Article Preview
Top1. Introduction
Wireless sensor network (WSN) consists of battery driven sensor nodes that perform the sensing, computation and wireless communication. WSN are resource constraints such as less communication bandwidth, limited energy supply, less storage and less computing power (Akyildiz & Cayirci, 2002). In recent past, there has been a significant improvement in WSN that improve the processor design and computing issues, but limitation of battery provision still exists (Chen & Zhang, 2012).
Wireless multimedia sensor network (WMSN) has been possible due to the production of cheap CMOS (Complementary Metal Oxide Semiconductor) camera and microphone sensors which can acquire multimedia information. WMSN contains sensor nodes equipped with cameras, microphones and other sensors producing multimedia contents. These sensor nodes retrieve much richer information in the form of image or video and hence provide more detailed and interesting data about the environment (Wang & Cao, 2011).
Sensor nodes (SNs) are battery driven devices and deployed in remote and hostile environment, so it is difficult to replace the batteries (Akyildiz & Chowdhury, 2007). Large numbers of sensor nodes are deployed to monitor the physical environment. In case of WMSN, similar data is produced by neighbouring sensor nodes and transmit observed data to the sink. Large amount of energy is consumed while data is transmitted to the sink. To reduce the energy consumption or prolong the network lifetime, we can apply data compression techniques that removes redundant data because more energy is consumed in data transmission as compared to data processing at sensor node (Xiao & Hui, 2012).
Data compression is a technique to reduce transmission bandwidth, storage cost, etc. by eliminates redundant data. There are two types of compression, lossy and lossless. Lossy compression identifies unnecessary information and eliminates that information. This compression often is used for audio and video data. Lossless compression usually exploits statistical redundancy to represent data more concisely without losing information (Suarjaya 2012). No information is lost in lossless compression. Data compression can also be referred as in-network processing technique for energy saving because size of data is reduced (Capo-chichi & Friedt, 2009).
In this paper, we use Principal Component Analysis (PCA) technique for data compression. PCA is a compression method widely used in statistical analysis and image processing (Borgne & Bontempi, 2007). PCA can efficiently remove redundant data from raw (sensor measured data) and also between the principal components (PCs) of neighbouring sensor nodes with a promise of ensuring the data reconstruction accuracy (Chen & Tian, 2013).
The purpose of PCA is to reduce the dimensionality of a data set (sample) by finding a new set of variables. New set of variables are smaller in size than the original set of variables, which nonetheless retains most of the sample's information. Correlations between the original data have been removed and new set of variable are generated which are called principal components (PCs), are uncorrelated, and are ordered by the fraction of the total information each retains.