Institutional Repository
Technical University of Crete
Technical University of Crete
EN
|
EL
| URI: | http://purl.tuc.gr/dl/dias/E66E0EC8-8717-4D85-B798-F132BFFAA7A6 | ||
| Year | 2007 | ||
| Type of Item | Peer-Reviewed Journal Publication | ||
| License |
|
||
| Bibliographic Citation | A. Deligiannakis, Y. Kotidis and N. Roussopoulos, "Dissemination of compressed historical information in sensor networks," VLDB J., vol. 16, no. 4, pp. 439-461, Oct. 2007. doi:10.1007/s00778-005-0173-5 https://doi.org/10.1007/s00778-005-0173-5 | ||
| Appears in Collections |
Sensor nodes are small devices that “measure” their environment and communicate feeds of low-level data values to a base station for further processing and archiving. Dissemination of these multi-valued feeds is challenging because of the limited resources (processing, bandwidth, energy) available in the nodes of the network. In this paper, we first describe the SBR algorithm for compressing multi-valued feeds containing historical data from each sensor. The key to our technique is the base signal, a series of values extracted from the real measurements that is used to provide piece-wise approximation of the measurements. While our basic technique exploits correlations among measurements taken on a single node, we further show how it can be adapted to exploit correlations among multiple nodes in a localized setting. Sensor nodes may form clusters and, within a cluster, a group leader identifies and coalesces similar measurements taken by different nodes. This localized mode of operation further improves the accuracy of the approximation, typically by a factor from 5 to 15. We provide detailed experiments of our algorithms and make direct comparisons against standard approximation techniques like Wavelets, Histograms and the Discrete Cosine Transform, on a variety of error metrics and for real data sets from different domains.
Copyright © DIAS 2013
