Speech Recognition in the Context of Variability: Lessons from Human Listeners

Explore the complexities of speech recognition and acoustic variability in this lecture by Dr. Shannon Heald from the University of Chicago. Delve into the challenges posed by acoustic-phonetic variability between talkers and learn how human listeners adapt to understand difficult speakers. Discover how leveraging systematic variance in speech signals allows listeners to maintain robust phonetic constancy despite noise or distortions. Examine traditional models of speech perception, the nature of talker variability, and the role of expectations and visual information in recognition. Gain insights into Dr. Heald's research on how listeners reorganize their processing of acoustic features when adapting to new talkers. Understand the implications of this work for developing more robust speech recognition systems that can handle real-world variability.