Approaches to Language Identification Using Gaussian Mixture Model and Linear Discriminant Analysis

The baseline system PRLM has the best performance on NIST language recognition evaluation tasks. But this system needs orthographically or phonetically transcribed utterances which can not be easily obtained from Chinese dialects and minority languages. So, the PRLM system is not used to these languages. To overcome this limitation, we present the Gaussian mixture model recognizer followed by language-dependent language model (GMM-LM) as an approach to language identification. In this paper, we focus on finding the optimum number of frames to train each GMM parameter and comparing two back-end processing approaches in GMM-LM system. The experiments show that the LDA processing approach can achieve average accuracy 78%, which is a 45% relative improvement over simple approach on 30s test data.