Sparse audio representations using the MCLT

We consider sparse representations of audio based around the modulated complex lapped transform (MCLT) and a generalized iteratively reweighted least squares algorithm which can be interpreted as a variation of expectation maximization. We compare this mildly overcomplete representation to the more traditional modified discrete cosine transform (MDCT) in terms of coding cost and explore the possibility of extending it to a dual-resolution analysis using a pair of MCLT transforms, illustrating its potential application for audio modification.