Signal representation using adaptive parallel mixed transform techniques

Transform based signal representations are central to lossy compression methods. Since most signals of interest contain both narrow and broad band frequency components, mixed transform techniques can frequently provide more efficient representations of such signals than individual transforms. This paper presents a new adaptive parallel mixed transform technique for signal representation. Formulations are derived, algorithms are described and results of simulations applied to speech signals are presented. Excellent performance is achieved with much less computation than required by existing mixed transform techniques, making real-time implementations feasible