Optimization And Implementation For G.729A Speech Codec Based On CKA5102 Platform

Speech is the most common form of data exchange in the modern communications system,. With the modern voice communications technology continues to development, communication bandwidth efficiency and communication quality has become the two prominent and conflicting problems. So designing a low-rate, high-quality and low-cost voice codecs is necessary.In order to meet these demand, ITU-T proposed in 1995 G.729 codec, the 8kbit/s speech coding algorithm beased on conjugate structure algerbraic code excited linear prediction (CS-ACELP). G.729A retained the decoding algorithm structure of G.729, and did a certain degree of downsizing on the algorithm complexity, but their the coding streaming format is identical. The G.729A algorithm's voice quality is equal to the 32kbit/s ADPCM algorithm under 8kbit/s code rate, while the delay is 15ms. This dissertation studies G.729A speech codec algorithm's optimization and implementation on a particular embedded hardware implementation platform.A digital voice system generally needs "CPU + DSP" dual-core chip or two chips for the whole application solutions. In this dissertation we hope to seek a single-CPU-cored MCU chip with domestic independent intellectual property to finish the speech codec, network/communication protocol stack and system control at the same time, so as to effectively reduce the cost of the whole application solution, meanwhile make a contribution to the application of domestic embedded CPU. Therefore the dissertation chooses the domestic 32-bit embedded CK510E core and the CK510E-based MCU chip CKA5102 as the hardware implementation platform for optimization and implementation of G.729A speech codec algorithm, and also designs the system board hardware for the CKA5102 platform.The dissertation analyzes the organization structure of the G.729A algorithm standard source code, ports the G.729A source code to the CKA5102 hardware implementation platform, and optimizes the G.729A source code on the algorithm level, code level and hardware platform level, finally adjusts the coding data format. We use multi-stage searching acceleratation and fixed code-book searching acceleratation for algorithm-level optimization, unnecessary overflow judgment deleting, function directly insertation and macro definition, assembly rewriting with DSP expanded command, and process optimization of algorithm procedures for hardware platform level optimization.The dissertation designs the algorithm testing scheme, and tests the optimized G.729 speech codec algorithm on the aspects of speech data validation, speech quality evaluation and treatment performance testing. The validation and testing results show that the optimized algorithm can realize G.729 codec in real-time based on CKA5102 hardware implementation platform with a satisfaction speech reconstruction.