Naomi Harte - ViSQOL, An objective measure for speech quality
Date:
Fri, 05/01/2015 - 11:30am - 1:00pm
Location:
CCRMA Seminar Room (Top Floor of the Knoll)
Event Type:
Hearing Seminar This talk gives an overview of ViSQOL – the Virtual Speech Quality Objective Listener. It is a signal-based, full-reference, intrusive metric that models human speech quality perception using a spectro-temporal measure of similarity between a reference and a test speech signal. The metric has been designed to be particularly robust for quality issues associated with Voice over IP (VoIP) transmission. The talk will explore how the original idea for associating visual similarity with spectrogram changes developed. I’ll show results from a full evaluation of the metric against PESQ and POLQA in a range of scenarios, including how it handles VoIP degradations. The research to develop ViSQOL was sponsored by Google Chrome in Mountainview CA.
Bio: Dr. Naomi Harte is a lecturer in the School of Engineering at Trinity College Dublin in Ireland. She was appointed as an SFI Engineering Initiative Lecturer in Digital Media in 2008, sponsored by DTS. She spent time as a Research Associate Academic in McMaster University in Canada. Prior to returning to academia, Naomi worked in high-tech start-ups in the field of DSP Systems Development, including her own start-up. Naomi’s specialist area is Human Speech Communication. Her industrial background brings a real-world approach to her research. Current projects focus on speech quality, audio-visual speech recognition, emotion in speech, ageing voices in speaker verification and bird species analysis. She is currently on sabbatical at ICSI in Berkeley.
FREE
Open to the Public