|
This is a
news report about the annual International Meeting of Autism Research (IMFAR) in Salt Lake City, 2015. Our projects get some
attention.
With the rapid increase in the prevalence of autism spectrum
disorder (ASD) in 1990s, there are approximately 50,000
individuals with ASD turning 18 years old every year. The
community-based employment rate, even for those individuals
with higher functioning capabilities, is very low. This
can be partly explained by socio-communicative skill deficits
which are a hallmark of ASD, such as poor eye-contact
and inappropriately modulated speech. There has been little
work in developing assistive technology to help individuals
with ASD to compensate for these deficits. LittleHelper is a new system based on a wearable augmentedreality
glass platform to provide customized supports for individuals with ASD in enhancing social
communication during a job interview. Using the built-in
camera and microphone, LittleHelper can detect the position
of the interviewer relative to the center of the camera
view, and measure the sound level of the user. Based on
these inputs, appropriate visual feedbacks are provided back
to the user through the optical head-mounted display.
Littlehelper takes advantage of the small display and mini camera of Google glass to help individuals
with autism during the job interview. The camera can find where the interviewer is and the display can
tell the interviewee with ASD where he/she should look at. I did the user study and also
implemented the first version of Littlehelper, which is a Google glass app.
LittleHelper takes data both from the built-in camera and the microphone. We estimate
an ambient noise floor level during the initial training phase when the user is not speaking. To reduce
computing load in volume estimation, we set the sample rate to 8000 Hz and sample
width to 16 bits, the smallest supported by the Android operating system. Using a window size of 1 second, we calculate
the root mean square (RMS) of the audio signal.
To detect the location of the interviewer’s face, we rely on
the Viola-Jones face detector from the open-source
OPENCV library adapted for Google Glass platform. The detected face size (in pixel) is also used as an estimation of the
distance between the interviewer and the interviewee. Because different distance means different appropriate speaking
volume (calculated as a RMS value).
|