Title :
Feature extraction and sufficient statistics in detection and classification
Author_Institution :
Sanders Associates Inc., Nashua, NH, USA
Abstract :
The effectiveness of sufficient statistics as features in the detection/classification process is studied. The concept of a sufficient statistic is reviewed and an empirical method of developing an `apparent´ sufficient statistic from training data is offered. Examples of the performance enhancement achieved when using such statistics on real world data in both linear and neural network classifiers are given
Keywords :
feature extraction; linear network analysis; neural nets; signal detection; statistical analysis; empirical method; feature extraction; linear network classifiers; neural network classifiers; performance enhancement; real world data; signal classification; signal detection; sufficient statistic; training data; Computer vision; Equations; Feature extraction; Neural networks; Probability density function; Statistical analysis; Statistical distributions; Statistics; Testing; Training data;
Conference_Titel :
Acoustics, Speech, and Signal Processing, 1996. ICASSP-96. Conference Proceedings., 1996 IEEE International Conference on
Conference_Location :
Atlanta, GA
Print_ISBN :
0-7803-3192-3
DOI :
10.1109/ICASSP.1996.550519
