Real Time Classifier For Industrial Wireless Sensor Network Using Neural Networks with Wavelet Preprocessors

• Main Authors: Akojwar, S.G., Patrikar, R.M.
• Format: Conference Proceeding
• Language: English
• Subjects: Bandwidth; Discrete transforms; Discrete wavelet transforms; Embedded computing; Microsensors; Neural networks; Noise reduction; Subspace constraints; Very large scale integration; Wireless sensor networks
• DOI: 10.1109/ICIT.2006.372319

Wireless sensor node is embedded of computation unit, sensing unit and a radio unit for communication. Amongst three units communication is the largest consumer of energy. Energy is the prime source for wireless sensor node to function. Hence every aspects of sensor node are designed with energy constraints. Neural Networks in particular the combination of ART1 and FuzzyART(FA) can be used very efficiently for developing Real time Classifier. Wireless sensor networks demand for the real time classification of sensor data. In this paper classification technique using ART1 and Fuzzy ART is discussed. ART1 and FA have very good architectural strategy, which makes it simple for VLSI implementation. The VLSI implementation of the proposed classifier can be a part of embedded microsensor. The paper discusses classification technique, which can reduce the energy need for communication and improves communications bandwidth. The proposed sensor clustering architecture can give distributed storage space for the sensor networks. Wavelet Transform is used as preprocessor for denoising the real word data from sensor node, this makes it much suitable for industrial environment. Many methods of wavelet transforms are available. Simplest Haar 1D transform is used for preprocessing and smoothing the sensor signals. The discrete wavelet transform implemented here helps to extract important feature in the sensor data like sudden changes at various scales.