Abstract: In this paper, a Kalman post-filter approach is proposed to improve the performance of the Switch Split Vector Quantization（SSVQ） of LSF parameters. In the proposed approach, the correlations between the successive and sub LSF vectors are recovered by the Kalman filter in the decoder end. The parameters of the Kalman filter are adjusted according to the classification results in the first step of the SSVQ. Experimental results show that the proposed approach can further reduce the average spectral distortion of 0.01~0.02 dB when the average spectral distortion of the SSVQ is around 0.9~1.0 dB.

Key words: speech coding, line-spectrum frequency, Switch Split Vector Quantization, Kalman filter

摘要： 针对LSP参数在转换分裂矢量量化（Switch Split Vector Quantization，SSVQ）中未能充分利用子矢量间相关性的不足，提出了一种LSP参数SSVQ的卡尔曼后滤波增强方法。该方法在解码端利用卡尔曼滤波器来进一步发掘子矢量间和连续帧矢量间的相关性，并结合SSVQ中分类的转换来自适应地调整卡尔曼滤波器的参数。实验结果表明，方法可在SSVQ的平均频谱误差为0.9~1.0 dB时进一步减少0.01~0.02 dB。

关键词: 语音编码, 线谱频率, 转换分裂矢量量化, 卡尔曼滤波