MMI-Man Machine Interface

It is known for years that stress, press, fear, happiness and many other feelings, can be measured by several techniques .

Abstract
It is known for years that stress, press, fear, happiness and many other feelings, can be measured by several techniques . Based on this knowledge, we would like to build an interface, that will enable people to communicate without using any muscle, and without making any physical effort.

Background
Our goal is to build a user-interface that will show a set of symbols to the subject and on the same time records a physiological signal from him.The interface should detect in Real-Time when the subject choose a symbol and show which symbol was chosen. This goal requires the subjects ability of changing the measured signal when he wish.Otherwise the interface will not detect a meaningful information.

The solution (or the basic approach)
The physiological signal that we measure is the DC potential across neurons of the autonomic nervous system.It is not simple to control this signal.There for we had build a Biofeedback train tool that will improve the subjects ability of controlling the signal.The tool gives feedback in means of audio and video to the subject according to the amplitude of the signal.

1
Figure 1 – The biofeedback tool. The height of the bar is function of the subject signal

The user-interface display a matrix of the alphabet. The rows in the matrix are highlited in random order.When the interface detect a significant change in the signal (which means that the subjects desired letter is placed in the highlighted row),the letters in this row start to highlighted randomly and the subject can choose the letter he wanted.The chosen letter appears on the screen and in this way the subject can write a whole word.

fi
Figure 2 -The user interface tool.The subject choose a line and letter from the matrix

Tools
In this project we used “pc-ecg 1200” – an E.C.G measure tool, and in order build our user interface and bio-feedback tools we have used Visual Basic.

Conclusions
The main target was to be able to write a whole word in the real-time implementation. We have not completely reached this target. In accuracy of 70% we are able to write one chosen letter. The real-time implementation functionality is not completed. We can point at some areas for improvement:
1. A much longer training session.
2. Try to use other Bio-feedback tools.
3. The algorithm should solve some problems that our algorithm did not solve, like reducing the low-frequency signal.

Acknowledgment
We are grateful to Yohanan Erez and to our project supervisor Dr. Daniel Langue for their help and guidance throughout this work.
We are also grateful to the Ollendorf Research Center Fund for supporting this project