Finger Movement Recognition During Ballistic Movements Using Electromyography

Individuals who perform repetitive and spontaneous movements with their fingers such as typing on a keyboard or playing instruments are susceptible to repetitive stress injuries. We denote such movements as “ballistic gestures”. The most common injury due to ballistic gestures is carpal tunnel syndrome, which occurs when typing on a keyboard using an improper typing style or hand posture. It then becomes important to pro-actively analyze an individual's finger movements and recommend changes according to their habits to prevent such injuries. In this work, we present a Gaussian mixture Hidden Markov Model classification method to classify finger movement during typing activities with an accuracy above 97%. Indeed, the model maintains such accuracy across typing speeds and different individuals. Further, we show the effects that sampling rate, electrode placement, typing speed, and movement epenthesis have on classification accuracy in ballistic gestures.