A SVOR based method for automatic scoring of prosody quality in Mandarin speech

摘要

Automatic Scoring of prosody quality in Mandarin is an essential part of computer-aided language learning system. In the paper, a method is investigated to automatically score prosody quality in Mandarin speech. The method is based on prosody comparison between the tested and standard sentence. The prosodic similarities are calculated from three aspects: tone, intonation and rhythm. Based on these similarities, a ranking algorithm named SVOR is proposed to predict the prosody quality score. Compared to SVM algorithm, the algorithm shows the better performance. It shows that human scoring of prosody quality is of ranking. In addition, detailed analyses of human scoring at the sentence-level are given.