Self-Adaptive Vowel-Frame Detection Algorithm Based on Energy Distribution Analysis in Frequency Domain
Limin Xu · Dianzi xuebao · 2007
This paper presents a self-adaptive vowel-frame detection algorithm based on energy distribution analysis in frequency domain.The method maps MEL-scale into frequency component and detect vowel-frame by compare the energy of LF and HF.Then conduct the time-domain-analysis with short-time energy and zero-crossing rate of vowel-frame,which have been verified.The method needs less computation cost and can adapt the transformation of volume.The experimental results on single word have demonstrated that the accurate rate is over 97% while the average frame-lost rate is 3.95%.The accurate rate of vowel-frame detection on continuous speech is exceeding 90%.