The 1997 classification evaluation is a text independent speaker detection (verification) task. the training set is composed of ``One session'', ``One handset'' and ``Two handset'' data. the speech duration for each speaker on each training condition is 1 minute.
The test set has 3 different speech duration 3 seconds, 10 seconds and 30 seconds. these tests segments have to be used on the three different training conditions.
Two types of results have to be given:
were
Details of the evaluation protocol are available on [4].
The 1997 evaluation focused on the different handset conditions.