This paper presents a series of dialect/accent identification results for three sets of dialects with discriminatively trained Gaussian mixture models and feature compensation using eigen-channel decomposition. The classification tasks evaluated in the paper include: 1) the Chinese language classes, 2) American and Indian accented English and 3) discrimination between three Arabic dialects. The first two tasks were evaluated on the 2007 NIST LRE corpus. The Arabic discrimination task was evaluated using data derived from the LDC Arabic set collected by Appen. Analysis is performed for the English accent problem studied and an approach to open set dialect scoring is introduced. The system resulted in equal error rates at or below 10% for each of the tasks studied.
Bibliographic reference. Torres-Carrasquillo, Pedro A. / Sturim, Douglas E. / Reynolds, Douglas A. / McCree, Alan (2008): "Eigen-channel compensation and discriminatively trained Gaussian mixture models for dialect and accent recognition", In INTERSPEECH-2008, 723-726.