基于耳周肌电信号的默念口令识别方法

The widespread use of smart devices has led to an increasing demand for wearable human–computer interaction technologies. To improve user acceptance, human–computer interaction technologies require high levels of interaction usability and concealment. This paper proposes a method for silent command recognition based on periauricular EMG signals. This method is easy to integrate with headphones equipped with integrated physiological signal acquisition, enables silent control of smart devices, and reduces social awkwardness. First, the command empirical principles are determined and constructed, and then the optimal command set is selected through screening. Second, the optimal periauricular sensor positions are chosen based on single-channel signal-to-noise ratio and classification accuracy. Third, a recognition model based on the CNN–Transformer structure is proposed to learn the spatiotemporal mapping between periauricular EMG signals and silent commands. Finally, extensive experiments evaluate the feasibility and stability of this method. Results demonstrate that the average accuracy of this method is 91.18%. The proposed method is superior to advanced models in similar tasks and is stable under command deformation and head motion. This method lays the technical foundation for commercial products of silent command recognition.