Reliable Audio Voice Identification

Typical voice comparison systems require consistent or well-controlled collection parameters that are unrealistic in the forensic environment. Real-world recordings may be over varied channels, in the presence of competing noise or speakers, or even recorded while the speaker is under duress. The Massachusetts Institute of Technology's Lincoln Laboratory developed an advanced voice identification system and reference for use in forensic casework. The system compares a new or unknown voice sample to a collection of known voice samples. Different voices saved on various storage media and recorded under diverse conditions can be compared, and unwanted noise can be eliminated. Phone-based speaker recognition technology is also included in this system. Requests for additional information should be sent to isfsubgroup@tswg.gov.