- Slides: 16
Mobility Aid For The Blind By Amir Gonnen
The oldest aids l Walking Cane l Guide dog l Problems: – Skills and Training phase – Range – Very little information conveyed
A Technological Solution Sensor Information about the environment Computer Digital Processing Output Device Perceptual input perceived by the blind This is where I’m going to focus
Recent Developments l Soundscapes from “The v. OICe” l “Stereo vision system for the Visually Impaired”, University of Guelph, Canada l The “Nav. Belt” l The “Guide. Cane”
“The v. OICe”
The “Nav. Belt”
The “Guide. Cane”
Problems l Too much information l Missing sub-conscious vision processes – Interpretation of a continuous 2 D signal into a set of recognizable objects in a 3 D space – Movement velocity and direction
Solution l Emulating some of the subconscious processes in a computer l Describe the world as a set of objects in a 3 D space
Intro to the Simulation
Do We Get A Better Perception? l The Simulation: – A 3 D room with several objects – Each object has its own sound l The sound is located at the object – Stereo sound the direction of the object – Volume How far is the object l The beat scanner – divides the field view into sectors – More accurate perceptual direction
More About the Simulation l The simulation solves several problems: – How to separate sounds of different objects, so that not all objects will be heard at once ? – How to give a good indication of object’s type, direction and distance at the same time ? l Tradeoff between Resolution in Time and in Space.
Experiment l Let the subject close her eyes and guess which object is visible, its direction and its distance. l Results: – Recognition was correct in 80%-90% of the cases. – Notable improvement over time.
Conclusion l Still a lot of work to be done – Computer vision: how do we recognize objects on 3 D space using a raw 2 D signal ? – How do we assign sounds to different objects ? – How much information can be perceived using sound in a given time frame ? l Real progress could be achieved when we understand able to emulate the cognitive processes of human vision