The cocktail party problem

Slides:



Advertisements
Similar presentations
Cochlear implants Current Biology
Advertisements

Gemmata obscuriglobus
Calibrating color vision
Multisensory Integration: What You See Is Where You Hear
Volume 24, Issue 23, Pages R1109-R1111 (December 2014)
Jean-Paul Noel, Mark Wallace, Randolph Blake  Current Biology 
Brain Responses in 4-Month-Old Infants Are Already Language Specific
Victoria F. Ratcliffe, David Reby  Current Biology 
Individual Differences Reveal the Basis of Consonance
Gene Evolution: Getting Something from Nothing
Generalizable Learning: Practice Makes Perfect — But at What?
Comparative Cognition: Action Imitation Using Episodic Memory
Sensory-Motor Integration: More Variability Reduces Individuality
Whistled Turkish alters language asymmetries
Visual Development: Learning Not to See
Cell Biology: Microtubule Collisions to the Rescue
Somatosensory Precision in Speech Production
Josh H. McDermott, Eero P. Simoncelli  Neuron 
Mosaic and regulative development: two faces of one coin
Volume 21, Issue 20, Pages R837-R838 (October 2011)
The Lombard effect Current Biology
Insect Neurobiology: An Eye to Forward Motion
Norm-Based Coding of Voice Identity in Human Auditory Cortex
Multisensory Integration: What You See Is Where You Hear
Infant cognition Current Biology
Kinesthetic information disambiguates visual motion signals
Thigmomorphogenesis Current Biology
Volume 25, Issue 24, Pages R1156-R1158 (December 2015)
Volume 20, Issue 6, Pages R282-R284 (March 2010)
Volume 23, Issue 23, Pages R1025-R1026 (December 2013)
Microbial Ecology: Community Coalescence Stirs Things Up
Motor Networks: Shifting Coalitions
Insect Hearing: Active Amplification in Tympanal Ears
Evolution of human vocal production
RecA Current Biology Volume 17, Issue 11, Pages R395-R397 (June 2007)
Visual Attention: Size Matters
Brain Responses in 4-Month-Old Infants Are Already Language Specific
Immediate susceptibility to visual illusions after sight onset
Honeybee Communication: A Signal for Danger
Vision: When Does Looking Bigger Mean Seeing Better?
The many colours of ‘the dress’
Cochlear implants Current Biology
Volume 54, Issue 6, Pages (June 2007)
Myopia: The Importance of Seeing Fine Detail
Fifty years of illumination about the natural levels of adaptation
Gene Evolution: Getting Something from Nothing
Volume 24, Issue 7, Pages R262-R263 (March 2014)
Elementary motion detectors
Attentive Tracking of Sound Sources
Evolution of sound localisation in land vertebrates
Volume 15, Issue 13, Pages R483-R484 (July 2005)
Mosaic and regulative development: two faces of one coin
Visual Development: Learning Not to See
Centrosome Size: Scaling Without Measuring
FOXO transcription factors
Function and Evolution of Vibrato-like Frequency Modulation in Mammals
Volume 17, Issue 15, Pages (August 2007)
Small RNAs: How Seeds Remember To Obey Their Mother
The challenge of measuring long-term positive aftereffects
Drosophila Connectomics: Mapping the Larval Eye’s Mind
Volume 28, Issue 2, Pages R58-R60 (January 2018)
Neuroscience: Tiny Eye Movements Link Vision and Attention
Adaptive Diversity: Hormones and Metabolism in Freshwaters
The Lombard effect Current Biology
Animal Communication: Origins of Sequential Structure in Birdsong
Basal bodies Current Biology
Thigmomorphogenesis Current Biology
Circadian Biology: The Early Bird Catches the Morning Shift
Volume 18, Issue 5, Pages R198-R202 (March 2008)
Volume 24, Issue 11, Pages R508-R510 (June 2014)
Presentation transcript:

The cocktail party problem Josh H. McDermott  Current Biology  Volume 19, Issue 22, Pages R1024-R1027 (December 2009) DOI: 10.1016/j.cub.2009.09.005 Copyright © 2009 Elsevier Ltd Terms and Conditions

Figure 1 A typical Manhattan cocktail party. The listener must follow the conversation of interest despite many concurrent sources of sound. (Image from Breakfast at Tiffany's: Paramount Pictures.) Current Biology 2009 19, R1024-R1027DOI: (10.1016/j.cub.2009.09.005) Copyright © 2009 Elsevier Ltd Terms and Conditions

Figure 2 Cocktail party acoustics. Spectrograms of a single ‘target’ utterance (top row), and the same utterance mixed with one, three and seven additional speech signals from different speakers. The mixtures approximate the signal that would enter the ear if the additional speakers were talking as loud as the target speaker, but were standing twice as far away from the listener (as might occur in cocktail party conditions). Spectrograms were computed from a filter bank with bandwidths and frequency spacing similar to those in the ear. Each spectrogram pixel represents the rms amplitude of the signal within a frequency band and time window. The spectrogram thus omits local phase information, which listeners are insensitive to in most cases. The grayscale denotes attenuation (in dB) from the maximum amplitude of all the pixels in all of the spectrograms, such that gray levels can be compared across spectrograms. Acoustic cues believed to contribute to sound segregation are indicated in the spectrogram of the target speech (top row). Spectrograms in the right column are identical to those on the left except for the superimposed color masks. Pixels labeled green are those where the original target speech signal is more than −50 dB but the mixture level is at least 5 dB higher. Pixels labeled red are those where the target was less and the mixture was more than −50 dB in amplitude. The sound signals used to generate the spectrograms can be listened to at http://www.cns.nyu.edu/∼jhm/cocktail_examples/. Current Biology 2009 19, R1024-R1027DOI: (10.1016/j.cub.2009.09.005) Copyright © 2009 Elsevier Ltd Terms and Conditions