A Rose by Any Other Name
First, Let's Define Our Terms
One of the most bewildering aspects of learning about an industry or technology is navigating through a jungle of unfamiliar terminology. If you are entering the world of voice ID with existing knowledge about speech recognition, the terminology pitfalls become even more treacherous because seemingly identical labels can have incompatible meanings. In this column, I define several terms you will encounter when you begin to explore the voice ID industry:
Speaker Verification: A technology, process, or product for verifying that a person is who she or he claims to be. It has a variety of other names including speaker authentication, voice verification, and voice authentication.
For example: I want to access a computer network or other system that is protected by speaker verification. The speaker verification security will ask me to identify myself. When I claim to be Oprah Winfrey, the system takes a sample of my voice and compares it with a stored sample of Oprah's voice. Since I am not Oprah Winfrey the system will identify me as an impostor and will not allow me access to the system.
Speaker Identification: A technology, process, or product for attaching an identity to the voice of an unknown person. A synonym is speaker recognition.
For example: a department store brings a tape recording of a bomb threat to a law-enforcement agency with a speaker identification system. The system contains a database of voice samples from known terrorists and other criminals. The unknown voice is compared with each of the voices in the database (or a clearly identified sub-group, such as male voices) to determine whether the speaker is known to the law-enforcement agency.
Voice ID: The name I use for the industry that provides speaker verification, speaker identification, and related products and technology. I also use this term when referring to both speaker verification and speaker identification (e.g., Voice ID products). There may be a better term. I would certainly like to hear some good suggestions.
Speaker Recognition: According to researchers and developers who have been in the industry for a long time, this is the true name for the voice ID industry. They look askance at the term "Voice ID."
There are two problems with the term "Speaker Recognition." One is that this term is also used as a synonym for speaker identification. The other problem is that it sounds almost exactly like Speech Recognition. The confusion that results in both instances is a strong argument for the use of an alternate term, such as Voice ID.
From time to time I will present more definitions, but if there are any terms that are causing you difficulty, please let me know by sending me e-mail via Speech Technology Magazine at speechmag@aol.com .
As a final note, I want to point out that there was a typographical error in the Voice IDeas column in the June/July issue. This column reported that "losses from telephone toll fraud have reached $3 million a year." Unfortunately, the actual losses amount to $3 billion a year.
Judith Markowitz is the technology editor of Speech Technology Magazineand is a leading independent analyst in the speech technology and voicebiometric fields. She can be reached at (773) 769-9243 or jmarkowitz@pobox.com.