Feature extraction in pattern recognition

Some Chemical Applications of Machine Intelligence

“…Preprocessing of data before classification has received considerable attention (30,31). In specific studies of chemical data, several preprocessing operations have been investigated, including converting spectroscopic peak intensities to their square roots (16) or their logarithms (23), generating cross terms (26), and using Fourier transforms (25).…”

Sonar Sensor Interpretation and Infrared Image Fusion for Mobile Robotics

“…Other research groups are actively developing algorithms for the automatic interpretation of sensor data (Leonard & Durrant-Whyte, 1992;Crowley, 1985;Dror et al, 1995;Jeon & Kim, 2001). Our goal is to bring to bear a high level knowledge of the physics of sonar backscattering and then to apply sophisticated discrimination methods of the type long established in other fields (Rosenfeld & Kak, 1982;Theodoridis & Koutroumbas, 1998;Tou, 1968). In section 2 we describe an algorithm to distinguish big trees, little trees and round metal poles based on the degree of asymmetry in the backscatter as the sonar beam is swept across them.…”

Sequential least-squares prediction based on spectral analysis

“…The representation of a general, nonperiodic random process observable on a finite interval by a converging sum of orthogonal functions with random coefficients has been studied extensively [I-41. It has been used as a signal estimator of a signal-plus-noise process [5], as a Perceptron model in problems of recognition and learning [6, 71, in optimalising and learning control 181, and for kernel approximation in the extraction of statistical features from observable data in pattern recognition [9]. The spectral representation has also been used successfully as a method of prediction for the short-term forecasting of electrical load demand [lo-141.…”