The Alpha-Digit object is exactly like the generic object except that the recognizer is used only to recognize Letters and digits and the object does not provide branching. An example of an application that might benefit from enhanced letter and digit recognition is a catalogue ordering system where the user may select a item number such as #ID570. The recognized digits are stored in the array variable '$stateX(recog)' where stateX is the name of the digit object. (See Tutorial 4 or a Tcl manual to learn about variables and see Tutorial 21 for a demonstration of the Alpha-Digit Object)

Double click the red port of the Alpha-Digit object to open the Alpha-Digit recognition window. Recognition vocabulary can be loaded in two ways: from a file, or entered directly in the vocabulary window.


To create a vocabulary file, open a text editor and enter the letter and digit items you wish to recognize on separate lines. Save your file. From the recognition window, select the "Use vocabulary file" checkbox. Select "Browse" to find and load the vocabulary file you just created.

For direct vocabulary entry, select the "Vocabulary" checkbox and enter the vocabulary letter and digit items you wish to recognize on separate lines. Optionally you may select "Compile" to speed up recognition at run time. Doing so will require you to name a file where RAD will save a processed version of the vocabulary you specified.

There are two tunable parameters for the alphabet recognizer:

Rejection Threshold: A low setting will reject more, a high setting will reject less.

Pruning: Controls the breadth of the Viterbi search recognition algorithm. A small number makes the recognition faster by narrowing the search tree. However a small number can decrease recognition accuracy by pruning too quickly. A large number increases the breadth of the search resulting in slower recognition. The size of the recognition vocabulary also effects the recognition speed.