Kannada Speech Recognition using Discrete Wavelet Transform – PCA

M. A. Anusuya and S. K. Katti

Computer Science Department, SJCE, Mysore, India.


This paper, proposes a new scheme for recognition of isolated words in kannada Language speech, based on the Discrete Wavelet Transform and PCA. We first compute the Discrete wavelet transform of a speech signal and then we calculate LPC coefficients. For this Principal Component Analysis procedure is applied for speech recognition. This paper also presents the comparative results with respect to the results given in [11]. Our results are superior than the results presented in [11] with respect to recognition accuracy. It also shows, the results compared with the different wavelet families along with the best wavelet family performance.

Keywords: Discrete wavelet transform(DWT), Linear predictive coding(LPC), Principal component analysis(PCA), Kannada speech recognition, Euclidean distance.

     Back to TOC