8. Automatic assessment of consonant omission and speech intelligibility in cleft palate speech
-
Abstract
The effective assessment of cleft palate speech has a great significance in clinical practice. Two algorithms are proposed to automatically detect the consonant omission and assess the speech intelligibility in cleft palate speech. The cleft palate speech database contains 530 participants fromthe Hospital of Stomatology, Sichuan University. The vocabulary of speech database includes all initial consonants and the most widely used vowels in Mandarin. All the speech recordings are assessed and annotated by professional speech-language pathologists. Based on the differences between vowels and initial consonants in Mandarin, this work combines the short-time autocorrelation function and the hierarchical clustering model to realize the automatic detection of consonant omission. The average detection accuracy is 82.75%. Based on the automatic continuous speech-recognition algorithm, the evaluation of speech intelligibility is completed. The recognition accuracy of automatic speech-recognition systemis proportional to the speech intelligibility. And the hypernasality grades and speech intelligibility are in inverse proportion. For the normal speech, the recognition accuracy of automatic speech recognition system is 96.41%. With the increase of hypernasality grades, the accuracy of automatic speech-recognition system reduces.
Abstract
The effective assessment of cleft palate speech has a great significance in clinical practice. Two algorithms are proposed to automatically detect the consonant omission and assess the speech intelligibility in cleft palate speech. The cleft palate speech database contains 530 participants fromthe Hospital of Stomatology, Sichuan University. The vocabulary of speech database includes all initial consonants and the most widely used vowels in Mandarin. All the speech recordings are assessed and annotated by professional speech-language pathologists. Based on the differences between vowels and initial consonants in Mandarin, this work combines the short-time autocorrelation function and the hierarchical clustering model to realize the automatic detection of consonant omission. The average detection accuracy is 82.75%. Based on the automatic continuous speech-recognition algorithm, the evaluation of speech intelligibility is completed. The recognition accuracy of automatic speech-recognition systemis proportional to the speech intelligibility. And the hypernasality grades and speech intelligibility are in inverse proportion. For the normal speech, the recognition accuracy of automatic speech recognition system is 96.41%. With the increase of hypernasality grades, the accuracy of automatic speech-recognition system reduces.
Chapters in this book
- Frontmatter I
- Foreword V
- Acknowledgments IX
- Contents XI
- List of Contributors XIII
- Introduction 1
-
Part I: Applying New Developments in Speech Signal Processing in the Diagnosis and Treatment of Speech and Communication Disorders and in the Detection of Risk to Neuro-Motor Functioning
- 1. Advancements in whispered speech detection for interactive/speech systems 9
- 2. Selective pole defocussing for detection of hypernasality 33
- 3. Acoustic-based tools and scripts for the automatic analysis of speech in clinical and non-clinical settings 69
- 4. Analysis of normal and pathological voices by novel chaotic titration method 87
-
Part II: Using Acoustic Modeling in the Detection and Treatment of Cognitive, Affective, and Developmental Disorders
- 5. Speech disorders in children and adults with mild-to-moderate intellectual disability 123
- 6. Autism and speech, language, and emotion – a survey 139
- 7. Clinical applications of speech technology for Autism Spectrum Disorder 161
-
Part III: Assessing and Quantifying Speech Intelligibility in Patients with Congenital Anatomical Defects, Disabling Conditions, and Degenerative Diseases
- 8. Automatic assessment of consonant omission and speech intelligibility in cleft palate speech 183
- 9. Distinctive auditory-based cues and rhythm metrics to assess the severity level of dysarthria 205
- 10. Quantification system of Parkinson’s disease 227
Chapters in this book
- Frontmatter I
- Foreword V
- Acknowledgments IX
- Contents XI
- List of Contributors XIII
- Introduction 1
-
Part I: Applying New Developments in Speech Signal Processing in the Diagnosis and Treatment of Speech and Communication Disorders and in the Detection of Risk to Neuro-Motor Functioning
- 1. Advancements in whispered speech detection for interactive/speech systems 9
- 2. Selective pole defocussing for detection of hypernasality 33
- 3. Acoustic-based tools and scripts for the automatic analysis of speech in clinical and non-clinical settings 69
- 4. Analysis of normal and pathological voices by novel chaotic titration method 87
-
Part II: Using Acoustic Modeling in the Detection and Treatment of Cognitive, Affective, and Developmental Disorders
- 5. Speech disorders in children and adults with mild-to-moderate intellectual disability 123
- 6. Autism and speech, language, and emotion – a survey 139
- 7. Clinical applications of speech technology for Autism Spectrum Disorder 161
-
Part III: Assessing and Quantifying Speech Intelligibility in Patients with Congenital Anatomical Defects, Disabling Conditions, and Degenerative Diseases
- 8. Automatic assessment of consonant omission and speech intelligibility in cleft palate speech 183
- 9. Distinctive auditory-based cues and rhythm metrics to assess the severity level of dysarthria 205
- 10. Quantification system of Parkinson’s disease 227
