Current automatic speech-recognition system approaches tend not to use information about speech production when analyzing speech signals. This project focuses on making improvements to an automatic speech-recognition system that uses acoustic cues and speech signal features to generate the words spoken from a signal. The goal of my project is to improve the system’ s accuracy by finding ways to integrate new information into the system’ s model. One way of achieving that is by using formant frequencies to infer different properties about the speaker and using Bayesian conditioning to update the probability distributions used by the model.

I am participating in SuperUROP because I would like to gain more research experience and apply what I have learned from previous research to the current project. I hope to learn more about acoustic signal analysis and how to apply this knowledge to the project. I am very excited to learn more about the applications of machine learning to natural languages and signals.