Energy-saving data approximation for data and queries in sensor networks

One way of conserving the scarce resources in a sensor network is to minimize the amount of data transmitted. This can be accomplished by data compression, aggregation or approximation. The current researches on sensor data compression mainly focus on lossless compression methods, they cannot achieve higher compression ratio than lossy data compression. In-network data aggregation and data approximation can be regarded as lossy data reduction methods. However, in-network data aggregation methods cannot record all the features of sensor data, thus queries referring to the historical data might not be answered. Moreover, the data cached in sensor networks should be used easily for answering queries. Based on the correlation and frequency spectrum analysis results of some types of slowly varying sensor data, we have presented two data approximation methods to reduce data transmission while make queries easy to answer. We have implemented these methods, tested on some real life data sets and compared with related methods. The results indicate that the algorithms are simple and deliver high data reduction ratios, while meeting the user's tolerance of errors.