Novel Application to Improve Communication for Children Affected by Autism Spectrum Disorder

Abstract

Autism Spectrum Disorder (ASD) is a spectrum of disorders that affects a child’s communication, social, and emotional skills. ASD is a major challenge for children because they are not able to communicate effectively with others, especially in terms of conveying their emotions. For the 40% of ASD children who are mute and are not able to verbalize their emotions, facial expressions are the primary indicator that family members and caregivers use to recognize their emotion. ASD children display facial features unique to each child, so those who are unfamiliar with the child such as teachers may find it difficult to interpret the child’s emotions. This creates a communication barrier between ASD children and the outside world, leading to frustration and isolation among the 13.7 million ASD children around the world. Current solutions to help ASD children socialize, such as speech practising or assisted learning apps, do not reduce this barrier. This is because these apps are not an immediate solution to this barrier, and can be effective only after months of practice by the child. Also, most of these solutions do not work for mute ASD children. Thus, there is a dire need for an individualized solution that interprets an ASD child’s emotion. My solution is the Cognitive Emotion Interpretation App (CEIA). CEIA uses Artificial Intelligence and Emotion Recognition Technology to map an ASD child’s facial expressions with an emotion. Through CEIA, people who are not familiar with the ASD child (teachers, extended family) can interpret the child’s emotion. When a user (parent, caregiver) downloads the app, they upload photos of the ASD child expressing different emotions and tag the picture with the emotion (e.g. Happy, Sad, Frustrated, Hungry). CEIA then extracts the child’s facial features, and the AI algorithm is trained to associate the picture with the emotion. When the user wants to interpret the child’s emotion, they take a photo of the child exhibiting the emotion and upload it to CEIA. The AI algorithm will evaluate the photo, and list the emotions that match with the highest accuracy. The user can also upload more photos at a later stage, and the AI algorithm will be retrained to take these new photos into its training dataset. A higher number of photos used in training generally yields a higher recognition accuracy, thus users are encouraged to upload many photos of the child’s emotions. The performance of the app will be evaluated on the following metrics: 1) accuracy of the emotion recognizer, 2) amount of time CEIA takes to recognize the emotion, and 3) CEIA’s ease of use. Accuracy will be measured by collecting a sample of a variety of emotions of different users, then measuring if CEIA correctly matched the emotion in the photo. This initial test of accuracy will provide a representative sample of the types of emotions CEIA will need to train on. CEIA will provide a much needed powerful tool to reduce the communication barrier between ASD children and their community.