Digital filtering: application on the driver´s impairment detection

We show the importance of a filtering step for real-time on-board applications such as the driver impairment detection system. Sampling on-board a vehicle produces noisy signals representing errors for the data processing module. Digital filtering is applied to signals from different sources including: lateral position, steering wheel angle and vehicle speed. Our filtering approach takes into account two phases for each signal. The first phase makes a correction of measures out of normal ranges. The second phase reduces the noise in high frequencies of the signal. For each variable, we tested a combination of: Kalman filter, least-means-squares, linear interpolation and elliptic filter. Results obtained show the better combination of filtering for each signal as follows: 1) for lateral position uses the linear interpolation and Kalman filter; 2) for steering wheel angle uses the Kalman filter and least-means-squares; and 3) for vehicle speed uses the elliptic filter