| Speech communication plays an important role in modern communication.G.729 that was issued by ITU-T in 1996 is widely used owning to its high compression rate and excellent quality in synthetic speech. Several algorithms such as voice activation detection, are added in G.729B in order to decrease the code rate and improve the spectral efficiency. However, Low bit-rate voice coding technology has became a key to voice communication because of the scarcity bandwidth resource.As DSP technology develops,application system with DSP core based is widely used in many areas,so how to make the speech coding run quickly in the DSP platform is important to voice technology and expand its applications.This paper achives the transplante to DM642 and optimizes G.729B algorithm according to the characteristic of C64X DSP, then improves the voice activity detection algorithm of G.729B.The main researches and innovations are as follows:Firstly, on the basis of studies on voice activation detection algorithm, a new detecting algorithm is proposed to improve performance to against the weak point that the performance of voice activation detection algorithm will get worse in low Signal Noise Ratio.This method detects the voice according to Higher Order Cumulant and other the autocorrelation of voice,with which a better performance is obtained in low Signal Noise Ratio and complex background noise environment.Secondly, this thesis has done a lot of research that focuses on the Structural characteristics of DM642 and AIC23B in the DSP platform. And also,the G.729B standard has been transplanted successfully in DSP platform.Particularly, G.729B real time system with TMS320DM642 as the system core has been achieved in CCS development environment.Thirdly, TMS320DM642 DSP system platform is optimized. A modified algorithm is presented under assembly level.The numerical results show that the code bit rate is improved obviously after optimizing to satisfy the real time system without sacrificing performance. |
PDF Full Text Request