Linguistic Cues for Heuristic Determination of Patterns of Speech in Noise

This project aims to investigate the benefits of using linguistic cues for extracting speech-of-interest from competing background noise. Given an algorithm which detects formants of a speaker of interest within an audio signal which includes background noise, and given that the limits of such an approach are reached when formants of the speaker of interest become substantially lost within the noise, are there linguistic cues within some formants, or linguistic relationships between some successive speech sounds, which can be used to infer with confidence which speech sounds from the speaker of interest are present beneath the noise floor of a noisy audio signal. The partner organization will benefit from this internship by extending the range of solutions it can offer to the market. If the proposed solution proves to be useful and novel, the partner organization will apply for patent protection. This could lead to increased sales and the additional employment within the company.

Faculty Supervisor:

Dr. Molly Babel

Student:

Michael McAuliffe

Partner:

Malaspina Labs Inc.

Discipline:

Languages and linguistics

Sector:

Life sciences

University:

University of British Columbia