A new filter bank theory for time-frequency representation

This paper presents a new formalism for the description of analysis/synthesis systems based on filter banks. Although this work was done in the context of speech coding systems, the results apply to a far broader class of problems. This new formalism has three specific advantages. First, it allows for the description of many different techniques and approaches with one unifying theory. Second, it enables the decoupling of the effects of the various distorting elements in the system design process, allowing each to be addressed separately. Finally, it allows for the relatively simple invention of entirely new techniques in such a way that the relationship of the new approaches to previously developed techniques is always clear.