Aiming at the limitation of the existing hand gesture recognition methods, considering the sparse multi-channel feature of MYO armband, this paper proposes a hand gesture recognition scheme which combines surface electromyogram (sEMG) and acceleration (ACC) information. Firstly, 5 participants wore MYO armbands, and simultaneously collected 8 groups of sEMG and ACC data of different gestures. Secondly, a new data preprocessing method is proposed, which will segment multi-channel sEMG and ACC data through high-energy sliding windows to obtain short-term effective active parts. Finally, a parallel sequence LSTM network is designed to extract and fuse the features of the two types of data. The accuracy of gesture recognition is 96.87%. The results show that the proposed scheme is simple and feasible.