Techniques for variable rate speech coding using wavelet representations

This paper presents two techniques for variable rate speech coding using wavelet representations. In the first method, the Daubechies wavelet is used to produce a wavelet representation of the speech signal. This representation is encoded using some of the properties of the speech signal to allocate bits to the various subbands. A technique to lower the bit rate using vector quantisation is discussed and perceptual results are reported. The second technique caters to the encoding of speech in the presence of road noise, as in the cellular environment. It has been observed that in the presence of noise, the wavelet extrema representation provides a “denoising effect”, and the temporal correlation of the speech signal can be exploited. A modification of Vetterli´s projection algorithm using spline interpolation has been found to give perceptually better results for speech encoded in the presence of noise. Examples of signals encoded by this method are shown. Preliminary results indicate that the first technique performs well for clean speech (comparable to GSM), and the second exhibits robustness to noise with quality almost comparable to GSM