Hanyang University said on the 23rd that Professor Jang Jun-hyeok's team (Department of Convergence Electronics) ranked first among domestic universities from the INTERSPEECH SASV (spoofing-aware speaker variation) Challenge 2022, the most well-known international academic conference regarding voice artificial intelligence (AI).

Hosted by the International Speech Communication Association (ISCA) and sponsored by NAVER, the competition was held from January to March to develop and share technologies to defend against spoofing attacks in the field of speaker recognition.

Hanyang University team of master's degree and PhD students Choi Jeong-hwan, Yang Jun-yeong, and Jeong Ye-rin (advisory professor Jang Jun-hyeok) developed a core deep learning algorithm that defends against external spoofing attacks such as voice synthesis or voice transfer. Especially, they showed a creative artificial intelligence (AI) technology that makes a deep learning model to learn and include knowledge of spoofed voice into speaker embedding.

The algorithm's performance score that Hanyang University submitted ranked third among other 23 universities and organizations from all over the world and ranked first among domestic universities. The awarding ceremony at the ISCA INTERSPEECH that will be held at Incheon Songdo Convention Center in September.