The Research And Design Of Audio Decoder Based On Software And Hardware Cooperation

Currently, the popular audio decoders are mostly implemented based on DSP solution, RISC CPU solution or ASIC solution. Among them, the decoders implemented based on DSP solution or RISC CPU solution are more flexible and their time to market periods are more shorter than the one implemented based on ASIC solution, but these decoders consume too much power, conflict with the demand of the mobile terminals. Although the decoders implemented based on ASIC solution could slove the power problem, they are not so flexible and their time to market periods is too long compared with the prior solution. The decoder based on hardware and software co-design technology considers of both lower power consumption and higher flexibility. As a result, we research the audio decoder by the hardware software co-design technology. The decoder here includes a CPU core and several hardware modules. The CPU can process the control intensive tasks and the hardware modules process the computational intensive tasks.The main content of the thesis includes the following aspects:First, the popular solutions for audio decoders are introduced and analyzed. And based on the comparison result, the MP3 and AAC audio decoder is designed and implemented using the hardware software co-design technology.Next, the basic principals and steps of the hardware software co-design technology are introduced and described. Then, the decode arithmetic of MP3 and AAC audio are analyzed and the results of arithmetic complexity analysis are presented according to the steps of co-design. Based on the complexity analysis, the decode arithmetic are divided into two parts, namely software part and hardware part.And then, the design of some hardware modules includes VLD and transformation part is presented. For Huffman decoder, a novel Huffman decode method for MP3 and AAC audio is applied, namely 16 search trees decode method. This method could accelerate Huffman decoding so much only at less expense of codebook storage. For the transformation part, a general IMDCT fast algorithm based on DCT-II core is adopted for the MP3 subband filters and the AAC.Then, the implementation of software and hardware cooperation is discussed in detail. After the implementation of hardware, the thesis presents the fix-point implementation of the float-point software, the result of comparison between PCMs between fix-point software and float-point software shows that the precision could meet the demand. Finally, the thesis does the evaluation of the MP3 audio decoder, includes evaluation of hardware resource usage and effect with the hardware modules. Based on the evaluation results, this design can implement the real time decoding of the MP3 mono audio stream at 32kHz sample rate.