Automated Tools to Support Early Detection via Retrospective Review

This project investigates the use of wireless, mobile sensors and computer vision to support the automatic capture, access, and retrospective review of children's play behaviors to support researchers with the identification of early indicators of autism spectrum disorder and other developmental delays. The Child'sPlay Smart Toys system uses wireless sensors embedded in toys to capture object-play data. The Video Filtering of Infant Social Games focuses on the automatic identification of social games, such as peak-a-boo, in video footage of infant/adult pairs.

Data collection Tools for special needs

This work is the result of a student project conducted over the Fall 2008 semester. This project focuses on the design of a system that eases the burden on instructors while maintaining accurate data collection for researchers in the education programs of children with developmental disabilities. The working prototype was developed and tested on a Tmobile G1 cellphone. We are currently working on another design concept.

Cognitive Modeling of Autism

This project involves the development of a cognitive model of autism centered around visual representations and processing, an account that could lead to new and improved paradigms for communication and education within the autism community.

Validation of Parent-Collected Observational Data in the Natural Environment for Use in Behavioral Interventions

We propose to study the feasibility of applying a human annotated video observation system to the problem of assessment and treatment of problem behavior exhibited by children with autism and related disabilities. Previous attempts to collect information on environment-behavior relationships for this population in the natural environment have been unsuccessful for a variety of reasons. It is our belief that human annotated video may resolve many of these factors that have been impediments in the past. Parents will be recruited for this study and asked to identify locations in the home where problem behavior most frequently occurs. Those locations will be equipped with a video camera that will record for a 12 hour period. In addition, parents will be provided with a device that allows them to indicate when a problem behavior has occurred and simultaneously annotate the video data. The goal of the study is to determine the validity of the parent collected data. This will be accomplished by comparing the instances of problem behavior they annotate to those scored by the trained observers.

Rapid ABC

Rapid ABC stands for Attend to their social environment, Establish Back-and-forth games, Engage others socially by Communicating using eye contact, facial expressions,vocalizations or words, and gestures.The goal of this project is to design a technological solutions that would facilitate the integration of an autism screening form into the medical office’s work flow and to engineer incentives that would provide value added data for the physician. Rapid ABC Screener, is a paper form screener used for screening Infants to indicate atypical social-emotional development that can be associated with autism spectrum disorders. The Screener is designed to supplement other screening instruments that assess language and motor skill development. Once this screening is complete manually by the physician, the screener can be converted to digital format and stored using our technological Rapid ABC Solution. This solution would generate data concerning the child’s performance compared to the normed values and if applicable, data from the child’s 18-month results. The staff can then give these to the physician and the parent. The data provided by solution could be used for research.


Towards Designing An Interactive and Intelligent Tool for Social Skill Development of Individuals with HFA
refl-ex is a system that can be used to train social situations for individuals with high-functioning autism. It is based on a constant loop of experience and reflection activities. The data gathered during one activity, determines the level for the next activity. In this way, the system adapts to the user's skill level and grows along as the user improves.


Baby Steps and KidCam

The Early Detection system is a project motivated by the Centers for Disease Control and Prevention's ActEarly campaign, which encourages new parents to look for the signs of autism as early as possible to ensure affected children get better treatment and intervention. We are working on a digital repository for new parents to keep track of various milestones in their child's development that are indicators of typical cognitive development (for example, babbling by age 6 months, or holding and shaking toys at age 3 months). We are looking at ways to automatically record developmental milestones and save video evidence to show to other caregivers. The system we are developing will also proactively ask parents to look for various milestones and provide warnings when a child has not yet achieved milestones by a specific age.

Abaris & Abaris UW

Abaris is a fully functioning prototype capture and access application to support therapists who perform Discrete Trial Training therapy, a current best practice intervention for children with autism. We have studied extensively the domain of discrete trial training through participant observation and interviews. We have developed a prototype using an Anoto digital pen and Nexidia voice indexing technology that allows for easy indexing of trials into a video session. The paper based form is very familiar to the paper ones the therapists previously used. The capture side of the interface allows for easier capture and less paperwork on a session by session basis. Abaris then provides an access interface for therapists and lead therapists to go back and review how the child is doing, look for inaccuracies, and easily show problem areas to other therapists for evaluation.


Capture and access technologies are particularly applicable to the monitoring, diagnosis, and intervention treatments of behavioral and learning disabilities in children. Behavior and learning data are pieces of information that can be captured, measured, mined and analyzed over time. Furthermore, the members of care teams are particularly motivated to do these activities, which may or may not be the case in traditional capture and access scenarios, such as meetings and classrooms. Capture and access applications created as a part of a cyclical system of diagnosis and treatment are also an interesting and special case wherein, the data captured in the past, once accessed and analyzed, affects the treatment plan and often the data to be captured in the future. CareLog is a mobile capture and access application for recording behavioral data in informal settings.

Wearable Sensors

We are in the early stages of exploring the use of sensors on a child with autism both during structured and unstructured activities. The goal is to run algorithms on this sensor data that can automatically detect certain physiological and (e.g. heart rate and galvanic behavioral responses (e.g. heart rate, galvanic skin response, and vocal/physical stimming). Automatically obtained indicators of this type can be used to guide therapists and researchers to salient information pertaining to the child. As a first step towards this goal, we have conducted preliminary experiments using a single neurotypical adult. The next step will be to validate the use of these same techniques with a CWA. Depending on our findings, we be able to link automatic sensor data analysis with Abaris to determine if certain physiological or behavioral occurrences have a correlation to the child's progress during therapy.


Walden Monitor is a combination wearable and Tablet PC based system that combines two existing paper-based data-collection instruments: the Child Behavior Observation System (CBOS) and the Pla-Chek (pronounced PLAY-check). CBOS and the Pla-Chek are used to record largely the same data in two different ways. The Pla-Chek is a paper spreadsheet used to record behavioral variables in the inclusive classrooms at the special school we studied. Each calendar quarter, research assistants enter the classroom for ten consecutive days and observe a particular child with autism. The research assistant mentally counts a ten-second interval, then records positive or negative results for twelve variables such as proximity to an adult (within 3 feet) or an adult interacting with the target child. The research assistant repeats this process twenty times. These data are also gathered using CBOS, in which a research assistant enters the classroom with a handheld video camera and records the child for five minutes. Another researcher watches the video and codes the variables on a spreadsheet similar to Pla-Chek. The teacher tabulates the data and includes it in written reports.

©2009 Georgia Tech • Georgia Tech Disclaimer