Research On Key Techniques Of AMR-WB+ Algorithm And Coded Stream Recognition

The multi-rate audio codec, which is adopted by 3G mobile communication system, satisfies the need of high quality coding and transmission of speech and music. The AMR-WB+ algorithm is a research focus of the area. The dissertation investigates in interrelated techniques of AMR-WB+ and stream recognition, containing three aspects below:First of all, the dissertation investigates in the calculation and quantization algorithms of ISF. Firstly, the definition of ISF and the transformation algorithm for the parameter and LPC are introduced, then a lower bit rate algorithm for ISF is proposed by improving on SSVQ introduced by other literature. After switched by switch codebook, input ISF vectors substract from the corresponding vectors in the switch codebook, instead of being quantized directly by split vector quantization. Then, difference vectors are quantized by SVQ. The experimental results show that the new algorithm can achieve transparent quantization at the bit rate of 37 bits/frame, which is less than that of SSVQ and AMR-WB+, respectively 44 bits/frame and 46 bits/frame. The memory requirement for codebooks of the new algorithm is only one third of those of SSVQ.The dissertation also investigates in TCX technique. The dissertation introduces TCX flow, and researches on the mechanism of quantizing frequency spectrum coefficients by lattice vector quantization. A 12.8kbps TCX codec algorithm is realized. The codec's availability for music signals is validated by average segment SNR and informal subject listening test.The dissertation's research contains coded stream features which are used for coded stream types recognition. The research objects in the dissertation are twelve kinds of coded stream including AMR-WB+. Stream features like frame length, probability of bit one and zero are studied. The frame types of AMR-WB+ are distinguished by frame length and frame header. A method called frame length combined with character frame is proposed for recogizing fixed rate parameter coding algorithms. The method in the first place determines frame length of unknown stream by inter-frame similarity, and then makes use of character frame templates and the corresponding match threshold for stream whose frame length is identical. The experimental results show that the methods work well for recognizing coding algorithms.