In this article, we propose a new feature which could be used for the framework of SVM-based language\r\nrecognition, by introducing the idea of total variability used in speaker recognition to language recognition. We\r\nconsider the new feature as low-dimensional representation of Gaussian mixture model supervector. Thus we\r\npropose multiple total variability (MTV) language recognition system based on total variability (TV) language\r\nrecognition system. Our experiments show that the total factor vector includes the language dependent\r\ninformation; what�s more, multiple total factor vector contains more language dependent information.\r\nExperimental results on 2007 National Institute of Standards and Technology (NIST) Language Recognition\r\nEvaluation (LRE) databases show that MTV outperforms TV in 30 s tasks, and both TV and MTV systems can achieve\r\nperformance similar to that obtained by state-of-the-art approaches. Best performance of our acoustic language\r\nrecognition systems can be further improved by combining these two new systems.
Loading....