The ICR (Intelligent Character Recognition), used to text handwritten and block letters recognition, can be considered an evolution and differentiation of OCR.

The moment when the OCR systems have been extended to the recognition of handwritten data in block letters, was used especially to the technology of neural networks: to emphasize that this new technology was based on artificial intelligence was then coined the new acronym, replacing the "O" Optical with the "I" Intelligent.

Contrary to what might imagine, most of the systems ICR is able to read only part of the text printed or typed, so if you have need to extract data to be printed manuscripts that you need to use data capture systems which are also available with OCR .

The ICR system can be distinguished in two categories: constrained ed unconstrained. The first are those who fail to recognize handwritten text in block letter in which the characters are completely separated from each other, while the latter are those that can tolerate some characters by touching them.

It is clear that the great variability and poor repeatability of the handwriting makes very complicated the work of these systems.

Usually, in order to enhance the performance of ICR systems, we use to an optimized design of forms, so that those who write the data is driven to write individual characters loose, in boxes printed appropriately colored that are already filtered during the scanning of the paper.

The use of pre-printed boxes avoids the scribe characters stick together, while their print with special colors makes possible the removal being scanned, improving the recognition.

A very important thing to keep in mind is that the ICR systems, being trained on real prototypes of writing, are very sensitive to the style of writing in the sample used, so that systems trained with samples of american characters, they work well when you need to extract data written from the american people can have errors while reading data written by the european people. To understand the motivation just compare a typical sequence of numbers written by an italian person with the one written by an american: will notice immediately that a "7" american is much more similar to a "1" italian, as well as a "4" american is easily confused with a "9" italian.

The differences in writing styles between Italian and Anglo-Saxon are significant: only a system developed ad hoc manages to not confuse, for example, a 1 with a 7, a 4 with a 9, etc.