Research On Efficient Parameter Quantization Algorithms For Low-bit-rate Speech Coding

In the digital communication networks, speech coding is one of the most basic and important fields. In order to meet the urgent need in the fields of wireless communication, the 2.4 kb/s and even lower bit-rate (1.2,0.8,0.6kb/s) speech coding is always one of the most important research topics in the area of speech coding. In some band-restricted situations, lower bit rate speech coding algorithms (such as 0.3 and 0.15kb/s) are also in need.In this paper, based on SELP, several efficient quantizing algorithms used on the 0.3kb/s vocoder are introduced. Besides that, the primary implement of 0.15kb/s vocoder is also presented here.SELP is a high efficiency low bit-rate speech coding algorithm. In the process of speech coding and transmission, with the bit rate getting lower, the bits can be used will get fewer. So the effective technique of vector quantization will directly influence the intelligibility and accurateness of the reconstructed speech.Two enhanced parameter quantization methods are presented in this paper. One is the MSVQ with inter-frame and inter-stage prediction, and the other is joint optimization of codebook sizes for several parameters. The simulation results proved that these two methods can efficiently reduce the distortion of vector quantization and enhance the performance of reconstructed speech.The 0.15kb/s vocoder is one of the leading research topics in the field of speech coding. One way to construct the framework of 0.15kb/s vocoder is introduced in this paper, and the basic testing results are also given out. But much more deep research is still in need.