Abstract: In order to protect drivers' safety under emergency, utilizing the driver's behavior & reaction research platform and cluster technology, a structure model between physiological reactions and the braking parameters, which include braking reaction time and maximum braking pedal power, was established. The traits of the driver's physiological reactions were discussed; the relationship was studied and analyzed. Then using the fuzzy adaptive network, the braking behavior was predicted based on the physiological data. The results show the fact that the physiological indices of the driver change in different emergency situations are obvious. The more driving experience the driver has, the little the physiological indices change. The prediction errors of the FAN were under 2%.