Development of an automatic speech analysis system based on human speech cognition models
The Speech Communication Group has worked on a speech perception model based on acoustic cues to distinctive features called landmarks. Landmarks can provide information on articulator-free features of speech sounds (such as obstruent and sonorant) as well as distinctive features that specify place of articulation and voicing. This speech information is valuable for sppech analysis and the Speech Communication Group is developing a speech analysis system based on this approach. This system is nearing its completion; the SuperUROP research project will finish implementation by integrating existing feature cue detection modules and developing remaining modules. The system will have applications to many areas including clinical speech analysis and automatic speech recognition.

I have had in an interest in programming for artificial intelligence and music for a long time. These interests led me to my SuperUROP project which works with a novel approach to speech analysis. I have previously taken linguistics machine learning and signals and systems which will be sure to help me. By the end of this project I hope to learn about how speech recognition programs work.