A Pattern Recognition Framework for Embedded Sensor Electronics
The recent developments in the area of high speed micro-electronics and computational intelligence has opened new opportunities in smart sensor design. In this paper a generic pattern recognition framework is presented for integrated sensor based system design. Two case studies are described for Rock-Image Classification and Pulse Shape Identification. Both applications use same framework that consist of pre-processing of sensor data, wavelet based data compression, feature extraction and neural net based feature classification. The rock identification combines multi-parameter analysis to improve the accuracy. The proposed system is tested using above two case studies for real time application. The average accuracy observed for pulse shape and rock type identification is 96% and 95% respectively.