Homogeneous Multi-Classifiers for Moving Vehicle Classification

Profoundly hearing impaired community cannot moderate wisely an acoustic noise emanated from the moving vehicles in outdoor. They are not able to distinguish either type or distance of a moving vehicle approaching from behind. Hence, a hearing impaired encounters a risky situation while they are in outdoor. In this paper, a simple system has been proposed to identify the type and distance of a moving vehicle using multi-classifier system (MCS). One-third octave filter band approach has been used for extracting the significant feature from the noise emanated by the moving vehicle. The extracted features were associated with the type and distance of the moving vehicle and the homogeneous MCS based on multilayer Perceptron (MLP), K-nearest neighbor (KNN) and support vector machines (SVM) has been developed. Pairwise diversity measure was developed and the diversities between the classifier teams were measured. In order to achieve diverse classifier, sampling based on repeated holdout and bootstrap methods were used and compared. The developed MCS based on repeated holdout method has performed well on ensemble accuracy. Meanwhile, the bootstrap method gives a better performance on diversity measure. For both sampling methods, as a base classifier, the SVM outperforms MLP and KNN base classifiers.