Multi channels sEMG-based joint angles estimation of lower limbs utilizing bidirectional recurrent neural network

Nowadays, rehabilitative robots that have received more attention in the field of rehabilitation can help patients in the rehabilitation training and reduce therapist workload. This paper suggests the use of surface electromyography (sEMG) signals and a bidirectional neural network (BRNN) for the estimation of the joint angles of lower limbs. The input of BRNN is the preprocessed sEMG signals and its outputs are the estimated joint angles of knee, ankle and hip. In order to prove the usefulness of the bidirectional neural network, four normal and healthy subjects and two spinal cord injury (SCI) patients take a part in the experimental tests. The healthy subjects exercise two movement modes containing the leg extension and the treadmill at various loads and speeds, while the SCI subjects conduct only the treadmill exercise. In order to record useful information, seven leg muscles were used and then the hip, knee and ankle joint angles were acquired at the same time. The experimental results show the satisfactory performance of the proposed method in the estimation of joint angles by employing surface electromyography signals for both groups. The proposed estimation method can be used for controlling the rehabilitation robot of SCI subjects based on sEMG signals.

Keywords:

joint angle estimation , sEMG signals , rehabilitation robots , bidirectional recurrent neural network