Spectral Features based Speech Recognition for Speech Interfacing to Control PC Windows

In this research work Speech interfacing for computer control has been developed in MATLAB platform that enables physically disabled people also to gain access to computer. This speech interfacing enables users to execute common computer commands by spoken utterances and not by using standard input devices such as keyboards and mouse. The approach used, in this work for computer control is direct execution of recognized vocal command rather than the mouse movement control or the text matching of text between one from application menu and the other that is produced by speech recognition system. The work is based on PC Windows (windows 7), and is carried out in English language. In this work, recognition of speech is carried out by Gaussian Mixture Model (GMM). Mel Frequency Cepstral Coefficients (MFCC) are used for transferring input speech into spectral components. Optimum recognition is achieved with the performance comparison of the GMMs by varying the number of MFCC, Gausses (components) and iterations. Also, the average speech command recognition performance is observed to be 74.38 percent.