Research on Automatic Language Identification System for Internet Speech
Jun Hai Guo · Computer and Information Technology · 2007
This paper reports research on automatic language identification with speeches of English,German,Japanese,French and Spanish downloaded/recorded from Internet and Chinese speech from exist corpus.GMM is built for each language with RASTA-PLP parameters and Greedy EM method.Open test with speeches from a variety of speakers is performed to verify the efficiency of the acoustic-based method for ALID of Internet speech.The effects of internet speech and non-internet speech on the identification result,as well as relations between the recognition rate,quantity of the training data and the number of GMM components are discussed.