Improved Phonotactic Language Recognition Using Collaborated Language Model

In this paper an approach to build a collaborated feature supervector is proposed and a collaborated language model (CLM) builded with binary decision tree feature surpervector and N-gram feature supervector is introduced and applied to deal with the problems of language identification tasks such as handling long term contexts and too many parameters. Experiments are carried out on the database of National Institute of Standards and Technology language recognition evaluation 2009 (NIST LRE 2009). The experimental results have confirmed that phonotactic language recognition system using the collaborated language model yields 1.07%, 2.68%, 13.48% in equal error rate (EER), which means 8.54%, 12.70% and 4.60% relative reduction for 30s, 10s, 3s compared to the baseline system, respectively.