QMUL School of Electronic Engineering and Computer Science
Centre for Digital Music Seminar Series
** Lecture by Dr Mathieu Lagrange, CNRS **
**Date/time: 11-11.30am, Wednesday, 21st November 2018 **
**Location: Engineering Building, room 2.16 ** Campus map: https://www.qmul.ac.uk/media/qmul/docs/about/Mile-End_map-May2018.pdf
Open to academics, students, alumni, public; all welcome. Admission is FREE, no pre-booking required.
Title: Beyond Fourier?
Abstract: The Short-Time Fourier Transform is nowadays at the core of most digital audio processing systems. However, the use of such mono frequency resolution approaches comes at a modelling cost that decades of research in signal processing has tried to mitigate. Today, new types of architectures based on deep learning called "sample based" as they output audio samples and not spectral frames seem to largely overcome those issues. Besides their efficiency, they have interesting architectural features. They are inherently multi-resolution and only assume causality and not stationarity over a fixed time frame. In the way they are constrained, they are also reminiscent of parametric sinusoidal modelling where only the controlling parameters are assumed to evolve slowly with time. In this talk, I will take an academic point of view of those matters and discuss the potential of these approaches for solving important research problems.
Bio: Mathieu Lagrange is a CNRS research scientist at LS2N, a French laboratory dedicated to cybernetics and computer science. He obtained his PhD in computer science at the University of Bordeaux in 2004, and visited several institutions, both in Canada (University of Victoria, McGill University) and in France (Telecom ParisTech, IRCAM). His research focuses on signal processing and machine learning algorithms applied to musical and environmental audio analysis and synthesis.