Trace Finger Kinematics from Surface Electromyography by Using Kalman Decoding Method

Decoding hand motion intention from surface electromyography (sEMG) is facing with challenges brought by the requirements of multiple DOF and flexibility. Compared with discrete action classification based on pattern recognition, continuous kinematics estimation is considered more natural and intuitive. However, the accuracy of the existing estimation techniques is limited and the computational burden is large. Kalman filtering is an algorithm that uses linear system state equations to optimally estimate system state through system input and output observation data. It is easy to implement computer programming and can update and process the data collected on the spot in real time. In this study, the control process of human upper limb motion was considered as a linear dynamic system, in which the joint angle and the surface EMG were considered as the system state variable and the system observed variable, respectively. A Kalman decoding method was introduced for finger kinematics estimation. The model for finger joint angle estimation from the sEMG signal was established by training data set. And then the joint angle estimation is carried out according to the Kalman filter iterative method. The effects of the method were verified on the publicly available database, getting average correlation coefficient (CC) value 0.70 with the max value 0.81 and the minimum value 0.50. The establishment of the Kalman estimation model can promote its extended research and application in the field of finger motion continuous estimation technology.