|
|
|
“Multi-Category Classification of Ground Vehicles Based on the Acoustic Data of Multiple-Terrains Using Fuzzy Logic Rule-Based Classifiers”
by Hongwei Wu and Jerry M. Mendel
September 2004
This report summarizes our studies conducted from July 2003 to July 2004 for the multi-category classification of ground vehicles based on the acoustic data of multiple-terrains. Data pre-processing (including elimination of redundant records, processing of data dis- tortion, and generation of prototypes), feature extraction, and uncertainty analysis were performed before classifiers were designed. We established the Bayesian classifier, and type-1 and interval type-2 fuzzy logic rule- based classifiers (FLRBC). These classifiers have similar architectures, consist of four sub- systems each for one terrain, and have one probability model (Bayesian classifier) or one fuzzy logic rule (type-1 and interval type-2 FLRBCs) for each kind of vehicle on each terrain. They differ in the way that this common architecture is implemented. Experiments were conducted to evaluate the performance of all classifiers. Experimental results revealed that 1) for the non-adaptive working mode, both the type-1 and interval type-2 FLRBCs have better performance (smaller mean and standard deviation of classifi- cation error rates) than the Bayesian classifier, and the interval type-2 FLRBC has better performance than the type-1 FLRBC; 2) each classifier has a smaller average but a slightly larger standard deviation of classification error rates in the adaptive working mode than in the non-adaptive mode; and 3) for the adaptive working mode, both the type-1 and interval type-2 FLRBCs have better performance than the Bayesian classifier, and the interval type-2 FLRBC has better performance than the type-1 FLRBC. We applied all classifiers obtained from the above experiment to blind data records (pro- vided to us by the Army Research Laboratory), and used spatio-temporal decision fusion techniques to obtain overall decisions from local decisions. For each blind data record, by varying the number of classifier designs and the number of data blocks, we obtained different overall decisions. With this report we have completed our study into the classification of ground vehicles based on their acoustic emissions by using fuzzy logic rule based classifiers. Our overall conclusion from this study is that FLRBCs always outperform a Bayesian classifier and look quite promising for real-time applications.