| Efficient utilization of energy has become a bottleneck to restrict the application of wireless sensor networks. It shows that most of the energy is consumed by the wireless communication module in sensor node for data transmission. The research of compression algorithms has become a new research focus, as the compression methods can reduce the number of bits to be transmitted by communication module to significantly reduce the energy requirement and increase the lifetime of the sensor node.They are a lot of lossy compression algorithms have been proposed for wireless sensor networks in recent year. But only a few papers have discussed the lossless compression algorithms in sensor nodes. Since most of the sensor nodes lossless algorithms are tailored from classic algorithms designed for personal computers, the compression performance of these algorithms cannot meet the requirement of practical application in WSN.This paper argues that the research of lossless compression algorithm is more significant in practical. The reasons can be list as follow. The lossless data is a necessary requirement for many advanced applications. The lossless transmission from sensor nodes is a prerequisite for many data fusion algorithms in cluster nodes. Most of lossy and/or distributed compression algorithms can be tailored from lossless algorithms to achieve higher compression ratio and better performance. Therefore, the research of lossless compression algorithm is carried out by this paper.Through analyzing the influence of sensor data’s probability distribution after discrete cosine transform (DCT) and/or discrete wavelet transform (DWT) and/or linear fitting, this paper considered that to encode result data after an entropy-decrease-transform is the only way to further improve the performance of lossless compression ratio. Then, several new entropy-decrease-transform algorithms and an entropy-encoding algorithm are proposed to prove this point. Hence, this paper provided a new method for design and realization of lossless compression in sensor nodes. The research works and innovations can be described as follow. Firstly, a new lossless compression model named Entropy-Decrease-Transform Compression Model is proposed in this paper. Since the research of source coding algorithms is close to perfect, to encode data after entropy-decrease-transform, which can reduce the entropy of sensor data through a reversible transform, has become the only way to further improve the performance of lossless compression ratio. The Entropy-Decrease-Transform Compression Model states that the forms of entropy-decrease-transform are unlimited and the best lossless compression ratio can be achieved only when the corresponding entropy-decrease-transform follows the sensor data’s probability distribution.Secondly, several new algorithms based on fitting-residuals-transform, which is a kind of entropy-decrease-transform, are proposed to achieve lossless compression. The research of fitting-residuals-transform started from a lossless algorithm based on one-dimensional linear regression model (i.e. FR algorithm). The algorithm calculates the fitting-residuals between sensor data and the fitting data based on the linear regression model. Then the fitting-residuals are input to an entropy encoder to achieve compression. By analyzing the shortcomings of FR algorithm, a new algorithm based on difference fitting residuals (DFR algorithm) is proposed to achieve a better compression results. As the inverse transform of DFR must be reversed, which is not suitable for real-time application, this paper proposed two algorithms (P-DFR and P-DFR2 algorithm) based on the difference fitting prediction residuals. Both algorithms can reduce the coding delay while maintaining high compression ratios.Thirdly, several new algorithms based on reversible-domain-transform, which is a kind of entropy-decrease-transform too, are proposed to achieve lossless compression. The research of the reversible-domain-transform started from the DCT transform in H.264. This paper proposed a simple DCT transform from H.264 (H264-DCT algorithm) by replace the complex quantitative, quantization and entropy encoder module with a combination quantization module and a simple Huffman encoder. As the H264-DCT algorithm remains the combination quantization module, it can only achieve quasi-reversible transform. Therefore, this paper proposed a reversible integer DCT-algorithm (S-DCT algorithm) based on the lifting scheme. S-DCT compression algorithm has achieved higher performance on compression ratio and computational complexity. Then, by analyzing the applications of discrete wavelet transform in WSN, this paper proposed an improved algorithm of lifting scheme 5/3 wavelet (DWT53 algorithm) to achieve lossless compression in sensor node. The compression ratio of DWT53 is better than S-DCT’s. After that, this paper has analyzed the compression mechanism of S-DCT and DWT53 and found that the core of reversible transform is the lifting scheme. Then, this paper has analyzed the characteristics of 2-point lifting scheme and its influence for sensor data’s probability distribution. Finally, a new entropy-decrease-transform named difference-median-difference transform (DMD transform) is proposed in this paper. The DMD can get better compression ratio for non-monotonic data than others.Finally, an entropy encoding algorithms is proposed to achieve better compression ratio. After proposed several entropy-decrease-transform algorithms, the paper found that the third-party entropy encoder used by these algorithms is difficult to play down the advantages of transforms. By analyzing the result data’s probability distribution of entropy-decrease-transforms, this paper proposed a new entropy encoder based on normal distribution (ND-Encoding algorithm). And then, after continuous improvement of the algorithm, a context adaptive normal distribution entropy encoding algorithm (CAND algorithm) is proposed in this paper. The algorithm can achieve better compression ratio for both slowly-varying data and non-slowly-varying data.Through the code analyzing and comprehensive testing, this paper has shown that all proposed algorithms can smoothly run in sensor node to achieve good compression ratio, despite a less computational effort. Hence, some conclusions can be made as follows. Through reversible entropy- decrease-transform to change the probability distribution of sensor data is the only way to improve the performance of lossless compression. The form of entropy-decrease-transform is not restricted to DCT and DWT. There are many reversible entropy-decrease-transforms can be used to achieve lossless compression, while most of them ware not be found by us yet. If any reversible algorithm can change the probability distribution of sensor data and the change can be gain to achieve effective entropy encoding, it would achieve efficient lossless data compression!... |
PDF Full Text Request