摘要： 无线传感器网络的能量和通信带宽有限，不适合大规模数据传输，需进行压缩处理。为此，研究无线传感器网络中基于回归的数据压缩问题，提出分段线性回归拟合算法和基于置信间隔的回归模型调整算法。分段线性回归拟合算法通过分段，使回归拟合适应环境数据周期性变化的规律。回归模型调整算法能够确定分段时机，使回归直线更加逼近动态变化的环境数据集。在Berkeley-Intel数据集上的实验结果表明，该算法在较小的重构精度下能达到3%的压缩比。

关键词: 无线传感器网络, 回归, 数据压缩, 分段, 置信间隔

Abstract: Wireless Sensor Network(WSN) are not fit for the transmission of large-scale data because of their limited energy and bandwidth, and thus sensory data have to be compressed. This paper studies the problem of regression-based sensor network data compression. A linear regression algorithm based on segment and a regression adjustment algorithm based on confidence interval are proposed. Segment technique makes regression adapt to periodical change of environments whereas regression adjustment decides the time of segment, and thus regression line can approximate the environment data set more accurately. Through experiment of the Berkeley-Intel data set, result shows that the compression ratio of the proposed algorithms comes to 3% in minor reconfiguration precision.

Key words: Wireless Sensor Network(WSN), regression, data compression, segment, confidence interval

中图分类号: 

• TP393