Hearing Seminars
CCRMA hosts a weekly Hearing seminar (aka Music 319). 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. Stanford students can (optionally) receive credit to attend, by enrolling in Music 319 "Research Seminar on Computational Models of Sound Perception." 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
Bryan Pardo on Voice Assist and Cerberus
Date:Fri, 03/06/2020 - 10:30am - 12:00pmLocation:CCRMA Seminar RoomEvent Type:Hearing SeminarIn this talk, Bryan Pardo will describe two recent projects developed at Northwestern University’s Interactive Audio Lab: VoiceAssist and Cerberus. VoiceAssist is a system that helps inexperienced users produce high quality recordings by providing real-time visual feedback on audio quality. Cerberus is a single deep learning architecture that can both separate an audio recording of a musical mixture into constituent single-instrument recordings and transcribe these instruments into a human-readable format at the same time.
FREEOpen to the PublicDecoding Speech and Language Representations from the Brain
Date:Fri, 02/21/2020 - 10:30am - 12:00pmLocation:CCRMA Seminar RoomEvent Type:Hearing SeminarGopala Anumanchipalli will be talking about his work to measure and reconstruct speech signals from human neural activity. Gopala is uniquely situated to do this work, having graduated from Prof. Alan Black’s speech synthesis lab at CMU and he is now part of Prof. Eddie Chang’s excellent ECoG lab at UCSF. At UCSF there are patients who are undergoing surgery for epilepsy, and some of the patients allow recordings to be made as they listen and respond to sounds. One important question is how is speech represented in the brain. Gopala’s work uses deep neural networks (DNNs) to reconstruct speech sounds from human ECoG data. Speech from spikes.
AbstractFREEOpen to the PublicSource coding of audio signals with a generative model
Date:Fri, 02/07/2020 - 10:30am - 12:00pmLocation:CCRMA Seminar RoomEvent Type:Hearing SeminarAudio coders have historically taken two approaches: 1) source coders that understand how a signal is generated, often from a vocal tract, and generate representation using a few time-varying quantized parameters, and 2) perceptual coders that capitalize on “flaws” in the perceptual system that drop parts of the signal are not heard.
FREEOpen to the PublicOpen tools for the development and evaluation of hearing devices
Date:Thu, 12/12/2019 - 4:00pm - 5:30pmLocation:CCRMA Seminar RoomEvent Type:Hearing Seminar*** Date change, now Thursday! Sorrry. ***
There is a wealth of knowledge about how to adjust our hearing, but most of it is locked up in commercial products. Recently the US National Institutes of Health sponsored a research program to develop a suite of hardware and software that are open source. By this means they hope to create a new research ecosystem that can build and test improvements to the current paradigms. And of course the open source ecosystem will make it easier to put new products and ideas into the field.FREEOpen to the PublicDecoding inner speech from intracranial recordings in the human brain
Date:Mon, 11/25/2019 - 4:00pm - 5:30pmLocation:CCRMA Seminar RoomEvent Type:Hearing SeminarStephanie Martin will talk about decoding imagined speech activity from just brain signals.
FREEOpen to the PublicAuditory Separation of a Conversation from Background via Attentional Gating
Date:Fri, 10/11/2019 - 10:30am - 12:00pmLocation:CCRMA Seminar RoomEvent Type:Hearing Seminar
The latest speech enhancement work has the potential to dramatically change the way we hear the world around us. This new work has dramatically improved the quality and latency of these algorithms, and it has the potential to change the way we hear the world around us, whether we have normal hearing or need assistance. These new systems build highly sophisticated models of speech, and can pick out the speech signal from the noise. Oh, yes.FREEOpen to the PublicAlain de Cheveigne on cleaning up brain data for analysis and decoding
Date:Fri, 10/04/2019 - 10:30am - 12:00pmLocation:CCRMA Seminar RoomEvent Type:Hearing Seminar
Brain signals as measured by EEG, MEG or even ECoG are inherently noisy. Not only are there only a few dozen sensors to measure billions of different neural sources, the electrical environment can change during an experiment. One would like techniques that can pull the signal out of the noise. This can be done with smart forms of noise control, de-trending and signal averaging.Alain de Cheveigne will be at CCRMA on Friday to discuss a panoply of techniques to enable you to find the signals you care about.FREEOpen to the PublicAndrew Oxenham: How far does musical training generalize?
Date:Fri, 09/27/2019 - 3:00pm - 4:30pmLocation:CCRMA Seminar RoomEvent Type:Hearing Seminar
I’m really pleased that Prof. Andrew Oxenham will be here at CCRMA on Friday afternoon to talk about: How far does musical training generalize? This is a special time (3PM) to accommodate his travel schedule. I’m glad we get to welcome him to Stanford.
Andrew has an amazing record studying auditory perception. He has looked at some of the most important problems in psychoacoustics and neurophysiology, with creative and robust experiments.FREEOpen to the PublicLearning Audio Embeddings: From Signal Representation, Audio Transformation to Understanding
Date:Fri, 05/31/2019 - 10:30am - 12:00pmLocation:CCRMA Seminar RoomEvent Type:Hearing SeminarPrateek Verma will lead a discussion about using embedding spaces, trained using deep neural networks (DNNs), to model and perform amazing feats with music and speech signals. This continues the theme from last week’s Hearing Seminar, where Rohit talked about using DNNs for speech recognition.
FREEOpen to the PublicRohit Prabhavalkar on Modern DNN Speech Recogntion
Date:Fri, 05/24/2019 - 10:30am - 12:00pmLocation:CCRMA Seminar RoomEvent Type:Hearing SeminarModern speech recognition works really well, especially compared to the HMM approaches of even 10 years ago. Most notably, current approaches abandon the feature calculations, timing models, and language models of yore, replacing them all with a single deep neural network. Amazing, and they work really well, converting millions of hours of speech to text, in all sorts of auditory conditions, in a tiny box that fits in our pockets, for everyday users.
FREEOpen to the Public
- 1 of 14
- ››
