Student: Jamel Price
Senior Researcher: UMES
Title: Designing an Automatic Speech Recognition System using MATLAB
This paper provides an introduction to one of the many projects being conducted at the University of Maryland Eastern Shore as part of the Chesapeake Information-Based Aeronautics Consortium (CIBAC) summer program. The goal of this project is to build a simple, yet complete and representative automatic speaker recognition system using Matlab. Matlab is a data-manipulation software package that allows data to be analyzed and visualized using existing functions as well as user-designed programs. There are two fundamental phases of the speech recognition process: training and recognition. During the training phase a database of reference “fingerprints” are created in order to be tested against in the recognition phase. A fingerprint represents the most basic, yet unique, features of a particular speakers’ voice. A fingerprint is merely a matrix of numbers in which each number represents the energy or average power that is heard in a particular frequency band during a specific interval [2]. Initially, a simple speech database provided by the Swiss Federal Institute of Technology [1] will be used to design and test the system. After a fairly representative system has been created, actual cockpit recordings and voice commands will be used in order to further improve the system as well as show that the system is applicable in a real world environment. With the use of the third party application jMatlab, a java based interpreter interface for Matlab, a platform independent version of the speech recognition system can be created which is key for completing the aforementioned goal.
|
