support throughout this project. Dictionary projection pursuit is a fast approximate form of projection pursuit that can be used with a wavelet packet dictionary to efficiently find features in high dimensional waveforms or images for pattern recognition purposes. Basis functions for a linear feature extractor are chosen so that the projection coefficients of the ensemble minimize or maximize a user-defined projection criterion subject to being approximately orthogonal to the subspace defined by the previously selected basis functions. A survey of possible projection criteria is presented. Index Terms: classification, dimensionality reduction, waveform, Fisher’s lin-ear discriminant, wavelet packet