Date of Publication :7th December 2016
Abstract: Despite the fact that Marathi language is most common languages worldwide, there has been only a little research on Marathi speech recognition as compared to other languages such as English and Chinese. The speech recognition algorithms are very useful for designing efficient and accurate speech recognition systems. In this paper, we have proposed the automatic speech recognition systems by using MFCC features compare by DTW template matching algorithm. Our proposed systems divided into three phases, in first phase, the input signal is preprocessed and reduce the noise. In second phase the features are extracted from the speech signal using MFCC algorithm. For improving the recognition accuracy, we have added the delta coefficient. And last, the features of each word in test database are compared to the training database using dynamic time warping (DTW) algorithm. The proposed systems achieve the recognition rate of about 96.5%, which is outperforming than other approaches.
Reference :