The Optimization And Realization Of G.729A-Based Speech Codec On IOS Platfrom

In recent years, IP telephony, wireless communications and digital satellite systems and other services for voice applications continued to emerge, and speech coding technology has made many breakthroughs. There are some practical coding schemes proposed by researchers such as G.711, G.721, G.729, etc. By comparing several of these programs, we use the G.729encoding scheme. G.729algorithms used speech coding of conjugate structure-ACELP. G.729A was its simplified program with complexity reduced by nearly half. Although G.729A had simplified the algorithms, its complexity was still relatively high. So it can’t be directly applied in practical applications that need real-time requirements. The solution is to optimize the algorithms. The main contents in this paper include:(1) Through analyzing the whole module of G.729A algorithms, it showed the encoder consisted of pre-processing, LP analysis, open-loop pitch search, closed-loop pitch search, fixed codebook search and memory update, and the peak of operation mainly focused on the operation of LP parameters and codebook search, occupying more than60%of the time. Therefore, this article focused on the optimization of the open-loop and closed-loop pitch search, and after analysis of the algorithm structure, we improved the adaptive codebook search model and using a4:1extraction method for sampling signal points and the best latency search point. Moreover, the voice data was processed by serial module which made the compression ratio not be16:1in theory, so the conversion processing from lbit to Word16was moved in this paper.(2) Optimization of G.729A need to achieve through the platform, and existing studies were mostly based on DSP and other hardware platforms, but the cost was high. And the software-based implementation is relatively flexible with low cost, combined with existing resources, this paper chooses iOS platform for the optimization of G.729A, namely, we develop a voice call software on iPhone. Finally, for testing, G.729A was transplanted to the iOS platform and applied in software of voice call.Through the final testing results and analysis, we know that the improved method of this paper reduces the complexity of G.729A algorithm, improved the coding time, and did improve the real time capabilities with slightly decline in speech quality.