In this paper we investigate methods and applications of pattern recognition. It focuses the problem of recognizing driving environment of a vehicle by using information obtained from some sensors of the vehicle, and we present an environmental recognition algorithm in which the recognition is dealt with fuzzy logic.
Previously, we presented a recognition algorithm based on fuzzy reasoning. The algorithm can not be applied to meet the demands of nonstandard drivers and changes of vehicle properties, because the membership functions are fixed once they are constructed. To cover such weakness we present an adaptive recognition algorithm with adaptive change of membership function.
A result of computer simulation of the recognition model by using actual driving samples taken from a number of drivers under five driving environments is presented. Seven feature extraction variables generated from thirteen sensor measurements are considered to be the recognition feature. The recognition algorithm is constituted with self-adaptive as well as non-supervised method as there is no extra source of knowledge for correcting the decision taken by the classifier. Here the supervisor is made by using α-cut of membership function of the representative class. We show the efficiency of the self-supervised learning system for recognition of driving environment.