Academy of Finland
Funding decision
Name Kinnunen, Tomi
Organisation University of Eastern Finland
Project title NOTCH: NOn-cooperaTive speaker CHaracterization
Decision No. 309629
Decision date 20.06.2017
Funding,period 01.09.2017 - 31.08.2021
Funding 596 515
WebFOCUS Report
Project description
Automatic speaker verification (ASV), the task of recognizing persons from their voices, is used increasingly in applications demanding high security. Similar to physical biometrics, ASV systems can unfortunately be subverted using so-called representation (or spoofing) attacks involving techniques such as replay, impersonation, speech synthesis or speaker transformation. Another problem, known as disguise, involves non-cooperative speakers attempting to avoid being identified as themselves. This project targets at both enhancing state-of-the-art in characterization and detection of non-cooperative speech modifications, as well as using ASV-inspired approaches to improve voice transformation techniques and obtaining parallel human-vs-machine benchmarks concerning spoofed and disguised speech. The project is to be carried out primarily at University of Eastern Finland (UEF) together with world-leading experts in voice modification.