Speech recognition systems have improved due to machine learning research, but this approach does not model human speech perception. This research is part of a larger project which models human speech analysis by processing acoustic cues and detecting signal patterns. A signal processing and cue detection framework is being developed to highlight areas of nasalization in a signal. These areas can be a cue to the distinction between nasal consonants and other consonants, so detecting nasalization is critical for speech understanding. This work has applications in clinical speech analysis and in perception models. The goal is to develop a system that can detect nasalization via Gaussian mixture models and to present a clear pipeline for the processing and modeling necessary to do so.

I am participating in SuperUROP because I am excited to explore speech processing in a research environment. With a solid foundation in computer science and a deep interest in natural language processing, I hope to further my knowledge in speech perception. By conducting research in this area through SuperUROP, I aim to drive a research project from conception to implementation and to bolster my scientific communication skills.