14-09-2017, 09:27 AM
There have been many attempts to manually identify scripts in offline documents. The most important feature of online document recognition is that they capture the temporal sequence of strokes while the document is being written. This allows us to analyze the individual strokes and use the additional temporary information for the identification of the writing.
The proposed system uses the characteristics of the connected components to classify six different scripts (Arabic, Han, Cyrillic, Devnagari, Hebrew and Roman) and reported 88 percent accuracy on the pages of the document. There are some important aspects of online documents that allow us to process them in a fundamentally different way to offline documents.
The most important feature of online documents is that they capture the temporal sequence of strokes while the document is being written. This allows us to analyze the individual strokes and use the additional time information for both handwriting identification and text recognition.
The system collects data first in the six languages mentioned. The script is created in these languages and stored in the file. Based on this collection, words in a particular language can be detected with the properties of each language. The classifier design uses the k-Nearest Neighbour method.
The proposed system uses the characteristics of the connected components to classify six different scripts (Arabic, Han, Cyrillic, Devnagari, Hebrew and Roman) and reported 88 percent accuracy on the pages of the document. There are some important aspects of online documents that allow us to process them in a fundamentally different way to offline documents.
The most important feature of online documents is that they capture the temporal sequence of strokes while the document is being written. This allows us to analyze the individual strokes and use the additional time information for both handwriting identification and text recognition.
The system collects data first in the six languages mentioned. The script is created in these languages and stored in the file. Based on this collection, words in a particular language can be detected with the properties of each language. The classifier design uses the k-Nearest Neighbour method.