MIDS Capstone Project Summer 2017

iSeeiSpeak

The Problem

For a young child with Autism, the world is a lonely place.

Without a ready ability to communicate children can become frustrated leading to acting out and further problems.One of us has a son who after becoming a few months old was observed to be particularly quiet:

"Because I already had a child with special needs, I was very sensitive to my son and aggressively pursued the help of development pediatricians for starting early speech therapy. The speech therapy and especially the employment of pictorial symbols by my son's therapist made a significant impact on my son’s positive communication development skills. Drawing on this real-world experience, I came up with the inspiration to develop an intelligent image-based communication system for nonverbal children."

Our Vision

Design an intelligent image-based communication system for nonverbal children. 

The Solution

We have a four stage solution:

  1. Segment, parse English language sentences and apply a custom ASL (American Sign Language) based filter.
  2. Build a parallel corpus that translates between English sentences and Image-ese (the visual language).
  3. A deep learning neural network to learn based on the training data from the parallel corpus.
  4. Refining the text to image translator.
iSeeiSpeak, an intelligent image-based communication system for nonverbal children
iSeeiSpeak, an intelligent image-based communication system for nonverbal children

Last updated:

September 11, 2017