Optimization And Implementation For G.722.1 Wideband Speech Codec

In today's speech communication system, most speech coding schemes are narrowband transmission encoding, whose bandwidth are usually less than 4KHz, such as the sampling frequency of the telephone voice is 8KHz, with 300 Hz to 3400 Hz bandwidth. With the rapid development of digital communication networks and people's demand to high quality of audio-visual services, wideband speech is receiving more favor. Compared with traditional narrow band speech,the wideband (50-7000 Hz), with a sampling frequency of 16KHz, is not directly against the interoperation of traditional PSTN,and provides a substantial quality improvement in terms of naturalness and intelligibility in communication, and thus giving a experience of face to face. Although a higher bitrate may be required, with people's requirement of high quality multimedia, wideband speech coding has extendend its application range widely, such as audio and video conferencing, VoIP(Voice over Internet Protocol) and digital audio broadcasting (DAB) and other fields.There are two kinds of key processing platforms suitable for digital audio processing on the currently market, which are respectively reduced instruction set computer (RISC) and digital signal processor (DSP). It's represented by the ARM and the MIPS that the embedded RISC is versatile, with an outstanding performance in peripheral control and transaction management, and DSP has a powerful data processing ability with high speed, which has a strong specificity. A complete speech codec solution includes the three parts: speech codec, the application part of network/communication protocol stack and system control. It's generally realized by using a multi-core architecture SoC combination of "RISC + DSP" or dual-chip to the implementation of the whole system solution.But it costs relatively expensive by using RISC to handle network/communication protocol stack and system control and using DSP to complete speech codec.So this thesis designed an embedded hardware development platform based on a relatively high performance and low-cost domestic embedded CPU CK-Core, by optimizing G.722.1 wideband speech codec algorithm and then realizing it real-time on the platform, which provided a key technology for getting inexpensive wideband speech codec system solution that finishing G.722.1 speech codec, network/communication protocol stack and system control based on an CK-Core single chip.This thesis is mainly divided into three parts: analysing and verifying G.722.1 wideband speech decoding algorithm; the choice of embedded system hardware platform for G.722.1 wideband speech codec algorithm; the realization and testing results of optimized G.722.1 wideband speech codec algorithm on the embedded platform.This paper introduced some classifications and performance index comparison of the current wideband speech coding technologies, mainly including a summarization of several ITU-T wideband speech coding standards; also recommended the related selection principles of embedded processor through combining the embedded system and digital audio technology.This paper analysed the basic principle of the ITU-T G.722.1 standard,which represents a provision of 50 Hz to 7 KHz audio bandwidth, according to 24 kbit/s or 32 kbit/s bitrate digital wideband working principle of the encoder algorithm; and made a functional verification to G.722.1 algorithm.The paper choosed the 32-bit embedded CPU CK-Core system chip CK5A6 to act as embedded system hardware platform for optimizing and realizing G.722.1 algorithm, and introduced the CK-Core hardware platform and embedded software development environment, and determined the embedded software development process about G.722.1 algorithm based on the CK-Core platform.The paper completed the G.722.1 algorithm transplantation on CK-Core platform, which established the algorithm optimizing targets on the platform, and through the theoretical and practical runtime analysis about the key processes of the G.722.1 algorithm, focused on the optimizing for G.722.1 in these three aspects: overlapping modulation transform/inverse overlapping modulation transform (MLT/IMLT), huffman encoding/decoding, basic arithmetic operations; at the same time, aimed at the characteristics of CK-Core platform and made it another major start point for optimizing. This paper around these optimization goals and optimization strategies, designed and implemented several specific optimization methods, and completed the majorization and realization of the G.722.1 algorithm based on the CK-Core platform.The test results indicate that the optimized G.722.1 algorithm gets its time complexity effectively reduced, which can rebuilt the speech realtime on the hardware platform with a high quality.