Automatic Isolated Speech Recognition System Using MFCC Analysis

This is all about a successfully implemented research project which is expected to be a commercial product in future. Objective of the research is to develop an Automatic Isolated Speech Recognition System (AISR) which will be able to classify four distinct words class enunciated as "GO", "STOP", "LEFT" and "RIGHT". Speech data are analyzed mainly with Mel Frequency Cepstrum Coefficients (MFCC) analysis along with some subsidiary DSP processing to create compact and prolific speech database. While Artificial Neural Network (ANN) is used as a classifier or recognizer to classify among the four distinct word classes. In order to justify feasibility and effectiveness of proposed ASR system a small scale application is implemented as well with designated speech classes.

Autorentext

Mohammad Faisal & Aasim Ullah are senior faculty members of International Islamic University of Chittagong(IIUC),Bangladesh. Daud Shakil, a bachelor student from same alma mater researched on Speech Recognition System under their supervision.