A Study Of G.718Audio Codec

With the continuous development of the speech coding technology, a series of codingstandards aiming at getting a low delay have been proposed. This includes recommendationITU-T G.728(1992), namely,16kbit/s LD-CELP, which obtains an overall delay less than2ms; recommendation ITU-T G.729(1996) whose bit rate is8kbit/s and delay is15ms withdecoded speech quality worse than that of G.728.ITU-T G.718is a new narrowband (NB) and wideband (WB) embedded variablebit-rate speech and audio coding standard, established by ITU-T in June2008.Specifically,the embedded hierarchical structure is used in the codec. It has a bit rate from8to32kbit/sand the overall algorithmic delay is42.875ms (for WB input and WB output) or43.875ms(for NB input and NB output). Research has been done previously to reduce the core delayand a scheme has been proposed and it can receive a delay that is25ms lower than that inG.718with a frame length of5ms.Concerning about the quality decline in reconstructed speech of the scheme mentionedabove, an improved scheme is proposed in this paper, presenting a better decoded speechquality than the scheme proposed earlier at the expense of10ms more delay. In this case,the scheme is based on a frame length of5ms and for WB input and output, the core layerdelay is17.875ms, for NB input and output, the delay is18.875ms.The results fromregarding tests show that there is an improvement in reconstructed speech than that in theearlier approach.