Authors:
João Freitas
1
;
António Teixeira
2
and
Miguel Sales Dias
3
Affiliations:
1
Microsoft Language Development Center and University of Aveiro, Portugal
;
2
University of Aveiro, Portugal
;
3
Microsoft Language Development Center and ISCTE-Lisbon University Institute, Portugal
Keyword(s):
Nasality Detection for Silent Speech Interaction, Velum Movement Detection, Ultrasonic Doppler, Nasal Vowels, Portuguese.
Related
Ontology
Subjects/Areas/Topics:
Applications
;
Assistive Technologies
;
Biosignal Acquisition, Analysis and Processing
;
Health Engineering and Technology Applications
;
Human-Computer Interaction
;
Methodologies and Methods
;
Neural Rehabilitation
;
Neurotechnology, Electronics and Informatics
;
Pattern Recognition
;
Physiological Computing Systems
;
Software Engineering
Abstract:
This paper describes an exploratory analysis on the usefulness of the information made available from Ultrasonic Doppler signal data collected from a single speaker, to detect velum movement associated to European Portuguese nasal vowels. This is directly related to the unsolved problem of detecting nasality in silent speech interfaces. The applied procedure uses Real-Time Magnetic Resonance Imaging (RT-MRI), collected from the same speaker providing a method to interpret the reflected ultrasonic data. By ensuring compatible scenario conditions and proper time alignment between the Ultrasonic Doppler signal data and the RT-MRI data, we are able to accurately estimate the time when the velum moves and the type of movement under a nasal vowel occurrence. The combination of these two sources revealed a moderate relation between the average energy of frequency bands around the carrier, indicating a probable presence of velum information in the Ultrasonic Doppler signal.