Institut Français des Sciences et Technologies des Transports, de l'Aménagement et des Réseaux

Type de document

CHAPITRE D'OUVRAGE (CO)

Langue

anglais

Auteur

HERNANDEZ ; KHARDI ; VALLET ; ESTEVE

Résumé / Abstract

This paper presents a new hybrid diagnostic methodology and the results obtained for hypovigilance detection. This new hybrid system involves statistical pre-processing, Artificial Neural Networks (ANN) and Fuzzy Logic (FL). The statistical pre-processing of data allows outliers to be eliminated (filtering) and to perform an efficient system construction. Then, each driver state is learned by a particular ANN constructive algorithm. The final step is the final decision by Fuzzy Logic algorithms, this is performed by ANNs generalization and a defuzzification method which allows real-time to be performed. This method is applied to several SAVE experiments which study the driver's impairment under real driving conditions. Experimental results indicating the early states of impairment at the wheel have been compared with those derived from the previous method. Using experimental data processing, we investigate the reliability and efficiency of this method which could probably be integrated into a system of monitoring the driver's states.