Application of Linear Discriminant Analysis to the Long-term Averaged Spectra of Simulated Disorders of Oral-Nasal Balance

作者：de Boer Gillian; Bressmann Tim^*

来源：Cleft Palate-Craniofacial Journal, 2016, 53(5): E163-E171.

DOI：10.1597/14-236

• acoustics
• hypernasality
• hyponasality
• mixed nasality
• spectrography

Objective: Acoustic studies of oral-nasal balance disorders to date have focused on hypernasality. However, in patients with cleft palate, nasal obstruction may also be present, so that hypernasality and hyponasality co-occur. In this study, normal speakers simulated different disorders of oral-nasal balance. Linear discriminant analysis was used to create a tentative diagnostic formula based on the long-term averaged spectra (LTAS) of the speech stimuli.
Materials and Methods: Eleven female participants were recorded while reading nonnasal and nasal speech stimuli. LTASs of the recordings were run for their normal oral-nasal balance and their simulations of hyponasal, hypernasal, and mixed oral-nasal balance. The amplitude values (in decibels) were extracted in 100-Hz intervals over a range of 4 kHz.
Results: A repeated-measures analysis of variance of the normalized amplitudes revealed a resonance condition-frequency band amplitude interaction effect(P < .001). A linear discriminant analysis of the participants' LTAS led to formulas correctly classifying 80.7% of the oral-nasal balance conditions.
Conclusion: The simulations produced distinctive spectra enabling the creation of formulas that predicted the oral-nasal balance above chance level. Future research with speakers with oral-nasal balance disorders will be needed to investigate the potential of this approach for the clinical diagnosis of disorders of oral-nasal balance.

• 出版日期2016-9