Inaugural lectures
Wednesday 16th November, 2011 at 5pm - Lecture Theatre T.0.03, Wolfson Building
Signal Processing Based Solutions to the Machine Cocktail Party Problem
Professor Jonathon Chambers, Electronic and Electrical Engineering Department
Have you ever considered how you are able to hold a conversation in a noisy environment such as the bar in the students union or at a dinner party, when there is so much background interference? And how useful it would be if this skill could be automated?
Mimicking the ability of a human to follow one conversation within a room when many people are speaking is known as the machine cocktail party problem. As humans we use our listening and visual senses together with sophisticated signal processing in our brains to solve this problem. In a machine, the ears and eyes of a human are simply replaced by microphones and video cameras; but, the challenge is processing the measured audio-visual signals to extract the conversation of interest.
In this lecture, Professor Jonathon Chambers will begin by guiding you through the fascinating field of adaptive signal processing which has been the focus of his research for the last 25 years and led to his elevation to IEEE Fellow on 1st January 2011.
He will then highlight the particular difficulties in blind or unsupervised learning, and introduce the concept of blind source separation as a framework for extracting the conversation of interest from an array of microphone measurements. He will also show the way in which video cues can aid in this process and demonstrate machine audio-visual source separation.
Finally, he will describe the future research directions in this field and provide examples of its wide applicability.