Explore the intricacies of speech perception and prediction in this 41-minute lecture by Matt Davis from the MRC Cognition and Brain Sciences Unit. Delve into the complexities of cochlear implants and vocoded speech simulation, examining how these technologies impact listening effort and speech comprehension. Investigate the systems involved in perceiving and producing syllables, and learn about Representational Similarity Analysis for speech. Discover the application of Bayes Theorem in speech perception and compare machine and human capabilities in this domain. Analyze the predictive coding view of speech perception, exploring how prior knowledge influences our understanding of speech. Examine perceptual learning of vocoded speech and its relationship to prior knowledge. Conclude by studying a computational model that illustrates the predictive coding view of perception and learning in speech processing.
Overview
Syllabus
Intro
A cochlear implant
Three ways to improve speech perception (1) Make the signal clearer (or noise quieter)
Simulating Cochlear Implants with Vocoded Speech
Listening Effort
Systems for perceiving and producing syllables
Representational Similarity Analysis for Speech
Bayes Theorem
Machines and Humans
Prediction is very difficult, especially about the future
Predictive coding view of speech perception
Prior knowledge and perception of speech
Predictive coding and speech perception
Perceptual learning of Vocoded Speech
Perceptual Learning and Prior Knowledge
Predictive coding of speech perception and learning
Computational model
Predictive Coding View of Perception and Learning
Taught by
MITCBMM
