Skip to main content

Voice recognition software can be duped: UW-Madison study

Professor of electrical and computer engineering Kassem Fawaz and doctoral student Shimaa Ahmed have developed a mathematical model that could allow almost anyone to imitate the resonance of another person’s voice with a simple tube. As long as a person could nail down the frequency of a voice with a PVC pipe that, based on the algorithm, had the correct width and length, they could trick the security technology 60% of the time.