| With the rapid development of computer technology and the Internet,computer assisted language learning has aroused widespread concern and got important use in the field of education,more and more students use computer assisted language learning as their language learning method.It will change the traditional way of language learning and the method of teaching greatly and improve the efficiency of language learning effectively.Automatic speech evaluation is one of the key techniques in computer assisted language learning,it is to use computer technology to evaluate the learner’s pronunciation, it can give the score,correct the incorrect pronunciation and help the learners to correct pronunciation errors effectively,to improve the learners’ learning efficiency further.This paper studies the pronunciation evaluation method commonly used at present and improves the improved KLD algorithm and proposes the minimum phoneme pair algorithm.For a phoneme,using the confusion probability among phonemes to constitute a confusion phoneme set,then in accordance with the confusion probability from big to small sort confusion phoneme set again,the two elements of the highest phoneme confusion probability is the phoneme’s minimum phoneme pair.The minimum phoneme pair algorithm can improve the accuracy of forced alignment,the precision of calculation and the distinguish ability of phoneme confusion model,promote of the performance of the system further.Through the experiment analysis,the performance of minimum phoneme pair algorithm is improved to 0.814,compared with improved KLD algorithm, the training time is reduced,some progress has been made.According to the widely existed pronunciation accuracy judging problems in language learning currently,this paper designs and develops a Mandarin online learning system based on automatic speech evaluation technology.The system consists of user management module,course selection module and online learning and evaluation module.Course selection module let the learners choose course according to their own learning needs and have more autonomy. Online learning and evaluation module allow learners to evaluate their pronunciation in real time.According to pronunciation,the system will give the corresponding pronunciation score and the improvement guidanceof pronunciation error.The teaching effect test shows that 88.7% of the learners think that the system can help them achieve the learning objectives. |