Speaker Matching criterion (voice recognition related)?

Leonardo

Hi folks,

I'm not a native English speaker but I work as a localizer and translator at a company that develops voice recording and recognition software.
Recently, there has been a good deal of debate about one of the phrases in the GUI of one of our software solutions, if it was rendered into English correctly or not.
The case is that we have a fetaure that analyzes two recordings and can tell you if the speaker on recording 1 and 2 is the same or whether they are two different persons. Now, there is a basic criterion that allows software to tell you if the speaker is the same person or if there are two of them. The questions is how to properly name this criterion in English? Would "Speaker Matching Criterion" do? The disputable word here is Matching, but if Criterion is not good and there's a better option, please mention that, too.

Many thanks in advance.
Hey Bander,

Thanks for your response.
Well, the key point is not recognizing allophones but making a conclusion if the speaker is the same on both recordings or if they are different. The software makes decision itself, although the "listener" who's working with the software may change the decision manually.
When "coining" the "matching" term I was thinking about an identic situation and remembered we are given "passwords do not match" message when password and password confirmation are different. The similiarity lies in there being two objects that may be the same or different, and in the example above "match" is used. Is our case totally different then?
Btw, I also googled it and found there is "voice matching software" term used for the same thing. Maybe Voice Matching Criteria will be a good option then?

Thank you!

Answer
Yes,"matching" may not be a good choice. It seems that the system you developed can regognize the different sounds of a consonant (allophones), and interpret them as identical (matching). In this special apparatus, the process would be "categorical perception", similar to human perception of speach sounds. And since recognition of sounds here is eventually interpreted by the listener (not the speaker), you should maybe call it "speech categorical perception (apparatus)".

Edit: On the allophones and our peculiar ways of articulating sounds, I said it's impossible to have a match if the apparatus was producing spectographs of the speech signals. We are not machines; however, we can regognize the speaker's voice regardless of how sounds are produced. That's why I thought of "categorical perception". Your system does that, but I don't know how.

give me details of voice security system?

pandu

Answer
Suggest you read http://www.biometrics.gov/docs/speakerrec.pdf

or search on speaker recognition


NIST conducted yearly evaluations of speaker recognition system
http://www.nist.gov/speech/tests/spk/index.htm

Powered by Yahoo! Answers

Title Post: Speaker Matching criterion (voice recognition related)?
Rating: 98% based on 989 ratings. 5 user reviews.
Author: Yukie

Thanks For Coming To My Blog