Research has shown that caregivers implementing
pivotal response treatment (PRT) with their child with autism
spectrum disorder (ASD) helps the child develop social and
communication skills. Evaluation of caregiver fidelity to PRT in
training programs and research studies relies on the evaluation
of video probes depicting the caregiver interacting with his
or her child. These video probes are reviewed by behavior
analysts and are dependent on manual processing to extract
data metrics. Using multimodal data processing techniques and
machine learning could alleviate the human cost of evaluating
the video probes by automating data analysis tasks.

Creating an ’Opportunity to Respond’ is one of the categories
used to evaluate caregiver fidelity to PRT implementation. A
caregiver is determined to have successfully demonstrated creating
an opportunity to respond when they have delivered an
appropriate instruction while she or he has the child’s attention.
Automatically determining when the caregiver has correctly
provided an opportunity to respond requires classifying the audio
and video data from the probes. Combining the modalities into a
single classification task can be undertaken using feature fusion
or decision fusion methods. Two decision fusion configurations,
and a feature fusion model were evaluated. The decision fusion
models achieved higher accuracy, however the feature fusion
model had a higher average