Hearing Seminars
CCRMA hosts a weekly Hearing seminar. All areas related to perception are discussed, but the group emphasizes topics that will help us understand how the auditory system works. Speakers are drawn from the group and visitors to the Stanford area. Most attendees are graduate students, faculty, or local researchers interested in psychology, music, engineering, neurophysiology, and linguistics. Meetings are usually from 10:30AM to 12:20 (or so, depending on questions) on Friday mornings in the CCRMA Seminar Room.
The current schedule is announced via a mailing list. To be added to the mailing list, send email to hearing-seminar-request@ccrma.stanford.edu. If you have any questions, please contact Malcolm Slaney at hearing-seminar-admin@ccrma.stanford.edu.
Recent Hearing Seminars
Brian Malone (UCSF) on Separating Signal From Noise in Auditory Cortex
Date:Fri, 10/19/2018 - 10:30am - 12:00pmLocation:Seminar Room at the KnollEvent Type:Hearing SeminarBrian Malone, a neuroscientist from UCSF, will be leading the discussion Friday on the role of noise in *improving* auditory processing. There is still a lot we don’t know about neural networks, both biological and artificial. Usually we think of noise as being detrimental. But in several cases noise improves performance. In quantized systems, like spiking neural networks, a bit of noise allows the signal to exceed a threshold and generate a signal, perhaps synchronously, so that we can more easily perceive low-level sounds. This is known as stochastic resonance. In machine learning, a random forest of trees combines several different classifiers, each randomly generated, to improve recognition results.
FREEOpen to the PublicMalcolm Slaney (Google) on decoding pitch, rhythmic, imagined, and surprise signals from the brain
Date:Fri, 10/05/2018 - 10:30am - 12:00pmLocation:Seminar Room at the KnollEvent Type:Hearing Seminar
I want to discuss the experiments and results from the Auditory Group at the Telluride Neuromorphic Engineering Cognition Workshop. For many years we have been doing experiments to decode the state of the brain from EEG signals. We’ve been wildly successful at decoding auditory attention, but discovering signals related to expectations, pitch and imagined sounds is harder. This year we succeeded.
Who: Malcolm Slaney (Google)
What Decoding pitch, rhythmic, imagined, and surprise signals from the brain
When: Fri, 10/05/2018 - 10:30am - 12:00pm
Where: Seminar Room at the Knoll
Why: We want to know how the brain processes more ecological signalsFREEOpen to the PublicNoah Fram on Detecting Musical Surprisal
Date:Fri, 06/01/2018 - 10:30am - 12:00pmLocation:CCRMA Seminar RoomEvent Type:Hearing SeminarWhat can be better than surprises and music? Are you surprised? What are the neural implications of surprise?
FREEOpen to the PublicHotword Sifter - Make Google Home hear you in cocktail parties
Date:Fri, 03/16/2018 - 10:30am - 12:00pmLocation:CCRMA Seminar RoomEvent Type:Hearing SeminarIt's the last unsolved problem in speech recognition. :-) How does one use minimal resources to recognize hotwords like "Hey Google" or "Alexa", in the face of noisy 24x7 environments. These keywords (or hotwords) are the signal to turn on full speech recognition on the new generation of always listening devices, and the hotword must be recognized accurately, and in multiple acoustic environments. Recognizing these keywords must be done on the device, as nobody wants all their speech transmitted to the cloud.
FREEOpen to the PublicHearing Sound with Auditory Hair Bundles
Date:Fri, 03/09/2018 - 10:30am - 12:00pmLocation:CCRMA Seminar RoomEvent Type:Hearing SeminarHow is all that beautiful music converted from acoustic waves into spikes? The key biological component are some very sensitive cilia, hair-like structures, that sit at the end of auditory nerve cells. They move in response to passing waves, and that unleases spikes the head toward the auditory brain. It really is quite magical.
Who: Wisam Reid (Stanford)
What: Relating the Cohesiveness of Auditory Hair Bundles in Mammals to their Function
When: Friday March 9th at 10:30AM
Where: CCRMA Seminar Room (Top Floor of the Knoll)
Why: These are really quite amazing structures.
