Accession Number : ADA444863


Title :   Speaker Segmentation and Clustering Using Gender Information


Descriptive Note : Conference paper


Corporate Author : GENERAL DYNAMICS ADVANCED INFORMATION SYSTEMS DAYTON OH


Personal Author(s) : Ore, Brian M ; Slyh, Raymond E ; Hansen, Eric G


Full Text : https://apps.dtic.mil/dtic/tr/fulltext/u2/a444863.pdf


Report Date : Feb 2006


Pagination or Media Count : 10


Abstract : This paper considers the segmentation and clustering of conversational speech for the two-wire training (3conv2w) and two-wire testing (1conv2w) conditions of the NIST 2005 Speaker Recognition Evaluation. A notable feature of the system described is that each file is labeled as containing either opposite- or same-gender speakers The speech segments for opposite-gender files are clustered by gender, while those for same-gender files are processed by agglomerative clustering. By using gender information in the clustering of the opposite-gender files, the equal error rate in the 3conv2w training condition was reduced from 15.2% to 9.9%. For the 1conv2w testing condition, clustering opposite-gender files by gender did not improve performance over agglomerative clustering; however, it was over 100 times faster than agglomerative clustering on the opposite-gender files.


Descriptors :   *SPEECH RECOGNITION , *SEX , CLUSTERING , SEGMENTED , SPEECH


Subject Categories : Voice Communications
      Linguistics


Distribution Statement : APPROVED FOR PUBLIC RELEASE