Research Of MPEG-4 AAC Codec Based On Risc

As a digital audio compression standard, MPEG-2/4 AAC(Advanced Audio Coding) already becomes the core algorithm of the high quality audio coder of the MPEG-4 Standard, and represents the trends of perceptual audio coding technology. To the multi-channel audio signals, MPEG-2/4 AAC can offer extremely high quality audio with bitrate at 64kbps per channel, so it can be applied in many fields.MPEG-2/4 AAC is one of the Hi-Fi audio coding standards, many state-of-arts coding tools, such as Modified Discrete Cosine Transform (MDCT), Time Domain Prediction, Temporal Noise Shaping, and so on, are consisted in this standard. The psychoacoustic model of humanity auditory system and information statistic theory are utilized to get maximum compress bitrate with the precondition of perfect reconstruct quality. For different application fields and software/hardware configuration, MPEG-2/4 AAC can provide different coding/decoding profiles. Different coding tool in different profile requires different computational complexity and memory size. So MPEG-2/4 AAC can provide all-round solutions to the storage and transfer of Hi-Fi audio. This dissertation is target for design an audio codec base on single MCU, and providing a low power, low cost, extensible audio solution.In this dissertation, the MPEG-4 AAC coder is selected as the research target, especially focused on the very computational consuming modules, such as psychoacoustic analysis, filterbank, quantization and coding module of perceptual audio coding technology. The highest computational complexity is the quantization and coding module. It is based on the "two iteration loop" algorithm: the scalefactors are adjusted in different direction by the outer/inner loop, eventually converge on the state of the consumation of coded bit under the constraint of bit reservior and the quantization noise under the signal masking threshold. For the reason of reduction in the numbers of two iteration loops, the initial value and the adjustment algorithm of scalefactors are modified. The coding module is extracted from inner iteration loop to improve coding efficieny. There has a same funtion module in psychoacoustic analysis module and filterbank module, which can transfer the time domain signal to frequency domain, has higher computational complexity. Just one of them is selected in this optimized coding scheme, the procedure of psychoacoustic analysis module is also modified for coding efficiency.MCU selection is the foundation of audio codec. The fundamental instruction set architecture and system architecture of processor RISC3201 are firstly introduced in detail in the 3th part of this dissertation, the aids of multimedia instruction set extension to the improvement of multimedia adaptation is also presented. Then the general procedure and technique of audio implementation on a fixed point system are proposed. At last, the function and characteristic of each module of the MPEG-4 AAC LC profile decoder are analyzed. The fix-pointed decoder algorithm and its optimized implementation are provided, with concerned decode results; the scheme of optimization in assembly style based on RISC3201 and decoder results are also provided.At last, the platform of hardware decoder implementation based on FPGA with the multimedia processor core RISC3201 is discussed in detail. On this platform, the MPEG-4 AAC LC test bitstream can be decoded in real time. The average instruction and average cycle time of decode one audio frame are obtained too. Finally, the instruction set extension which can improve audio decoder efficiency based on RISC3201 microprocessor is putted forward.