Practical Study and Implementation of an Isolated Word Speech Recognition System.

Document Type : Research Studies

Authors

Communication Engineering Department., Faculty of Engineering., El-Mansoura University., Mansoura., Egypt.

Abstract

The art and science of speech recognition have been advanced to the state where it is now possible to communicate reliably with a machine by speaking to it in a disciplined manner using a vocabulary of moderate size. It is the purpose of this paper to present a hardware implementation of a simple real time speech recognition system for isolated words. The system determines the energy of the spoken word through a bank of four channels band pass filters. The output from each filter is passed through a rectifier followed by a low pass filter to provide an average level which is proportional to the total signal energy within the particular filter band. The energy measurements from each channel is then passed to a personal computer where a recognition algorithm program is then executed. The program detects the starting and ending points of each spoken word. compares the energy features with the reference stored in the library, and chooses the most close pattern that matches the pattern of the spoken word. Results from experiments conducted by simulation and by the hardware prototype using Arabic speech show that the system provides recognition score between 76 % and 94%. 

Main Subjects