Automatic Detection of Nasal Leak in Cleft Palate Speech Based on an Improved Group Delay Method
Xinyi Liu, Weiwei Tian, Heng Yin, Ling Song He · 2018
Cleft palate is a kind of congenital defect, and nasal leak is one of common symptoms in cleft palate speech.The automatic detection of nasal leak in cleft palate speech can provide a certain auxiliary basis for clinical diagnosis.Aiming at the characteristics of nasal leak in cleft palate speech, an improved method based on group delay is proposed.Since group delay parameter can greatly characterize nonlinear signals.Gammatone filter combined with the characteristics of human auditory is used to demodulate speech signal, and group delay parameters are calculated for each frequency band.It is used as features to identify nasal leak in cleft palate speech.The proposed algorithm is tested on 351 speech samples provided by West China Hospital of Stomatology Sichuan University.Compared with the classical group delay method, the accuracy of this improved method is 91.74% for the detection of nasal leak in cleft palate speech.