The Key Technology Research Of Energy Efficient Video Compression In Wireless Sensor Networks

As a interdiscipinary research field of new and traditional branch of science, wireless sensor networks (WSNs) are applied widely and valuable in practical, and attracted more and more attention of experts all of world. In China, the research of WSNs was promoted to a national strategy since a research report,'Research on pilot field direction of knowledge innovation engineering', field of information and automatic, was published in 1999. In this report, WSNs was one of key areas which shall be broken through in the field of information and automatic.Energy saving was the key problem of WSNs. The data access by the traditional WSNs nodes are simple, the energy consumed by data processing in the model of node energy consumed was negligible always. While the energy consumed by data transmission was most part of the total node energy consumed. So the research of WSNs was focused on the problem involved the data transmission, include the flow control, congestion control, routing and energy control, to solve the performance optimization problem under the constraints of energy in data transmission.While the video sensor was added into WSNs, the data accessed by video sensor is extremely huge. This requires the video to be efficiently compressed before transmission, otherwise, the required network bandwidth and power consumption for wireless transmission are tremendous. However, video compression always needs sophisticated computation, the most part of WSNs node energy was consumed by video compression, which far exceed the energy consumed by video transmission. To decrease the energy consumed by WSNs node, some methods must be implement to decrease the compression power and reduce the encode complexity. The video encode standard are faced to application of broadcasting, the video compression are complicated, is not fit for energy control of WSNs. One of methods is to transfer the complexity of encoder to decoder of WSNs, reduce the compression power of WSNs node. The distributed source coding had proved that this transfer was feasible in theory.Based on the theory of distributed source coding, the research was focused on reduce the video compression power while guaranteed the performance of rate distortion. Based on the review of WSNs research of all of world and the distributed video coding in detailed, analyze the main challenges of this research. The study was focused on this problem, the main contents and achievements include:1) Based on the rateless idea, a kinds of low power encode scheme of distributed video coding was presented, corresponding brief propagation decoding algorithm was proposed, the complexity of encode computation was tested in detailed, and then the power of encoder was estimated. Firstly, the limitations of fixed-rate LDPC was discussed, A novel rate adaptive scheme based on rateless LDPC was proposed, rate estimation scheme include encoder rate control (ERC) and hybrid rate control (HRC) were also designed, the corresponding belief propagation (BP) decoding algorithm was presented to enhance its decoding performance. The results of experiments show that the performance of PSNR of proposed methods was increased 1-2dB than the traditional intra encode standard. The computation power reduces about 5-40% than the traditional intra encode. The performance of PSNR of proposed methods was worse than traditional inter encode about 2 dB, but the computation power of inter encode is about 50-60 times of the proposed methods.2) To improve the robust of the video transmission of WSNs node, the digital fountain techniques were used to distributed video coding. In a brief analysis of the basic characteristics of fountain code techniques, then the Raptor codes was discussed which has relatively good performance. The memory-less channel model was used in distributed video coding currently, which was not accorded with the practical cases. So the statistical correlation between the source-side information was modeled as a hidden Markov model, and the joint source-channel design was used. The Raptor was used for distributed video coding, and then the corresponding belief propagation decoding algorithm was designed. The experiments result shows that, on the same packed loss rate, the PSNR performance of proposed system was better than the LDPCA and H.26L FGS 1 dB, and when the packed loss rate increase, the superiority is evident. The power of proposed encoding system reduce about 30% compared with intra encode, while the power of inter encode is 50-60 times than the proposed distributed video coding.3) Based on the interactive encode and decode, a kinds of energy aware distributed video coding scheme was proposed. Using interactive encoding and decoding schemes, establishing universal coding and memory finite state channel model. Then a kinds of linear interactive encode and decode algorithm based on LDPC was proposed, and a new BP decoding algorithm is presented for LDPC decoding also, which applies to the case where the statistical correlation between source-side information can be modeled as a finite state channel. The results of experiments show that the PSNR performance of proposed methods was increased 1-3 dB, the power of proposed encoding system reduce about 20% compared with intra encode, while the power of inter encode is 50 times than the proposed distributed video coding.4) A practical wireless communication system was constructed. The encoder of DVC, JPEG and H.264 was implemented in the energy constrainted PDA, the encoded data was sent by WLAN, the decoding was completed in the server. Four different type standard sequences were used for testing the complexity and the PSNR perfomance of DVC, JPEG and H.264. The results of testing show that the encoding complexity of DVC was lower than the JPEG and H.264, especialy, the encoding complextiy of H.264 was 10 times of DVC. So the DVC is more appreciated for energy limited application. Project was supported by "211"Project of Guangdong Province (Guangdong Development and Reform Commission, No.431), Natural Science Foundation of China (No.60871025), Natural Science Foundation of Guangdong Province, China (No.8151009001000060).