Marathi Digit Recognition Using Lip Geometric Shape Features and Dynamic Time Warping

The aim of our proposed research work is to identify language of spoken utterance using visual speech recognition and include Marathi language in language identification (LID) system. In this paper we have focused on the task of identifying first three digits in Marathi language. For this first Lips are extracted from video frames of face images and then landmark points on the lips are detected. Then lip features like height, width, area are calculated using these landmark points. In this paper we have analyse different geometric shape features of lips like height, width, area and height width ratio in the task of first three Marathi Digit Recognition using visual speech recognition. Dynamic Time warping is used for classification of time series sequence of each utterance. The result of the experiment shows that the recognition rate with feature Delta W (width) is nearly 63% which is highest as compare to other features.