Automatic Detection Algorithm of Pharyngeal Fricatives in Cleft Palate Speech Based on LPIF and Feature Selection

Abstract

Cleft palate is one of common congenital malformations that has huge impacts on the physical and psychological health of patients. Pharyngeal fricative in cleft palate speech is a kind of common compensatory articulations, which is produced by retracting tongue position to posterior pharyngeal wall and narrowing velopharyngeal opening. In this paper, based on voice mechanism of pharyngeal fricatives in cleft palate speech, linear prediction inverse filter was used to extract glottal waveform and estimate vocal tract model coefficients. Then four features were extracted, including pitch period, glottal flow derivative waveform, vocal tract area and vocal tract gain. The cross-correlation function was used to calculate the correlation between features. Glottal flow derivative waveform was removed since it had strong correlation with the others. KNN classifier was applied to realize the automatic pharyngeal fricatives detection in cleft palate speech, which reaches the detection accuracy of 98.40%.

Acoustic analysis and detection of pharyngeal fricative in cleft palate speech using correlation of signals in independent frequency bands and octave spectrum prominent peak

“…The highest detection accuracy of pharyngeal fricative speech in this work is 89%. Moreover, the specificity, sensitivity, and AUC values of CSIFs and OSPP features are larger than those obtained with EMAs [23], FMAs [24], ICPD [25], and VTG + VTA + PP [26].…”

“…In this subsection, the comparative experiments are conducted based on features proposed in previous researches [23][24][25][26] and this work. The researches [23,24] study the spectral characteristics of pharyngeal fricative speech based on one-third octave spectrum.…”

“…In research [25], He et al present the ICPD feature based on the variation of energy in speech signals caused by the changed place of articulation in pharyngeal fricative. Fu et al [26] extract the VTG + VTA + PP feature set from the glottal waveform and the coefficients of the vocal tract model. The capabilities of these features in detecting pharyngeal fricatives are tested using the database in this work, which contains six types of initial consonants.…”

“…In clinic, clinical doctors need more objective measurements to evaluate the speech quality of pharyngeal fricatives [21,22]. To overcome this limitation, related studies that have realized automatic detection of pharyngeal fricative have been conducted by Xiao et al [23,24], He et al [25], and Fu et al [26]. Xiao et al [23,24] propose features based on one-third octave spectrum, which can reflect the characteristics of the energy distribution of speech signal.…”

“…There are 221 speech samples (127 pharyngeal fricatives /s/, and 94 normal speech /s/) in their classification work. Fu et al [26] implement preliminary experiments based on speech signals collected from 10 patients and 4 control subjects. The imbalanced sample sizes of pharyngeal fricatives and normal speech could result in the overfitting of the model.…”