Research On Performance Improvement For ITU-T G.729

Digital speech has preponderance over analog speech in reliability,robustness and security during communication. However,digital speech needs more bandwidth than the analog signal. Especially with the requirement for communication frequency increasing,it's necessary to code speech signal at low rates. The aim of research on speech coding is to obtain high quality speech coding at medium and low rate,namely at rate of 16kb/s or below.This thesis chooses ITU-T G.729 as research object for its good all-around quality. G.729 not only has short code-decode delay but provides good synthesized speech at 8kb/s. Moreover it can work reliably in many circumstances. However G.729's computation complexity is a little higher for application. Taking the advantage of discontinuous transmission,we embed a new algorithm of voice active detection (VAD) into the original algorithm of G.729 to improve its efficiency. This algorithm can also reduce the average computation complexity of G.729. A VHDL implementation for this new algorithm is also provided in this thesis. The thesis is organized as following.Chapter I dissertates the scope of current study on speech coder and briefly describes the whole work. The reason for improvement scheme is also explained in this chapter. Chapter 2 overviews the fundament of ITU-T G.729. Us computational consumption during inactive voice period arouses us of the idea to apply VAD into G.729. Chapter 3 introduces some traditional VAD parameters. Experiment results show their shortages under varying noisy situations. Chapter 4 and chapter 5 detail our new algorithm and its implementation. To overcome the traditional VAD parameters' poor performance in non-stationary noise,a new VAD algorithm based on LP residue is proposed in chapter 4. Experimental results show that the new VAD algorithm improves G.729's overall performance in varying noisy situations. Experimental results of comparing with other VAD are also presented. In chapter 5 we design chip of this improved algorithm with VHDL. The synthesized result,namely the layout,can be easily downloaded into the FPGA to form an embedded system. This chip can be used in the fields of telecommunications and consumable electronic products. Chapter 6 gives the conclusion and proposes the future work.