Abstracts Submitted: 257
Number of Users: 304
View Abstracts Submitted
Back to home Page
Automatic speech recognition in a noisy environment is a challenging research problem even today. The paper presents a new method for robust recognition of Malayalam speech in noisy environment using DWT based Visual Features. The lip reading method is used to comprehend spoken utterances in noisy environment. The ability of lip reading enables the individuals with hearing impairment to communicate with the society. The lip imitation skill can be improved with the help of a lip-reading application. This paper aims to create a preliminary analysis on the lip-reading application for the recognition of Malayalam speech. As a part of this experiment we have created a dataset of Malayalam speech from 50 participants. Viola Jones Object Detection (VJOD) algorithm is used in preprocessing steps for facial recognition and lip localization. The images of a mouth are obtained after the lip localization and then it’s used for feature exaction using Discrete Wavelet Transform (DWT). Fifty percent samples collected is used for training the HMM model and other fifty percent samples are used for testing the model. The results show that the DWT based on visual features gives a good recognition rate for the Malayalam speech in a noisy environment.
© Copyright 2017 All Rights Reserved