Reseach On Joint Source-channel Coding Technology

Interference, fading and noise exists in the wireless channel, and in order to achievea lower error bit rate, a strong error correction coding method is needed. However,system complexity and delay will increase applying the strong error-correcting codingmethod. Under this circumstance, the Joint source channel coding technology is the mainway to solve the contradiction.Two kinds of joint source channel coding systems were set up in this paper, and ajoint source channel model was built. Considering the quantization and channel noise, thejoint channel (quantization, inverse quantization and binary symmetric channel) ismodeled as a Gaussian-Bernoulli-Gaussian (GBG) channel model. The performance ofthe joint source channel coding system which was based on Real BCH coding wasanalyzed firstly. A joint source-channel coding system which is based on over-sampledfilter banks was set up thereafter. Over-sampled filter banks as error correction code leadto a new method in the field of communication. Poly-phase decomposition of OFBS wascarried out to find a new threshold calculation method based on Bayesian hypothesistesting.Matlab simulation result shows that the joint source channel coding based on RealBCH code has a good effect in image transmission under conditions of differenttransition probabilities. The image transmission quality is very good when the channeltransition probability is smaller, and the image quality drops gently when the channelconditions deteriorate. Because of adding redundancy in the process of Real BCH coding,it will weaken the coding efficiency. Redundancies are produced by the signal throughthe filter banks with information of original signal components. The system can restoreout high quality signal at the receiving end when the channel quality is better, and theregain signal quality is decreasing when the channel quality becomes poor. Even if thechannel is bad, the quality of return signal is still acceptable. The new threshold methodbased on over-sampled filter banks is concise and effective.