Efficient Electromyography-Based Typing System: Towards a Novel Approach to HCI Text Input.

Area of Science:

• Biomedical Engineering
• Human-Computer Interaction
• Signal Processing

Background:

• Electromyography (EMG) is effective for static gestures and medical uses.
• Real-time EMG applications like typing are challenging due to continuous signal-to-discrete output conversion.
• Existing methods struggle with the accuracy and speed required for interactive typing.

Purpose of the Study:

• To develop a novel EMG typing system for real-time, keyboard-free interaction.
• To address the challenge of continuous EMG signal processing for discrete typing output.
• To improve the accuracy and response time of EMG-based typing systems.

Main Methods:

• Utilized Connectionist Temporal Classification (CTC) for efficient continuous EMG signal recognition.
• Implemented a parallel inference approach to enhance recognition accuracy.
• Developed and tested a prototype EMG typing system.

Main Results:

• Achieved a character error rate of 3.8% on the test set.
• Reported a word error rate of 7.1%.
• Demonstrated a system response time of less than 100 milliseconds.

Conclusions:

• The novel EMG typing system is feasible for real-time, keyboard-free interaction.
• The system shows significant potential for advancing human-computer interaction.
• The combination of CTC and parallel inference effectively handles continuous EMG signals for typing.