Energy-Saving DataApproximation forData andQueriesin Sensor Networs
Loket M. V Ramakrishna · 2006
Onewayofconserving thescarce resources ina sensor network istominimize theamountofdatatransmitted. Thiscanbeaccomplished bydatacompression, aggregation or approximation. Thecurrent researches onsensor datacompres- sionmainly focus onlossless compression methods, theycannot achieve higher compression ratio thanlossy datacompression. In-network dataaggregation anddataapproximation canbe regarded aslossy datareduction methods. However, in-network dataaggregation methodscannotrecord allthefeatures of sensor data, thusqueries referring tothehistorical datamight notbeanswered. Moreover, thedatacached insensor networks shouldbeusedeasily foranswering queries. Basedon the correlation andfrequency spectrum analysis results ofsome types ofslowly varying sensor data, wehavepresented twodata approximation methods toreduce datatransmission while make queries easytoanswer. We haveimplemented thesemethods, tested onsomereallife datasetsandcompared withrelated methods. Theresults indicate thatthealgorithms aresimple anddeliver highdatareduction ratios, while meeting theuser's tolerance oferrors.