This repository contains experiments in translation-invariant analysis-synthesis from audio signals. The problem is the following:

How to build a nonlinear representation of audio signals that provides a perceptually good reconstruction, under the constraint of invariance to translation ?

The code was written by Vincent Lostanlen, a PhD student of Stéphane Mallat at the École normale supérieure, and is released as free software under the MIT license.