Mumbai University > Computer Engineering > Sem 7 > Soft Computing

Marks: 10 Marks

Year: May 2016

Handwriting recognition is undoubtedly one of the most challenging areas of pattern recognition. It is extremely useful in a wide range of real world practical problems, including documentation analysis, mailing address interpretation, bank check processing, signature verification, document verification and many others [1].Several pattern recognition approaches have been applied to both on- line and off-line handwriting recognition, including statistical methods, structural and syntactic methods, and neural networks. Some reading systems identify strokes; others try to identify Characters, groups of characters, or entire words .

Neural networks are composed of simple elements operating in parallel. These elements are inspired by biological nervous systems. As in nature, the network Function is determined largely by the connections between elements. We can train a neural network to perform a particular function by adjusting the values Of the connections (weights) between elements. Commonly neural networks are adjusted, or trained, so that a particular input leads to a specific target output. Such a situation is shown below. There, the Network is adjusted, based on a comparison of the output and the target, until the network output matches the target. Typically many such input/target pairs are used, in this supervised learning, to train a network.

Computerized document processing has been growing rapidly since the 1980’s because of the exponentially increasing amount of daily received documents and the more powerful and affordable computer systems. Intuitively, the conversion of textual blocks into ASCII codes represents one of the most important tasks in document processing [5]. Our strategy of reclassifying characters is to incorporate typographical structure analysis which categorizes characters in the first step, and therefore it reduces the scope of character Recognition. Automatic Postal sorting, automatic bank cheque processing are application of Character recognition. In the work on character recognition has been reviewed. Optical Character Recognition (OCR) is used to recognize printed and handwritten characters. There are numerous approaches that address the problem and they vary in the features extracted from the graphical representation of the Characters.

1. Image Preprocessing:

The first phase in our character recognition process is converting the image to Binary image by thresholding the given character image. Two intensity values are available in binary image. These values are Black and White. We are use zero for Black and one for white. Thus the color of the character is White and the background is black.

Pre-processing covers all those functions of feature extraction to produce a original image. the feature extraction components of the character recognition used directly.

The steps in pre-processing involves

• Size normalization: Bicubic interpolation is used for standard sized image.

• Binarization: it is process of converting a gray scale image into binary image by thresholding

• Smoothing: the erosion and dilation smooth the Boundaries of objects.

• Edge detection: morphological gradient operators are used in edge detection because they enhance intensity of edges of characters.

2. SEGMENTATION

After scanning the document, the document image is subjected to pre-processing for background noise elimination, and binarization to generate the bit map image of the text. The pre-processed image is divided into lines, words and characters Line Segmentation ,Word Segmentation. Character Recognition

• Line segmentation-To separate the text lines, line segmentation is used.
• Word segmentation- Word segmentation is provide the space between words.
• Character Recognition-It is providing the Spacing between the characters. so it is called segmentation

3. FEED FORWARD NETWORK

A single-layer network of S logsig neurons having R inputs is shown below in full detail on the left and with a layer diagram on the right. Feed forward networks often have one or more hidden layers of sigmoid neurons followed by an output layer of linear neurons. Multiple layers of neurons with nonlinear transfer functions allow the network to learn nonlinear and linear relationships between input and output vectors. The linear output layer lets the network produce values outside the range –1 to +1.On the other hand, if you want to constrain the outputs of a network (such as between 0 and 1), then the output layer should use a sigmoid transfer function (such as logsig).

4. IMAGE PRE-PROCESSING

The next stage is image pre-processing module. Image pre-processing relates to the preparation of an image for later analysis and use. Images captured by a camera or a similar technique are not necessarily in a form that can be used by image analysis routines. Some may need improvement to reduce noise; other may need to simplified, enhanced, altered, segmented, filtered, etc[4]. The first step in the pre- processing block is to transform the color image into a gray scale image and this result to noisy gray scale image. In the next step, filtering is used in order to cancel the presented noise. Then, edge detection algorithm is applied for obtaining edge of the noiseless gray scale image. Image pre-processing module is consist of following operations-

i. Gray scale image

ii. Noise removal

iii. Edge detection

iv. Feature Extraction