Porting And Optimization Of AAC-LC Algorithm For X86and ARM Platforms

As the most advanced perceptual audio coding technology, MPEG-4AAC hasbeen applied widely in the transmission and storage of digital audio signal and it is thecore of high quality audio coding in MPEG-4. Compared with other MPEG audiocoding standards, it has higher compression ratio, better compression quality and moreflexible sound track configuration. However, the performance improvement is at theexpense of computation complexity and bad real time feature. So it is necessary andmeaningful to optimize the algorithms of AAC both in academic and industrialpurposes.Among all the profiles of AAC coding algorithms, the AAC-LC (AAC LowComplexity) has lower calculation complexity and therefore it is suitable for applicationin which the hardware only has lower processing power. The theory of perceptualcoding for general audio is presented in the thesis first, then structure and principle ofAAC-LC encoder is introduced, next the methods for optimizing the functional blocksin AAC-LC encoder algorithm is described, the ways to porting and optimization of thealgorithm to X86and ARM platforms are also detailed in the thesis. In the AAC-LCaudio coding algorithm, the complexity of encoder is much higher than that of thedecoder; and in the encoder, the psychoacoustic model, filterbank, frequency prediction,Mid/Side Stereo, temporal noise shaping and quantization modules play a important rolein the algorithm and take the main parts of calculation load of the encoder. As a result,the thesis pays great attentions to the encoder and made big progresses in optimizationfor them.Performance evaluation has shown that, while keeping the audio encoding qualityunchanged, the calculation complexity of the encoder for X86and ARM have beenreduced more than30%and75%respectively, the optimized algorithm is able to beused in real applications.