Automatic Speaker Recognition System
The proposed work provides a description of an Automatic Speaker Recognition System (ASR). It particularly documents all the stages involved in the proposed ASR system starting from the preprocessing stage to the decision making stage. The main aim of this work is to achieve a system with high robustness and user friendly. Voice samples from three different users are used as acoustic material. Feature extraction is done by computing Mel Frequency Cepstral Coefficients (MFCC) which is used to create reference template. For the purpose of feature matching, Dynamic Time Warping (DTW) algorithm is used wherein DTW distance is computed between the test signal and the reference signal. Decision is made by comparing the distance with a predefined threshold value.
KeywordsDiscrete Cosine Transform Speech Signal Dynamic Time Warping Speaker Verification Speaker Identification
Unable to display preview. Download preview PDF.
- 1.Abdulla, W., Chow, D., Sin, G.: Cross-words reference template for DTW-based speech recognition systems. In: Proc. IEEE TENCON, Bangalore, India (2003)Google Scholar
- 2.Furui, S.: Digital Speech Processing, Synthesis and Recognition. Marcel Dekker, New York (2001)Google Scholar
- 3.Ezzaidi, H., Rouat, J., O’Shaughnessy, D.: Towards combining pitch and MFCC for speaker identification systems. Presented at the Eurospeech Conf., Aalborg, Denmark (2001) Paper No. 2825 (unpublished)Google Scholar
- 4.Boves, L., et al.: Design and Recording of Large Data Bases for Use in Speaker Verification and Identification. In: ESCA Workshop on Automatic Speaker Recognition, Martigny (CH), pp. 43–46 (1994)Google Scholar
- 5.Staroniewicz, P., Majevski, W.: SVM Based Text-Dependent Speaker Identification For Large Set of Voices. In: Proc. 12th European Signal Processing Conference (EUSIPCO), vol. 1, pp. 333–336 (2004)Google Scholar
- 6.Reynolds, D.A.: Experimental evaluation of features for robust speaker identification. IEEE Transactions on Publication, Speech and Audio Processing 2(4), 63964 (1994)Google Scholar