Speaker Identification in a Collaborative Environment

At present, there is an increasing demand from industries for new generation of interactive technologies to support productivity and minimize time wasted for meeting room set up. In response to this demand, SMART technologies is investing into development of new UC&C technologies. The goal of this research will be the development of a software for reliable speaker identification and authentication in a meeting room environment. The software will be able to recognize users in the meeting room with the flexibility in dealing with noise and variable recording quality, including various voice samples. The developed software will take into an account the various machine learning methods to deal with proper training, the insufficient amount of samples, and the physical characteristics of data sets used to validate the methodology. It will be adaptive to different types of voice input data: the live voice and the voice from the teleconference/cell phone. This would require development of sophisticated machine learning methods which will also further enhance the areas of data mining, pattern recognition, and signal processing.

Faculty Supervisor:

Marina Gavrilova

Student:

Faisal Ahmed

Partner:

SMART Technologies

Discipline:

Computer science

Sector:

Information and communications technologies

University:

University of Calgary