Codebook Design And Fast Algorithm Research In Embedded Variable Speech Coding

In order to compete for the next generation speech coding standar (G.VBR) of ITU (International Telecommunications Union), the research on embedded speech coding, which is the new project in the speech coding field, was launched and an embedded coding codec was developed at the same time. The bitstream produced by this codec is favorable for the transmission in network and efficiently resolve the packet loss problem.The embedded algebraic codebook structure is the foundation of the whole algorithm and is also the important content of this thesis. By adding the number of the pulses in algebraic codebook, the proposed algorithm can describe the excitation signal more precisely and acquire an embedded bit stream. The searching for the impulses of each layer is not independent but based on the impulses which have already found in lower layers. Moreover, there is a unique target signal for each layer, which makes the algerbraic codebook searching relatively independent and the contributiton of the enhancement layer can be used in update for the synthesise filter and target signal and results in a quality improvement in enhancement layers.An adaptive codebook structure and its update method are also proposed in this thesis. Under this new structure, all the ACELP layers have their own adaptive codebook buffers and use each layer's excitation to update its adaptive codebook. This kind of adaptive codebook structure is favorable for finding the best coding parameters for each layer and make different layer independent relatively. The embedded speech coder with this structure can achieve good performance both in core layer and enhancement layers.At last, the test results of the ACELP related layers in the proposed embedded speech codec were given, which shows that the candidate codec has fulfilled the performace requirements drafted by the ITU and achieved a good quality at the first three layers.