Tree-Structured Linear Approximation for Data Compression over WSNs

In wireless sensor networks (WSNs), how to reduce the power consumption thus lengthen the system life time is one of the key issues to sustain the services. According to the radio model, packet transmission depletes a much more substantial amount of the energy budget when compared to sensing and pro...

Full description

Saved in:
Bibliographic Details
Main Authors: Wang, Chu-Ming, Yen, Chia-Cheng, Yang, Wen-Yen, Wang, Jia-Shung
Format: Conference Proceeding
Language:English
Subjects:
Adaptation models
Approximation
Bit rate
compression
convex hull
Data models
Detection
Distortion
linear approximation
Mathematical analysis
Mathematical models
model-based scheme
outliers removal
Rate-distortion
Remote sensors
Resource management
Sensors
Wireless networks
Wireless sensor networks
WSNs
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:In wireless sensor networks (WSNs), how to reduce the power consumption thus lengthen the system life time is one of the key issues to sustain the services. According to the radio model, packet transmission depletes a much more substantial amount of the energy budget when compared to sensing and processing. Therefore, it is desirable to compress or filter the sensing data effectively in order to save the transmission power eventually. Recently, the model-based scheme is proved to be a promising solution, which usually approximate temporal data by a piecewise linear function. In this paper, a tree-structured linear approximation scheme is proposed to compress sensing data according to an optimal rate-distortion (R-D) relationship. The main design goals are two: (1) providing a bottom-up procedure to explore the best-fit piecewise partition for modeling globally, (2) considering the heterogeneity of sensors simultaneously using our proposed rate-distortion adjustment. That is, a distortion allocation procedure is designed to allocate the distortions to sensor nodes for aware of the heterogeneous properties. Thus the proposed spatio-temporal scheme is adaptable to heterogeneous sensors, various sampling rate, and outliers of data. A real-world dataset simulation is applied to demonstrate the effectiveness. For nearly all combinations with distortion requirements, the proposed method shows better performance than the earlier approaches in terms of data reduction.
ISSN:2325-2944
DOI:10.1109/DCOSS.2016.37