Comparison between the MFCC and DWT applied to the roadway classification

Currently, in the field of road safety, research is moving towards the use of electronic driving support systems that are capable of simulating human perception. These systems have more and more facilities, flexibilities and human development, to respond effectively against the delicate situations in the real world, which require the development of more efficient, fast, accurate signal processing and decision algorithms. This paper presents a classification of intelligent real-time roadway into 4 classes: asphalt, gravel, snow-covered road, stone road. This system combines between a descriptor, the acoustic signal produced by the tire-road friction, based either on the Mel Frequency Cepstrum Coefficient algorithm or on the Discrete Wavelet Transform algorithm and a classifier of artificial neuron like Multilayer Perception network. This paper also presents a comparison of results obtained in terms of execution time and in terms of the correct classification for the 2 systems: a system consisting of Mel Frequency Cepstrum Coefficient descriptor and an artificial neuron network classifier Multilayer Perception type and another system composed of a Discrete Wavelet Transform descriptor and the same type of classifier.