Davis and Shah's work on Gesture Recognition using cameras

The research undertaken by Davis and Shah employed a simple black and white camera, and a glove which is black, except for white fingertips. A four-state finite state machine was used to recognise a simple set of seven gestures that would be used in interaction with computers -- like left, right, up, down, rotate, grab, stop. The four states of the finite state machine were:

1. Keep hand still in start position.

2. Move hand position slowly to gesture position.

3. Keep hand in position for desired duration (ie how far left you want to go).

4. Move fingers back to start position.

As can be seen, this is a very simple system, and it works quite well. However, it has some significant shortcomings for use in full sign language -- such as it does not monitor hand position, and requires explicit stop/start signals, and only operates at 4 Hz. Furthermore, the ``finite state machine'' approach is very limited and is really a ``posture'' recognition system rather than a ``gesture'' recognition system. The finite state machine has, in some of the experiments, gone prematurely into the wrong state, and in such situations, it is difficult to get it back into a correct state.