In this photo-essay, we will give a description of how the data is recorded and saved for processing.
The PowerGlove is connected to the SGI. A program, called gloveread, outputs the frames received by the glove in a standard format. The output of this is piped to a program called gloverecord. Gloverecord then asks the user to sign a particular sign.
In this case, we'll be looking at the sign for "same". The gloss for this sign, from the Auslan dictionary, is shown below.
The user begins by pressing the "Center" button of the glove. Gloverecord starts capturing the information from the glove at this point. It saves all frames between this and the end in a separate file.
The user then continues to make the sign.
The user then presses the "A" button on the glove to indicate that the signs is complete:
If the user is for some reason unsatisfied with the last sign, he can press "B" on the glove and he can start the sign again.
Obviously, in future, our goal is automatic segmentation of the signs. However, research in other areas, such as speech recognition and artificial intelligence in general, suggests that this is a very difficult problem.