Application of GMM models to spoken language recognition
Adam Dustor, Paweł Szwarc · International Conference Mixed Design of Integrated Circuits and Systems · 2009
This paper presents research on automatic spoken language recognition based on statistical pattern recognition. As a model of identified language Gaussian Mixture Model was applied, both with diagonal and full covariance matrix. The influence of GMM order and parameterizations of speech signal on the recognition results were examined. Tests were done for 10 languages. Obtained results were discussed.