A dynamic gap dimension reduction approach for high order n-gram phonotactic language recognition

In this paper, we demonstrate the feasibility and usefulness of using high-order n-grams for n = 4, 5, 6,7 in SVM-based phonotactic language recognition by using a dynamic gap dimension reduction algorithm, which can reduce the affection of insert error as the same time. For evaluating the approaches, experiments were carried out on the NIST-LRE2009 database in which systems were built by means of open software (HTK, SRILM) and an English phone decoder. The experimental results on NIST-LRE2009 30s, 10s, 3s closed-set test by proposed method are 2.77%, 7.87%, 19.73% (meaning a relative reduction of 6.73%, 5.74%, 8.44% compared with the baseline system) in terms of equal error rate (EER) and the fusion of the systems with the proposed approaches yielded 2.26%, 7.62%, 18.06%. Details of implementation and experimental results are presented in this paper.