February 26, 2001

The readings for this class were:

• Whitney, Chapter 5

The outline for the class is:

1. Announcements
2. Questions
3. Auditory system description
   • How do we Hear?
   • Sound
   • resonance
   • the auditory pathway
   • Wernicke's Aphasia
4. What information is used to identify and distinguish speech sounds?
   • "She uses standard oil." has exactly 3 breaks in the sound pattern, none of them between words. In fact, all 3 occur in the word "standard."
   • speech spectrogram example
   • Spectral Cues Homepage
   • sinewave synthesis
5. Where do categories of speech perception come from?
   • Kuhl and speech perception
   • Waveforms for /pa/ and /ba/
   • The Motor Theory of Speech Perception
   • FLMP and the McGurk Effect
6. Models of word recognition
   • Cohort model
   • TRACE model
   • Neural dynamics of phonemic integration
     • Gray Ship example from Grossberg & Myers, 2000.

Digressions/Miscellaneous

• Laddle Rat Rotten Hut
• The Shaw Brothers, "Alfalfalal Talk"

Overhead Notes

Parting Thoughts

• Read Whitney, Chapter 6

Self-test Questions

• Explain why it is easier to get an automated speech recognition system to perform well if it only has to recognize words spoken by a particular individual. Next, imagine that you wanted to improve the performance of an automated system by having it make use of top-down processing in word recognition. What kinds of information would you have to build into the system's memory?
• The word "dog" is recognized faster than the word "distance" at the end of a sentence beginning with, "The boy walked the ..." Contrast the explanations for this effect that would be provided by the cohort model and the TRACE model.
• A la "Laddle Rat Rotten Hut," create a sentence where the sounds of the words have an auditory pattern of a different set of words. For example, "I scream!" could be "Eyes cream!" or "Ice cream!"