Henrique S. Malvar

Transform/subband coding of speech with the lapped orthogonal transform

Efficient transform/subband signal coders can be designed using the lapped orthogonal transform (LOT). With the LOT, blocking effects in the reconstructed signal are absent, as demonstrated previously in image coding contexts. What limited the application of the LOT to a broader class of signals, including speech, was the lack of a fast LOT algorithm for large block sizes. The authors define a new version of the LOT that is fast-computable for any block size. A speech processing example demonstrates the efficiency of the LOT in reducing blocking effects.<<ETX>>

Tidbits on tunable analog filters and image demosaicing

This is a somewhat unusual paper. It combines some history of the authors personal experiences with Prof. Sanjit Mitra with a couple of examples of the authors work, in which the author feels that Prof. Mitras work and approaches towards research and impact have had a significant influence.