Abstract
This paper proposes a general design framework for a wrist-worn surface electromyography (sEMG) device to support one-handed typing in Mixed Reality (MR) environments. Current MR interaction methods-such as midair gestures, gaze input, and handheld controllers-are inaccessible to users with motor limitations due to their physical demands and limited feedback. To address this, we present a compact and battery-powered sEMG system that detects thumb-to-finger gestures and maps them to a virtual T9inspired keyboard layout distributed across the user's palm and fingers. The proposed system uses six sEMG sensors placed beneath the palmar side of the fingers, combined with an embedded nRF52840 module for wireless signal acquisition and filtering. Each gesture corresponds to one of twelve anatomical zones representing character groups. Signals are preprocessed with bandpass and notch filters, segmented with sliding windows, and transformed using time-domain features such as RMS, MAV, and waveform length. A lightweight convolutional neural network (CNN) is employed to classify gestures with realtime capability. The framework supports per-user calibration to address signal variability and includes a structured plan for collecting gesture data from both healthy and disabled participants. This approach enables tactile, efficient, and accessible typing in MR using a single hand. It reduces physical fatigue and enhances usability for individuals with reduced dexterity. The system is designed in alignment with United Nations Sustainable Development Goal 10.2, promoting inclusive access to digital environments. Future work includes prototyping, model training, and usability testing under realworld MR scenarios.

This publication has 24 references indexed in Scilit: