Automatic speech recognition system of tamil language using linear discriminant analysis

Author: 
Sundarapandiyan S and Shanthi N

This paper presents a method of dimensionality reduction in Tamil language speech recognition system. The method uses Linear Discriminant analysis (LDA). The intent of using LDA is to reduce the training time of the acoustical model also reduces the size of the feature vector. In General an automatic speech recognition system uses high dimensional acoustic vectors for train the system. LDA converts high dimensional acoustic feature vector to low-dimensional acoustic feature vector. The system uses the standard ELDA Tamil corpus. An automatic speech recognition system using LDA produces better result than the other dimensionality reduction techniques such as Principal Component Analysis (PCA), Multi dimensional Scaling(MDS), Locally-Linear Embedding(LLE) used speech recognition systems.

Download PDF: 
DOI: 
DOI: http://dx.doi.org/10.24327/ijcar.2017.6301.0913
Select Volume: 
Volume6