Robust spectrum quantization for LP parameter enhancement
Volodya Grancharov, Sriram Srinivasan, Jonas Samuelsson, Willem Bastiaan Kleijn · 2004
In this paper, we investigate the denoising properties of ro-bust vector quantization of the speech spectrum parameters in combination with a Kalman lter. The underlying assump-tion is that the high-energy speech regions can be used to re-construct the low-energy regions destroyed by noise. This can be achieved through vector quantization with a properly weighted distortion measure. The performance of the pro-posed system, Kalman ltering with prior vector quantiza-tion, is compared with existing schemes for parameter esti-mation used in Kalman ltering. The results indicate signif-icant improvement over the reference systems in both objec-tive and subjective tests. 1.