This study propose an algorithm for estimating walking patterns by the use of foot-pressure sensors embedded in shoes. The algorithm can classify each step into seven patterns: standing, level walking, walking downstairs, walking upstairs, walking downhill, walking uphill and running. The classification is based on foot contact duration, maximum foot pressure upon impact, maximum foot pressure upon lifting foot, maximum pressure at the forepart of the foot upon foot impact and maximum foot pressure at the toe upon lifting foot. To examine the algorithm, the foot pressure of eight healthy male subjects was measured while walking on a level floor, up stairs and down stairs. Experimental results showed that the estimated accuracy of walking patterns was over 94%. The estimated accuracy of the number of steps based on the foot pressure was 100%, while that of the pedometer on the market was 85 to 94%.