Research On CNG Algorithm Based On Mask Effect Theory

With the rapid development of communication technique, the speech coding is gradually updating from narrow band to wideband. However, the expansion of bandwidth will increase the transmission rate obviously. As non-voice time accounts for a large proportion of the whole process of voice communications, it is a waste of resources for the networks and terminals using the same coding rate. Therefore, we use CNG technology in the non-voice time, encoding with low rate and decoding a comfort background noise.Firstly, the thesis analyzes the theory of CNG technology, and introduces two of the representative recommendated standards: G.729 Annex B and G.722.2 (AWR-WB) Annex A. The traditional speech coding technology for both narrowband and wideband uses the CELP model, and the CNG technology is mostly based on the simplified model of CELP. However, the audio coding technology for ultra wideband mostly uses perceptual coding which is based on the psychoacoustic model. And perceptual coding is based on mask effect theory by adding quantified noise. As a result, the thesis analyzes the mask effect theory, and discusses its application in audio encoding.By discussing the existing voice and audio encoding technology, the thesis designs an CNG algorithm which is based on the mask effect theory. And the performance of the CNG codec is tested. The synthesized comfort noise is satisfactory.