Optimization And Implementation Of MPEG-4 AAC Audio Decoder

ACC is short for Advanced Audio Codec. Its' standard is made by Apple, SONY, AT&T, Dolby, Fraunhofer IIS company (The former MP3 standard makers) and other multimedia giants. As an advanced audio coding technology, AAC(Advanced Audio Coding) has become the audio coding standard of MPEG-4. AAC supports up to 48 main audio channels. Hardware platform DM6447 is the latest generation digital multimedia SOC(System on Chip) based on Da Vinci technology of IC giant Texas Instruments (Texas Instruments). DM6447 integrated a high-performance c64x+ DSP core and a ARM926EJ-S core. The strong performance of DM6447 makes the implementation of real-time MPEG-4 AAC decoder on it possible.First, this thesis introduces the development of AAC technology in recent years, then analyzes the modules of AAC decoder which include noiseless decoding, inverse quantization, rescaling, joint stereo decoding, temporal noise shaping, long time prediction and filterband. Besides, the means to optimize the inverse quantization module and the synthesis filterband module are given. With regard to the inverse quantization module optimization, this thesis uses a modified look-up table method. In addition, this thesis brings forward a method to optimize the synthesis filterbank module which uses IFFT method to calculate IMDCT and makes the time complexity from O ( N 2) down to O (( N )log( N / 4)). Besides, considering AAC decoder running on the C64x+ DSP side, for deriving the supreme performance, this thesis optimizes the program code of the AAC decoder from using some C language programming techniques and utilizing C6000 compiler aspects. The real-time AAC decoder is realized on DM6447 finally.In the end of this thesis , the result of the optimization of MPEG-4 AAC decoder and the analysis of the waveform of time domain and the spectra are given. Experiment results indicate that after being optimized,the AAC decoder has the ability to real time decode the audio bitstream, and the decoding efficiency is greatly improved after optimization. The spectrums before and after decoding are similar, and the playback effect is perfact.