Two-dimensional cepstral distance measure for speech recognition

The weighting factors for the coefficients of the two-dimensional cepstrum (TDC) are derived. By using the proposed distance measure, a vector-quantization (VQ)-based speech recognition method is introduced. The advantage of this speech recognition method is its simplicity in implementation. The effect of the weighting factors on the TDC distance measure is also investigated. Experimental results show that the TDC distance measure is very efficient for isolated-word speech recognition. The proposed method is very promising for syllabic languages such as Mandarin Chinese.<>