Hackerman Hall B17 @ 3400 N. Charles Street, Baltimore, MD 21218
New Twists for New Tricks, Making Audio Deep Learning Practical
Deep Learning has undeniably changed the way we thinkabout audio processing during the last five years. But in the process, it has ushered a number of new practical problems that hinder wide use and adoption. In this talk I will present a collection of projects that are geared towards resolving such issues. I will cover neural network reformulations that allow us to implement audio systems that are a couple orders of magnitude more efficient in hardware; I will talk about systems that learn audio enhancement tasks without being trained on before/after examples;and I will show how we can train models to perform complex tasks such as source separation in a fully unsupervised manner without even presenting clean audio examples to them. These projects have been motivated by real-world problems where both computational power and the ability to gather goodtraining data is limited. We show that even under these limitations we can approximately match the performance of systems that require significantly more resources and effort.
Paris Smaragdis is an associate professor at the Computer Science and the Electrical and Computer Engineering departments of the University of Illinois at Urbana-Champaign. He completed his masters, PhD, and postdoctoral studies at MIT, performing research on computational audition. In 2006 he was selected by MIT’s Technology Review as one of the year’s top young technology innovators for his work on machine listening, in 2015 he was elevated to an IEEE Fellow for contributions in audio source separation and audio processing, and during 2016-2017he is an IEEE Signal Processing Society Distinguished Lecturer. He has authored more than 150 papers on various aspects of audio signal processing, holds more than 50 patents worldwide, and his research has been productized by multiple companies. He has previously been the chair of the LVA/ICA community, and the chair of the IEEE Machine Learning for Signal Processing Technical Committee. He is currently the chair of the IEEE Audio and Acoustics Signal Processing Technical Committee, a senior area editor of IEEE Transactions of Signal Processing, and a member of IEEE Signal Processing Society’s Board of Governors.