The Department of Environmental Sciences at Renault is designing a system for detecting lapses of driver alertness. This system, mounted on board the vehicle, will be designed to warn the driver of any deterioration in his state of alertness. The principle is based on the analysis, in real time, of movements lent by the driver to his steering wheel. The design of such a system requires a study phase in order to know at any time the alertness level of the subject on the basis of physiological signals, so as to be able to determine the parameters derived from the steering-wheel angle signal which, by themselves, will be most suitable for distinguishing between two states of alertness. We describe here a method for defining a physiological reference for the driver’s alertness level.

Fatigue is responsible for 26% of the fatal crashes occurring on the motorway in France. The loss of control of the vehicle can be the consequence of drowsiness not perceived by the driver sufficiently early, in spite of the existence of numerous signs which are the precursors of such a state. The Department of Environmental Sciences at Renault is at present designing an on-board system for detecting lapses of driver alertness, the purpose of which is to notify the driver as early as possible of any deterioration in his state of alertness. This driving aid system should not require of the driver any additional task specific to alertness monitoring, but must be inherent in the task of driving. The principle is based on the real-time analysis of movements lent by the driver to his steering wheel. A microprocessor constantly compares the values adopted by certain parameters derived from the steering-wheel angle signal, with the values of a socalled “high alertness” reference, recorded for example at the start of the circuit. The selected parameters must be characteristic of the differences existing between a high alertness signal and a low alertness signal. They must, in particular, highlight any changes in the precision of directional corrections. This system is described in more detail in some of our earlier publications [1].

Here, we shall deal rather with the study phase aimed at obtaining a reliable physiological reference for the driver’s alertness level. This level must be known objectively to be able to determine the parameters derived from the steering-wheel angle signal, following tions of alertness as closely as possible. We shall therefore describe a method for analysis of physiological signals indicating the alertness level, from recordings performed on driving simulator.