Improvement of PNS Tools for MPEG-4 General Audio Coding

• Main Authors: Ji-da Chen, 陳繼大
• Other Authors: Hsueh-Ming Hang
• Format: Others
• Language: en_US
• Published: 2002
• Online Access: http://ndltd.ncl.edu.tw/handle/48743806671736288617

碩士 === 國立交通大學 === 電子工程系 === 90 === MPEG-4 is the latest multimedia compression standard defined by ISO/IEC MPEG (Moving Picture Expert Group). Its audio part, GA (General Audio), aims at data rates from 12kbit/s to 64kbit/s for encoding a pair of stereo channels to 320kb/s for encoding five-channel audio with ‘indistinguishable quality’. The basic structure of MPEG-4 GA is the same as that of the MPEG-2 AAC (Advance Audio Coding), the previous MPEG audio standard. Yet, it adds in several tools to improve its coding efficiency. This thesis selects one of the new tools, PNS (Perceptual Noise Substitution), as the research topic. We try to improve the audio coding efficiency by the proper use of this tool. PNS is an efficient method of coding the noise-like components in an audio signal. Although the data format and the decoder operation of PNS have been standardized, the method of choosing PNS parameters and the decision of using PNS or not are not standardized. In addition, we find that if we do not reduce the overhead of PNS, it would take up too many bits and thus decrease the coding efficiency. Therefore, we propose an effective and efficient procedure that detects the noise part of sound, and add a decision rule inside the bit-allocation loop to decide if PNS should be used. Simulations indicate that the new procedure saves bits in coding the noise-like components and thus may increase the overall audio coding performance.