LSTM. (Image by SIAT) 

Media Contact:
ZHANG Xiaomin
Email: xm.zhang@siat.ac.cn