Hand sign classification techniques based on forearm electromyogram signals

This paper describes classification techniques to distinguish hand signs based only on electromyogram signals of a forearm. Relationship between finger gesture and forearm electromyogram is investigated by two signal processing approaches; an empirical thresholding method and meta heuristic method. The former method judges muscle activity according to the criteria experimentally determined in advance, and evaluates activity pattern of muscles. The latter learns the electromyogram characteristics and automatically creates classification algorithm applying genetic programming. Discrimination experiments of typical hand signs are carried out to evaluate the effectiveness of the proposed methods.