The University of Auckland

Project #111: Can speaker verification be distorted via movement?

Back

Description:

 

These days speech recordings from mobile phones are increasingly being used as Forensic evidence in court.  One of the things about mobile phones is that people can move around considerably as they speak into their phones.  The question is how much does this movement impact on the speech signal? This project will involve creating a corpus of speech when people are moving and being still (lying, sitting, standing) and then directly measuring the changes in speech features extracted from the speech signal. Finally, the project will investigate whether movement impacts on a person’s speech sufficiently to make it difficult for speaker identification. The findings from this project have implications for speech forensics, speaker recognition, speech recognition and speech coding.

Outcome:

Prerequisites

None

Specialisations

Categories

Supervisor

Co-supervisor

Team

Lab

Lab allocations have not been finalised