Dialect Distance Assessment Based on 2-Dimensional Pitch Slope Features and Kullback Leibler Divergence
Conference paper (preprint)
TEXAS UNIV AT DALLAS CENTER FOR ROBUST SPEECH SYSTEMS (CRSS)
Pagination or Media Count:
Dialect variations of a language have a severe impact on the performance of speech systems. Knowing how close or separate dialects are in a given language space provides useful information to predict or improve, system performance when there is a mismatch between train and test data. Distance measures have been used in several applications of speech processing, including speech recognition, speech coding, and speech synthesis. Apart from phonetic measures, little if any work has been done on dialect distance measurement. This method of dialect separation assessment based on modeling 2D pitch slope patterns within dialects is proposed. Kullback-Leibler divergence is employed to compare the obtained statistical models. The presented scheme is evaluated on a corpus of Arabic dialects. The sensitivity of the proposed measure to changes on input data is quantified. It is also shown in a perceptive evaluation that the presented objective approach of dialect distance measurement correlates well with subjective distances.
- Voice Communications