Research
Funded Research
Model-based, single-microphone speech enhancement
Electrical Engineering Professor Phillip DeLeon has received a grant from the Sarnoff Corporation on behalf of Rosettex Technology for his project entitled “Model-Based, Single-Microphone Speech Enhancement.” This project seeks to develop algorithms and software which can remove noise from single-microphone, noisy speech signals. The algorithms will first be trained using an individual's "clean" speech signal and a statistical model constructed for the speaker. Then noisy speech signals from the user are evaluated against the model and adaptively filtered in order to remove the noise. This work is guided by prior research in speaker identification where individual models are developed for a population and an unknown speaker is identified based on a speech sample. The award is for $178,571.