Location: Shure Incorporated, 5800 W. Touhy Ave, Niles, IL 60714
Speaker(s): John Woodruff, Knowles Electronics
Listeners with normal hearing can recognize the source of a sound, localize the point of origin, and separate the information provided by an individual source from competing sound sources. There is a longstanding interest in developing algorithms to achieve these capabilities in commercial products. Performance for some problems, such as automatic speech recognition (ASR), has improved substantially in recent years. Conventional signal processing techniques, however, are still widely deployed for the problem of sound separation in spite of well-known limitations.
Supervised learning algorithms have been central to the advances achieved in ASR, and such algorithms are poised to displace or augment long-standing signal processing methods used for sound separation. Recent literature has shown that new approaches to sound separation enabled by machine learning may lead to transformative differences in user experience. One example is improving speech intelligibility in a noisy environment for a hearing aid user. Many technical challenges remain to be overcome before we see widespread deployment of these methods.
In this discussion we will cover the basic concepts and acoustic cues involved in conventional approaches to sound separation, such as beamforming and speech enhancement. We will also introduce recent supervised learning approaches to sound separation and discuss where these can be used in combination with, or to replace conventional methods. Finally, we will talk about the challenges in deploying supervised learning methods for sound separation in real-world products.
About the Presenter:
John Woodruff leads audio processing algorithm development for Knowles’ Intelligent Audio division in Mountain View, CA. He has been with Knowles and Audience since 2012, developing algorithms for detection, localization, classification, separation and enhancement of audio signals, and helping to deploy those algorithms in smart phones, laptops and other consumer devices. Prior to joining Audience, he worked on algorithms for sound separation and localization, pitch tracking, and music remixing in the Perception and Neurodynamics Lab at Ohio State University and the Interactive Audio Lab at Northwestern University. John received a Ph.D. in computer science and engineering from Ohio State University, a M.Music in music technology from Northwestern University, and a B.Sc. in mathematics from the University of Michigan.
Other Business: Dinner (optional, but please RSVP) will begin at 6:30pm. Contact Giles Davis ([email protected]) by Wednesday, February 22nd if you would like to join us. Pizza and salad from Lou Malnati’s will be provided. Please let Giles know if you have a preference for vegetarian, gluten-free, etc. Price is $10 for non-members and $8 for members and students (please bring cash).
Posted: Monday, February 13, 2017
« Stereo Mic Technique for Augmented Ambience… | Main | Tour of the Old Town School of Folk Music »
