Design and Implementation of a Spoken Letter recognition system for controlling the Upper Prosthetics

Abstract

The humanoid robot represents a promising application in the medical, handicapped and industrial limbs fields. The control of the hand fingers of the spoken command is one of the importance of these applications. This paper presents the proposed design of the hand-fingers controls circuit through the spoken letter. This design is based on the control of servo motors for each finger by Arduino controller. The Arduino has received a coded signal from the recognition algorithm to give the signals with different' duration to the servo motors to take the required hand-fingers form which is corresponding to the spoken letters. Where these codes relate to the statistical features of that letter, which are extracted from the spoken signals. Then, by using the Multilayer Feed-Forward Neural Network algorithm as a classifier, the simulation recognition, which has been gained is 96.33% while the experimental recognition is 94.8 %. The difference between these results is caused by the reject some patterns caused by the motor's errors.