Abstract
A noble system, which can recognize both finger and gesture signs and can synthesize theircorresponding voice, has been developed, in order to enable people unfamiliar with a sign language to make aconversation with hearing-handicapped persons. Finger and gesture signs are sensed both with bending andacceleration sensors. Acquired sensor data are transferred through wireless sensor networks to a host computer.Finger signs are recognized from bending information of finger, and gesture signs are discriminated by analyzingarm's movements by hidden Marcov model method. Finally, recognized signs are converted into correspondingvoices by a free software of Aques Talk or PWM function of microcomputers. As a result of evaluation, it has beenproved that 60-100% of finger signs and more than 90% of gesture signs consisting of six kinds of gesture aresuccessfully discriminated and converted into voices.
