Abstract
Our study introduces a surface electromyography (sEMG)-based Cheonjiin speller system developed to assist individuals with restricted hand mobility. The interface incorporates a directional control framework-comprising up, down, left, right, and select commands-integrated with a Korean keyboard layout to enable efficient and accessible text input. Two-channel surface EMG signals were recorded from the rectus femoris and gastrocnemius muscles at a sampling rate of 200 Hz using an EMG acquisition module. The signals were processed in real time using notch and bandpass filtering, followed by full-wave rectification. To decode user intent, three physiologically interpretable time-domain features-root mean square (RMS), slope sign change (SSC), and peak amplitude-were extracted and subsequently used for classification. The Cheonjiin speller was implemented in Python 3.10.8 and operated through directional cursor navigation. System performance was quantitatively evaluated in two experiments: in Experiment 1, recognition accuracy for five discrete commands reached an average of 90.0%, while Experiment 2, involving continuous Korean word and sentence input, achieved an average accuracy of 88.65%. Across both experimental conditions, the system attained an average information transfer rate (ITR) of 96.19 bits/min, confirming efficient real-time communication capability. The results demonstrate that high recognition performance can be achieved using simple, low-computation features without deep learning models, confirming the feasibility of real-time implementation in resource-limited environments. Overall, the proposed speller system exhibits high operability, accessibility, and practical usability in constrained conditions and holds potential for integration into augmentative and alternative communication (AAC) systems for users with motor impairments. Moreover, its lightweight architecture, minimal computational load, and flexible directional control structure make it adaptable to a wide range of assistive and wearable technology applications.