You Only Listen Once
The YOLO system is a practical speech-only login system. The system scales to many users, requires few speech samples for registration and needs only one speech sample at login. To prevent an impersonation attack (the primary weakness of most speech authentication systems), a novel impostor detection method based on an OTF speech recognition model is used. There are a number of pressing use cases for such technology, such as
- Transaction Authentication
- Password Reset via Voice
- Voice Controlled Locks (eg. opening a door or unlocking a car)
- Access Control for Computers
- Information Retrieval
System Diagram
Speech Processing
Speaker Verification
Speaker Identification
Presence Detection
ECE Capstone Project @ Carnegie Mellon
Authors: Ryan Brigden, Nikhil Rangarajan, and Richa Ravi