One of the more successful algorithms used in speech pattern matching is dynamic time warping, however, when applied to large vocabulary systems or when used for connected speech recognition, it uses a considerable amount of processing time. The most time consuming part of the algorithm is the calculation of the distance between two frames of parameters. If the two frames are highly dissimilar then the exact distance between them is not very important, and the time spent calculating it is wasted. This paper suggests the use of a quick distance measure which produces a fixed distance for obviously dissimilar frames, while passing the remainder on to a more detailed distance measure. The quick measure described is capable of providing a factor of four improvement in the speed of the dynamic time warping algorithm.