AutiGlove: Understanding autistic children using Biosensors and Python!
What is Autism? Autism is defined as a neurobehavioral developmental disorder that shows deficits in communication skills and social behaviour of a child. ASD is a lifelong disorder which has no cure but diagnosing early and providing effective treatment shows positive outcomes in the later stages of life.
How assistive technology is transforming the lives of Autistic children? Assistive devices like Google Glass and Robot Nao are using latest technologies like ArtificiaI Intelligence and Internet of Things to help provide cognitive behavioural and operational therapy to an autistic child. These devices primarily use facial / speech Recognition to understand and converse with an autistic child. However, this method tends to have very less accuracy since children with ASD primarily have difficulty in expressing and communicating. Their behaviour is often repetitive and confusing which leads to a prolonged and less effective therapy.
Recognizing emotions using physiological signals of the body is relatively new and unexplored domain. Research studies have shown that there is a positive correlation between the physiological changes of human body and the emotional triggers occurring in the body. To corroborate this research finding, AutiGlove was developed.
What is AutiGlove?
The Hardware Part: AutiGlove is a cost-effective, non-invasive wearable glove built using biomedical sensors like Galvanic Skin Response (GSR) sensor, Pulse sensor and Temperature sensor. It captures the real-time physiological signals like Skin Conductance, Heart Rate Variability and Human Body Temperature when the participant is subjected to a video/picture stimuli. This triggers an emotional response and the readings captured are sent via a bluetooth module interfaced with a 3.3 V Arduino Pro Mini. The AutiGlove prototype was tested at an orphanage for autistic children in Bangalore which in iteself was a challenge. For a Data Scientist like myself, who works on readily available public datasets, this experiment at the orphanage helped to understand the importance of data acquisistion stage, the techniques involved and the challenges that challenges that need to be overcomed.
Role of Machine Learning: The dataset obtained was subjected to pre-processing. Various Biosignal Processing Python tools like Neurokit and BioSPPY were used for feature extraction. The features extracted (Key to the entire project. Shall be explained at the conference, if selected ) helped to predict three emotional states of an autistic child using supervised machine learning algorithms (K-NN, SVM, Decision Trees) for multi-class classification. The prototype aims to be an important therapeutic tool for parents and caretakers of autistic children.
Outline of Poster Presentation:
- Developing AutiGlove prototype and overcoming challenges during data acquisition
- Live Demo of AutiGlove Prototype
- Introduce various biosignal processing tools available in Python (Neurokit, BioSPPy) for feature extraction
- Mapping the Physiological Signals with human emotions using supervised Machine Learning Algorithms
- Basic knowledge of Python
- Basic understanding of micro-controllers and sensors