MFCC coefficient and ANN classifier applied to roadway classification

Currently, Intelligent driver assistance systems provide more satisfactory solutions to the field of road safety. These are based on the use of powerful image or signal processing algorithms in terms of speed, accuracy of calculation and decision making. This document describes an intelligent roadway classification system in real-time in 4 classes: asphalt, gravel, snow and stone road. This system relies on the use and combination of two algorithms; the Mel Frequency Cepstrum Coefficient algorithm that extracts the characteristics of each type of road from the sound due to friction of the vehicle wheels with the road, and Multilayer Perceptron algorithm, which receives its characteristics to its inputs and classifies them. This document also describes the stages of the process of learning and testing. The robustness of this system is shown by the obtained satisfactory experimental results that combine between the respective advantages of the MFCC descriptor and MLP classifier.