Russian Version

A Breakthrough in Artificial Vision!

New concept of artificial vision – Tactile Vision! Newly patented means of coding and transmitting the signal – Artificial Vision Device!

Methods of Training

We prefer not to go into details of the methods and processes involved in the training experiments. You can find them in the complete report. But it’s important to stress that the work was divided into two main blocks:

First: Searching for the best ways for processing and transmitting the information;

Second: Determining the full range of functional abilities of the device, and figuring out and perfecting methods of training.

The work was going in both directions simultaneously, and therefore decisions that would be made would influence the other block of the work. That might lower the pure scientific accuracy of the experiment. But, in any case, the satisfactory results of the experiments are as follows:

1.1 First we faced the same difficulties as our predecessors – to learn how to recognize contrasting objects and shadows (ability to see in 3-D). Out of all the available methods of preliminary processing of information, we found the most useful to be the division of each camera-shot into positive (p) and negative (n). The radical improvement of distinction and speeding-up the training process were reached under the sequence of four shots – 3 positive shots and 1 negative shot, conveyed as P-P-N-P. We consider it beneficial to select sharp angels of a still object (and with better software, moving objects as well).

1.2 We have found it the most comfortable and effective to use square electrical impulses.

1.3 If the frequency of shots coming from the camera was equal to 25 hertz, we had the ability to change the frequency of electrical signals on the tactile device. The best compromise between the suitable level of comfort and achieving good distinction of objects is using a frequency of 80 hertz.

2.1 In terms of our philosophy, as mentioned earlier, what is important is that a subject is able to experience the extent to which we are able to compensate for his/her informational deficit, connected with blindness, and to widen the functional abilities of the subject. The results of the experiments allow us to state that after a rather short period of training (30-40 hours), the subject is able to:

• successfully distinguish among several objects within the field of vision;

• successfully distinguish, find and use objects located within reach (for example, build a little house from toy bricks). The objects can have the same shape, but differ only by a picture on the surface (for example, different brands of cigarette packages), and the subject would still distinguish between them;

• navigate and move in a space not specifically designed for the blind. Also, carry out certain orders, like finding and moving an object (for example, to find and bring a glass, located near the window in another room). Of course, the completion of those tasks demands more time compared to the same actions being performed by someone with normal vision but, as we believe, the speed is not as significant as the accuracy in navigating in the new space;

• distinguish up to three printed symbols within the field of his/her vision, thus giving the subject the ability to read, using the method of scanning.

2.2 For the methods of training, the most crucial thing was the order of learning: first, the closest space (objects within reach of the hand); then, medium range (objects that can be reached by changing the position of the body); and, last, remote space (objects that can be reached by moving in space). This order is extremely important. The training is very similar to the development of an infant. As a baby starts to focus his eyes on toys above his bed that can be reached by hand, the subject at the beginning of training reaches by hand for an object, placed on the nearest stand, and learns to combine this idea with signals conveyed from the device.

We would make the task more difficult as the subject makes more accurate movements – we would move the object and ask the subject to stand up in order to reach it.

When we started experiments with the object placed on a stand at a distance, we would mark the floor (drawing a white line on the floor showing the way to the object and moving away all the obstacles from his/her path).

Moreover, as we were working with adults, we could explain the task and give instructions and directions in the process of fulfilling of the task, which certainly made training more effective.

When the subject would finish working in close, medium and distant space, he/she would start navigating in a difficult space, where obstacles would not be put away, the paths would not be marked on the floor, and the objects would be placed behind other objects, or hidden in shadows (closer to real world conditions).

Treating the device as something that can compensate for vision deficit, rather than being a “vision prosthetic appliance,” helped us reject our original idea of attaching the camera to the subject’s head. The experiments showed that if the subject holds the camera in his/her hand, the speed of fulfilling the tasks would increase sharply, thus shortening the period of training.

Due to our approach, we found it reasonable to look for additional ways to increase the effect of the visual deficit compensation, use other methods for orientation in space and prepare surrounding space and given information. We tried to use the device and a cane at the same time. The results proved that it can be a successful method both in training and adaptation of subjects who are used to a cane.

The results of the experiments with printed text showed that for maximum success it is necessary to create a special pattern of camera movement that would provide sequential scanning of the text by a camera with regulated speed.

We have not yet conducted open-air experiments but, for the subject’s full freedom of movement, some special arrangements might be needed (special posters, navigation marks and so on).

<< Back | Next >>

Contacts