Yesterday, I gave an invited tutorial on time-frequency processing with sparse approximation at the Multivariate Analysis for the Technical and Life Sciences (MATLS) day — which is part of the larger 2011 Vision Days conference at the Technical University of Denmark. The audience was extremely diverse, with many coming from statistics and chemometrics, to the food sciences. More and more, I am delving into the world of chemometrics as it is a wonderful blend of signal processing, statistics and estimation, and pattern recognition. And uniting such techniques with sparse approximation methods — not just as feature extraction, but also for classification and dictionary learning — and acoustics, feels natural.
At our department today, one of the organizers of MATLS, Dr. Stina Frosch, is giving a seminar about her work on measuring food quality with acoustic measurements. Naturally, we are working together to tease out the relevant data from acoustic measurements of falling frozen prawns for estimating their ice glaze content. However, what we need are some “shrimp-invariant features,” e.g., features that are invariant under changes in shrimp size. This work will no doubt lead to fantastic possibilities for paper titles. And if I can unite this with my work in “dark energy,” my moniker could become “the dark energy shrimp.”