Tim Bunnell Center for Pediatric Auditory & Speech Sciences - PowerPoint PPT Presentation

Tim Bunnell Center for Pediatric Auditory & Speech Sciences Nemours/Alfred I. duPont Hospital for Children Wilmington, DE

Nemours Children’s Health System  Clinical – Hospitals in Wilmington & Orlando – Clinics & satellites in DE, FL, PA, NJ, MD, GA – Over 700 physicians – Over 30 pediatric specialties – Over 1.2 million encounters per year – Around 300,000 unique patients seen per year  Research – Almost $9M in NIH funding for 2014 – 11 Research Centers – ~40 Laboratories – Focus on Neurodevelopmental & Musculoskeletal disease, Diabetes & Obesity, Asthma & Cystic Fibrosis, Cancer, Applied Genomics, and Healthcare Delivery Science.

CPASS  Four Labs + Bioinformatics – Auditory Physiology & Psycho-acoustics – Head, Morlet – Balance & Vestibular Disorders – Head, O’Reilly – Craniofacial Outcomes Research – Head, Vallino – Speech Research Lab – Head, Bunnell  Acoustic Phonetics  Speech Perception/Production  Clinical Speech Technology – Applications involving speech recognition & synthesis technology – Bioinformatics – Head, Bunnell

Clinical Speech Technology Cluster Dendrogram  Utterance Verification / Human vs Machine CSIM Scoring 0.7 Classification 0.6 1.0 0.5 Hearing Status – Acoustic Phenotyping NH * 0.4 CI * – Used as functional hearing Height 0.3 0.8 nor 0.2 evaluation Human − Based CSIM Score 0.1 – Auditory/verbal therapy 0.0 0.6 s08 s14 s07 s13 s05 s15 s03 s16 s01 s10 s06 s12 s02 s11 s04 s09 s17 s18 – Used as objective speech intelligibility measure 0.4 Subject 0.2 Age 36 42 48 0.0 72 0.0 0.2 0.4 0.6 0.8 1.0 ASR−based CSIM Score

Speech Synthesis for Assistive Technology  Problem – Speech Generating devices are: – Limited in choice of voices – Impersonal – Lacking expressiveness  Existing Solutions – Voice banking – Voice conversion/creation – Parametric synthesis to modulate prosodic/expressive features

ModelTalker TTS System XML Control Speech File Files MTVC Feature Extraction This is a demonstration of the ModelTalker HMM Training Speech Synthesis System. Data Pruning Child2 Child Adult2 DB Construction Dictionary Unit Select. & Database Linguistic Rules

Going forward for personal voices…  User Needs – More Expressive! Lacking in Unit Selection without massive amounts of data  Not modeled well in statistical parametric synthesis  – More Natural Issue particularly for parametric synthesis  Fewer ‘glitches’ in unit selection  – Lower Barriers to creation Fewer hours of recording  Improved morphing  Research Needs  – Phonetics/Phonology Improve acoustic models of emotion and expressiveness   Improve models of the time-varying structure of speech (!) – Engineering Improvements in signal processing to model voice/vocal-tract interaction  – Capture the features of an individual’s speech in a small number of dimensions than can be manipulated in expressively useful ways.