Speaker recognition identifies a person from characteristics of voices. It answers the question who is speaking by analyzing acoustic patterns that reflect anatomy and learned behavioral patterns found between individuals.
The field dates back some four decades as of 2019 with the first international patent arriving on the 1st of January 1983 from CSELTUS4752958 A by Michele Cavazza and Alberto Ciaramella. Worlds of Wonder commercialized early training technology in 1987 with the Julie doll.
The Scobey, Coronach Border Crossing used the technology to allow enrolled local residents to cross the Canada United States border when inspection stations closed at night. Voice Strategies of Warren Michigan developed the system for U.S. Immigration and Naturalization Service.
Vice News and The Guardian demonstrated this capability in 2023 when standard financial speaker-authentication systems fell to AI-generated voices. Attackers needed about five minutes of target's voice samples to succeed against these security measures.
Barclays Wealth became the first financial firm to deploy voice biometrics in 2013 while HSBC offered 15 million customers biometric banking software in February 2016. These systems enabled users to access online accounts using their voice.