Bring your cilia to CCRMA and Wisam will talk about how they do their important job.
FREEOpen to the PublicYaqing Su on Neural Coding of Pitch Cues in the Auditory Midbrain
Date:Fri, 02/16/2018 - 10:30am - 12:00pmLocation:CCRMA Seminar RoomEvent Type:Hearing SeminarNeural Coding of Pitch Cues in the Auditory MidbrainNeural Coding of Pitch Cues in the Auditory Midbrain music perception that pitch??!!?!? And just how do we perceive pitch? After many decades it’s still a source of many disagreements, and much research.
Pitch is critical to auditory perception. It’s an important cue for separating sounds in a cocktail party. It binds a speech signal together so you hear speech instead of a bunch of chirps. Yet, the exact mechanism that our brain uses to hear the pitch of a signal is not known. Yaqing Su will be talking about evidence she has found that suggests that pitch is perceived with a combination of measures, effectively sidestepping the argument about place vs. time. This will be fun.FREEOpen to the PublicDecoding EEG Signals by Malcolm Slaney
Date:Fri, 02/02/2018 - 10:30am - 12:00pmLocation:CCRMA Seminar RoomEvent Type:Hearing SeminarThere are many ways to interpret EEG signals, but the newest approach eschew averaging and perform single-shot decoding. This allows for real-time information about how a subject is perceiving a sound. How does one connect arbitrary audio to the resulting EEG signals? Or how does one predict the audio that resulted in a particular EEG signal? Can I tell you the sound you are attending to? I'll be talking about recent work using linear and correlation methods to build better models that connect audition and EEG signals. This a form of system identification, and has been shown to work for decoding which of two audio signals one is attending to. Cool stuff.
Who: Malcolm Slaney (Google and CCRMA)
What: EEG Decoding Algorithms and ResultsFREEOpen to the PublicModeling fine time structure in the brain
Date:Fri, 01/19/2018 - 10:30am - 12:00pmLocation:CCRMA Seminar RoomEvent Type:Hearing SeminarWe really don't know how auditory *information* is represented in the brain. The easiest approach relies on the rate profile, simply a measure of how fast a group of neurons is firing. But the rate profile ignores a lot of information in the exact timing of each spike. And at a higher level there are oscillatory representations. All of these representations can be used to encode information, but which is most important and leads to the best model of auditory perception?
FREEOpen to the PublicAudio Style Transformations using Deep Neural Networks by Prateek Verma
Date:Fri, 01/12/2018 - 10:30am - 12:00pmLocation:CCRMA Seminar RoomEvent Type:Hearing Seminar
Deep Neural Networks (DNNs) have been wildly successful for many tasks, but none of the tasks are as wondrous as the success that DNNs have had on transferring the style of one painting to another painter’s art. This magical trick is accomplished by mixing and matching the low-level feature analysis layers between different styles of paining, so a painting in one style is rendered in another with different kinds of brush works.
But can we do this for audio? Prateek Verma has been experimenting with this, and will talk about his work and results. What does audio style mean, and how does one capture it?
Who: Prateek Verma (Stanford CCRMA)
What: Audio Style Transformations using Deep Neural NetworksFREEOpen to the PublicElisabetta Chicca on spike-based learning
Date:Thu, 12/07/2017 - 5:00pm - 5:50pmLocation:CCRMA Seminar RoomEvent Type:Hearing SeminarSpike...Spike.....Spike..Spike........Spike. That's how the brain communicates. But how do you spikes help you learn? And more importantly, how do you modify a spiking network to learn new things (like speech and music)? The colloquial rule is "those who fire together, wire together" but in practice you need causality. A more realistic learning model is called spike-timing-dependent plasticity, where a particular neural connection is strengthened when an input spike immediately precedes the output spike. An input spike must happen before the output. This is the backprop, for those of you from the machine-learning or DNN world, of the real neural world.
FREEOpen to the Public
