Focusing on the signal timing estimation for intersections through the GPS data from buses, an improved interpolation method is proposed to estimate the travel time for individual vehicles at intersection. A signal cycle is divided through clustering followed by classification. By sing GPS data sequence near the stop line, a method for estimating cycle boundary and cycle length is given. Combined with the results of cycle boundary estimation, a new method for estimating red time is proposed. The simulation and field tests for the estimation method are carried out under different traffic flows. Results show that the signal timing estimation method using bus GPS data is applicable for different traffic flows, and the estimation effect is obviously better than existing methods.