A Compressibility-Based Clustering Algorithm for Hierarchical Compressive Data Gathering

• Main Authors: Ming-ZhiWei, 魏銘志
• Other Authors: Kun-Chan Lan
• Format: Others
• Language: en_US
• Published: 2015
• Online Access: http://ndltd.ncl.edu.tw/handle/97639196847795796903

碩士 === 國立成功大學 === 資訊工程學系 === 103 === For most wireless sensor networks (WSN), wireless transmission is a major contributor to power consumption. Data gathering is one of the important functions in WSN, but this introduces a lot of wireless transmission overhead. Data compression is the most common approach, to reducing the transmission required for data gathering. However, conventional data compression will cause heavy in-node computation, and thus applying compressive sensing (CS) with wireless data gathering has attracted growing attention. With regard existing CS-based data gathering work, Hierarchical Compressive Data Gathering (HCDG) is the most transmission-efficient architecture. For HCDG, the clustering algorithm used has a close relation with the amount of transmission data that is gathered, and existing HCDG works adopt the Random Clustering method as their clustering algorithm, although this still has a lot of transmission. We thus proposed a Compressibility-Based Clustering Algorithm (CBCA) for HCDG, demonstrate that CBCA has lower transmission than the Random Clustering method, and find the optimal parameters of CBCA through mathematical analysis. We used water level data collected form an inundation monitoring system in our simulation experiment, and the results verify the mathematical analysis, and demonstrate the recovery quality still maintains a percent root difference (PRD) 〈5% at the optimal parameters of CBCA.