Environmental noise analysis for robust automatic speech recognition

Abstract

Most of the speech communication applications viz. telephony, handsfree communication, voice recording, automatic speech recognition, interactive voice response system, human-machine interfaces, etc. that require at least one microphone, desired speech signal is usually contaminated by background noise and reverberation. As a result, the speech signal has to be ‘‘cleaned’’ with digital signal processing tools before it is played out, transmitted, or stored. The noise estimation and reduction techniques will help to clean and attenuate the noise component in speech data, known as Speech Enhancement. In this paper, we recorded the speech in different environmental conditions and estimated the noise signal/background noise distribution in speech. Now the speech is enhanced by using the compliment of Weiner-Hopf optimal filter. And this enhanced speech signal is given for training and testing the Automatic Speech Recognition (ASR) system, which will improve the word accuracy. The analysis of speech and results presented in this paper are produced using MATLAB.