Research On Hidden Markov Model Based Distributed Arithmetic Coding

Although wireless sensor network has become a maturing technology, the power of each sensor node is very limited. In order to prolong the life of the whole network, the energy consumption of each node needs to be reduced in the application. In the traditional joint source coding method, the correlated sources need to communicate to each other in encoding process, so the encoder consumes more energy than the decoder. These coding methods are not suitable for applications such as wireless sensor networks. In order to solve the problem, some scholars have proposed a coding method called distributed source coding which is used to encode correlated sources recently. In this new coding method correlated sources perform their own encoding independently without communication. Distributed arithmetic coding which based on the distributed source coding theory uses arithmetic coding as its core process. This paper includes contents as follows:Firstly, it analyses the problem of channel coding based distributed source coding and introduces distributed arithmetic coding. After researching on the principle of distributed arithmetic coding and distributed source coding theory, this paper analyses the theory limit of distributed arithmetic coding according to Slepian-Wolf bound. Then it gets the range of the parameter k's value which is used to amplify source symbol's sub-interval. By the study on Hidden Markov Model (HMM), the encoding source X and correlated side information Y of distributed arithmetic coding can be corresponded to the observer sequence and the hidden sequence of Hidden Markov Model respectively. HMM based distributed arithmetic coding is a new method which combines the Hidden Markov Model and distributed arithmetic coding. Although the coding efficiency of distributed arithmetic coding can be improved by using overlap region, if the codeword lies in the overlap region a big decoder tree is generated to keep the ambiguity in the decoding process. So an improved Viterbi algorithm is introduced to prune the big tree and select the best path of the tree as the final decoded result.Secondly, this paper explores a method to apply HMM based distributed arithmetic coding into image data. According to the spatial redundancy characteristics of neighbor pixel data in an image, the neighbor lines of image are regarded as the observer sequence and the hidden sequence respectively to build the Hidden Markov Model. Then, this paper describes the progress of encoding and decoding image data using HMM based distributed arithmetic coding in detail.Finally, two experiments which use simulation binary correlated sequences based on Hidden Markov Model and actual image pixel data to verify Hidden Markov Model based distributed arithmetic coding are carried out respectively. By several different experiments and analysis on their results, it can be found that compared to the traditional arithmetic coding, coding method proposed in this paper can obtain better coding efficiency. Furthermore, when hidden Markov model is used for data modeling, using the Viterbi algorithm in the decoder can get a higher accuracy rate than using maximum a posteriori estimate in decoder.