D2-A: AAC Technology to Supplement Intelligibility of Residual Speech
The goals of this project are to develop design specifications, develop prototype systems,
and evaluate technologies that facilitate the comprehension of supplemented speech.
For more information about this project, contact David Beukelman or Kevin Caves.
Progress to Date
The prototype has been developed. A project has been completed in which 7 speakers with very low intelligibility (15 to 20%) used to prototype to supplement their speech. With high levels of word prediction accuracy, the speech of these individuals was supplemented to be approximately 80-90% intelligible.
A second project evaluated the impact of alphabet supplementation (of various levels of completeness) on speech intelligibility. Specifically, four experimental conditions were utilized--all words supplemented, all word supplemented except nouns, all words supplemented except verbs, or all words supplemented except functor words. This study demonstrates that not all words in an utterance must be supplemented in order to achieve the supplementation effect. |
D2-B: Recognition of Dysarthric Speech
Automatic speech recognition (ASR) systems, such as ViaVoiceT, and DragonDictate ®
are being used more frequently by the general population and individuals with physical disabilities.
While there has been major improvement in ASR, commercially-available systems do not work well for
individuals with dysarthric speech. Our goal is to build a portable system with text and/or speech
output to recognize dysarthric speech.
For more information about this project, contact Kevin Caves . |
D2-C: Gesture Recognition
This activity builds on existing work conducted in the current AAC-RERC with NavAir,
Orlando , FL as part of CRADA established between Duke University and the NavAir in January
2001. Specifically, it seeks to determine if movements made by people with disabilities can
be used to operate AAC and AT. An individual can "show" the system a movement that the system
can then learn and that movement can be used as an input to an AAC or computer system. The
proposed system attempts to utilize a person's ideal movement patterns and use them to operate
technology, rather than attempt to fit the person to a piece of technology.
For more information about this project, contact Kevin Caves . |
D2-D: Brain Interface
The goal of this activity is to develop the interface module for an implanted brain
neuroprosthesis that can be used to access AAC/AT for individuals with "locked in" syndrome
secondary to stroke or neurological disorder.
Duke University 's Center for Neuroengineering is conducting a study in which
individuals with severe physical limitations will be implanted with commercially-available,
FDA-approved microelectrodes within the brain. The microelectrode signals will be directed to
a control computer which will process the neuronal signal into device control commands specific
for each device. The output from the implanted device will be analyzed to determine the maximum
"channels" of control.
The AAC-RERC intends to develop control signal processing for a variety of typical external
devices, and then establish the training format necessary to enhance the user's ability to use
these devices for functional purposes. For example, if a subject can control a single channel
and can control it above a threshold, the control computer can be trained to interpret this as
a switch closure for use with a commercially-available AAC system. If the subject can reliably
control the threshold for two channels, two-switch scanning or encoding is possible. If the
subject can reliably and consistently raise and lower a signal, a level of proportionality could
be achieved. For example, if a subject could raise and lower two discrete signals, these signals
could be used to control the X-Y position of a mouse pointer in a graphical user interface.
For more information about this project, contact Kevin Caves or Frank DeRuyter. |
