Neuro-fuzzy based techniques for acoustic signal classification
Vassilis Kodogiannis · 2001
Abstract: Neural networks and fuzzy systems are currently finding practical applications, ranging from ‘soft’ regulatory control in consumer products to accurate modelling of non-linear systems. This paper presents the design of a classification system for vehicle acoustic signal classification. Vehicle acoustic signals have long been considered as unwanted traffic noise. Acoustic signals generated by each vehicle have been used to detect its presence and classify its type. Training and testing data used in this paper were collected from roadside sensor station at the Valle d’Aosta highway in north-western Italy. Two main systems, multilayer perceptron networks and adaptive fuzzy logic systems, are considered, analysed, and used as classifiers. The results indicate that the fuzzy classifier based on a novel proposed defuzzification method, namely balance of area (BOA), provide more accurate classifications compared to the conventional classifier systems.