Automatic Evaluation of Hypernasality and Consonant Misarticulation in Cleft Palate Speech

Abstract

The automatic evaluation of CP (Cleft Plate) speech has various clinical applications. In this work, automatic classification of hypernasality levels and detection of consonant omission methods are proposed. Considering that the data collection is a major bottleneck in the field of CP speech signal processing, an extensive CP speech database is used. This database is collected over 10 years by the Hospital of Stomatology, Sichuan University, which has the largest number of CLP (Cleft Lip and Palate) patients in China. The vocabulary used for this database covers all the initial consonants and the most widely used vowels in Mandarin. Based on the production process of hypernasality, a two-step classification algorithm is proposed to identify four levels of hypernasality: normal, mild, moderate and severe. In order to locate consonant omission, an effective classification algorithm is proposed through analyzing the acoustic characteristics of initial consonants and finals. The classification accuracy for four hypernasality levels reaches up to 83% and the correct identification of consonant omission is over 94%.Index Terms-Cleft palate speech, consonant misarticulation, consonant omission, hypernasality.

Estimation of Hypernasality Scores from Cleft Lip and Palate Speech

“…The presence of extra-nasal formants around 250 Hz and 1000 Hz in vowel spectrum, increase in the first formant bandwidth, reduction in second formant strength, and increase in the spectral flatness are considered as the important acoustic cues of hypernasality [5,6,7]. Melfrequency cepstral coefficients (MFCCs), glottal source related features (jitter and shimmer), wavelet transform based features, Gaussian mixture model (GMM) and support vector based classifiers have been explored for the hypernasality detection from word and sentence level data [8,9,10]. Also, automatic classification of speech into normal, mild, moderate, and severe levels of hypernasality is proposed in Ref.…”

“…Also, automatic classification of speech into normal, mild, moderate, and severe levels of hypernasality is proposed in Ref. [8,9], where GMMs are explicitly trained for these classes. However, these methods have limitation to use for the clinical applications.…”

Pitch-Adaptive Front-end Feature for Hypernasality Detection

“…The presence of nasal peak in low-frequency region around the first formant F1, reduction in strength of F1 and hence broadening of F1 are some important spectral cues proposed for nasalized vowels [7], [8] which are used by the researchers for the hypernasality detection. The important works reported in the literature for hypernasality detection are based on Teager energy operator [9], Teager energy operator plus Mel frequency cepstral coefficient (MFCC) [10], linear prediction cepstral coefficient (LPCC) [11], high spectral resolution group delay spectrum [4], set of features based on acoustic, noise and cepstral analysis, nonlinear dynamic and entropy measurements [12], [13], [14], energy distribution [15], [16] and zero time windowing [17], [18].…”

Automatic Evaluation of Soft Articulatory Contact for Stuttering Treatment

“…There have been many studies on the automatic measurement of characteristics of consonants, such as voice onset time (VOT) [10]- [12], or the use of acoustic features of consonants to detect or assess disordered speech [13]- [16]. Regarding stuttering, methods for detecting repetition and/or prolongation have been proposed [17]- [19].…”