Position-independent decoding of movement intention for proportional myoelectric interfaces

In this decade, myoelectric interfaces based on pattern recognition have gained considerable attention thanks to their naturalness enabling human intentions to be conveyed to and in control of a machine. However, the high variations of electromyogram signal patterns caused by arm position changes prohibit application to the real world. In this paper, we propose a novel method of decoding movement intentions robust to arm position changes towards proportional myoelectric interfaces. Specifically, we devise the position-independent decoding that estimates the likelihood of different arm positions, which we predefine during a training step, and also decodes the movement intention in a unified framework. The proposed method has an advantage that could be used to decode the movement intentions on untrained arm positions in a realistic scenario. Our experimental results showed that the proposed method could successfully decode the continuous movement intentions (e.g., flexion/extension and radial/ulnar deviation) on both trained and untrained arm positions. Our study also proved the effectiveness of the proposed method by comparing the existing methods in terms of the decoded trajectories as movement intentions in untrained arm positions.