Study Of Algorithm Optimization For MPEG-4 AAC Real-time Audio Encode

As the most advanced perceptual audio coding technology, MEPG-4 AAC ensures the quality of audio coding and provides high performance of coding and compressing as well. It has become a widely-used audio coding standard. However, the AAC standard's algorithm complexity is so high that it can not meet the requirement of real-time coding without optimization. Therefore this thesis does the optimization research on its relative algorithms.Firstly, In order to satisfy the requirement of optimization for real-time performance and consider the high modulated characteristic of AAC standard, this thesis proposes optimal choices on encoder models considering computation cost, time complexity and actual profit without significantly affecting the quality of the audio coding. This approach reduces the intermediate steps of coding, simplify the structure of AAC encoder and thus decrease the computation complexity.Secondly, this thesis proposes a new time-domain peak-rate-changing-based window switching algorithm to eliminate the drawbacks of existing time-domain window switching algorithms by dynamic block and windows switching two-phase submission. This algorithm takes the characteristics of signal, makes use of time masking to improve the efficiency of switching algorithm, avoids the excessive use of short window and improves the efficiency of coding.Finally, to solve the problem caused by AAC quantify modules, this thesis proposes a simple efficient single-cycle algorithm structure using SDI algorithm to realize the self-adapted initialization of quantifying factors and perceptual entropy prediction. This algorithm significantly decreases the number of iterations of quantifying process and improves the coding efficiency and real-time performance of quantifying modules.