A study on a very low bit-rate speech coder based upon self similarity of speech signal.

点击翻译按钮获取中文摘要

In this research, a new segment quantization method for coding LSP coefficients of speech signal. In order to design an efficient segment coder, a segmentation method using the self similarity of LSP coefficients is employed. This method reduces redundant information between LSP frames by using a segmentation scheme which is similar to the Lempel-Ziv coding method.As experimental results, the proposed method was able to quantize the LSP test set at lower bit rate with lower cepstral distortion, in comparison with two conventional vector quantizer (Split-VQ and MSVQ), and its performance was comparable to a vector quantizer with MA prediction.