This paper proposes a neuro-fuzzy architecture that can be used in vehicles for the prevention of road accidents. The reaction time of a driver who is in an accident situation is predicted thanks to a network of neurons that admits the physiological and psychological parameters of the latter. To this neural network is associated a unit using the fuzzy logic which provides a modulated warning signal, for a prompt reaction of the driver. The results obtained at the output of the neural module show a match between the test and validation values whose best response is obtained with a correlation coefficient of around 0.98. The Matlab software was used to model our architecture and simulate certain scenarios. As a result, we obtained a ten-neuron network at the input layer and a neuron at the output layer. At the exit of the blur module, we observe the variation of the alert rate according to the anxiety, the inter-vehicle distance and the reaction time. The results show that, depending on the age, sex, accident history, driving experience and anxiety trait, the system calculates the reaction time and then proposes an appropriate warning signal. depending on the type of situation.