ODYSSEY 2004 - The Speaker and Language Recognition Workshop

May 31 - June 3, 2004
Toledo, Spain

Analysis of Multitarget Detection for Speaker and Language Recognition

Elliot Singer, Douglas A. Reynolds

MIT Lincoln Laboratory, Lexington, MA, USA

The general multitarget detection (open-set identification) task is the intersection of the more familiar tasks of close-set identification and open-set verification/detection. In the multitarget detection task, an input of unknown class is processed by a bank of parallel detectors and a decision is required as to whether the input is from among the target classes and, if so, which one. In this paper, we show analytically how the performance of a multitarget detector can be predicted from the open-set detection performance of the individual detectors of which it is constructed. We use this analytical framework to establish the relationship between the multitarget detector’s closed-set identification error rate and its open-set detector miss and false alarm probabilities. Experiments performed using standard speaker and language corpora are described that demonstrate the validity of the analysis.

Full Paper

Bibliographic reference.  Singer, Elliot / Reynolds, Douglas A. (2004): "Analysis of multitarget detection for speaker and language recognition", In ODYS-2004, 301-308.