Research Of Low Bits Rate Wideband Speech Codec

With the development of communications technology and information technology, wideband speech would be applied to videoconference, conference call, as well as to multi-user/ multi-channel communications domain of ISDN. The resource of frequency band is more and more invaluableness, the demands for transmission efficiency and storage of speech grow higher and higher. One of the means to resolve these problems is compressing speech signal, using fewer data and lowest distortion to represent the original speech, namely carry through speech codec. Therefore the study of speech coding, especially for the developing of wideband speech, the study of low rates wideband speech coding has been playing a very important role for practicality significance.In this thesis we sum up the actuality of wideband speech coding development at present, and detail adoption different technology of coding to implication wideband speech codec. Coding rates from G.722(ADPCM) 64kb/s which is wideband speech codec standard put forward by CCITTreduce to 16kb/s 13 kb/s which are realized by CELP coding arithmetic that is usually adoptive currently. G.729 standard and the principle of Mixed Excitation Line Prediction (MELP) coding decoding are completely and systemic introduced. The theory of sub-band coding is researched, basing on the theory of sub-band coding, so 14.4kbit/s and 10.4kbit/s wideband speech coding can be simulated to realization.Realization two different rates scheme for lowness bits rate wideband speech coding are presented. According to the theory of sub-band coding, 0~8kHz band is split into two equal bands, lower band 0~4kHz and higher band 4~8kHz. When imitation realization 14.4kbit/s coding the lower band adopt G.729A and higher band adopt low bits rate G.729D. For the sake of depressing the bits rate even more. When imitation realization 10.4kbit/s coding, the higher band adopt the 2.4kbit/s MELP low speed speech codec. The synthesize speech which is put out come by low bits rate wideband coding can reach MOS 3.5 by audition test. That indicate the quality of synthesize speech can achieve scheduled request.