I. Introduction
Recently, a new method for constructing nonuniform modulated lapped transforms (MLTs) was introduced [1], [2] by combining subband filters of a uniform MLT. We shall refer to this method as subband merging. The combined subband filters have better time localization than the noncombined filters at the expense of a decrease in frequency localization. Since the nonuniform filter banks are obtained by simply taking linear combinations of the filters of a uniform MLT, the proposed method allows for an efficient implementation of time-varying transforms without the need for transition filters. It was shown that subband merging can be used beneficially for reducing ringing artifacts, e.g., reverberation and pre-echo, in audio and speech coding [2]. The design of such transforms, however, was restricted to combining two or four subband filters only, and no systematic design procedure was given.