This paper introduces a new method for finding all subsequences similar to a given time series sequence. The method takes into account noise .offset translation and amplitude scaling. Based on a piecewise linear representation, the speed is exceptionally fast.