Sunil S. Morade¹ and SupravaPatnaik², ¹SVNIT, India and ²The Xavier Institute of Engineering, India

In lip reading, selection of features and classifier plays crucial roles. Goal of this work is to compare the common feature extraction modules and classifiers. Two well-known image transformed models, namely Discrete Cosine Transform (DCT) and Discrete Wavelet Transform (DWT) are studied. A competent feature extraction module cascaded with a robust classifier can result a novel automatic lip reading system. We have compared performance of Back Propagation Neural Network (BPNN) algorithm with that of K-Nearest Neighborhood (KNN) algorithm. Both being from class of artificial intelligence needs training.Hence we have also examined the computational complexity associated with the training phase of both classifiers. The CUAVE database is used for experimentation and performance comparison. It is observed that BPNN is a better classifier than KNN.

Discrete Cosine Transform (DCT), Discrete Wavelet Transform (DWT), K Nearest Neighborhood (KNN),lip reading, BP Neural Network (BPNN)