Intellegent Word Recognition Techniques: A Servay

Resumen

The field of pattern recognition known as word recognition (WR) has undergone extensive
research and development throughout the years. To access and edit this data, display it online,
use it for text to speech translation, etc., word recognition is used to design reading machines
for offline viewing of written by hand and computer printed characters from an image
captured by camera, or Portable Document Format (PDF), many other application depend on
the word recognition. In this paper, the methodes of text recognition based on machin
learning and deep learning investigated, the advantage and disadvantage of each methodes
presented, the hybride methodes for text recognition also presented. Additionally, various
datasets type that are often cited in the literature are reviewed and summarized in this survey.
The performances of the state of the art techniques are demonstrated and explored using these
datasets. Many factores are shown to be effective in performance mesurmant of the text
recognition system such as the dataset content and the model used effected directly to the
accurcy of the system. The most accurent methodes for recognition shown to be the mehtodes
based on convolutional neural network (CNN). For the models depend on the CNN
technology the dataset size must be high. Two type of the recogniton based on the type of
dataset also presented. The dataset contain prented text and handwitten text. It hase been
shown that there is no suffitiont and comperhensive dataset based on printed text. Finally, the
conclution of this servay presented and some suggestion about the future research
possibilities for text detection and recognition also presented.