Images are first converted into binary format using Otsu thresholding algorithm. In writer identification, features do not correspond to a single value, but a probability distribution function (PDF) extracted from the handwriting images to characterize writer individuality. The following features have been considered in this study.
This feature makes it possible to discriminate between fast writers who create smooth handwriting and slow writers who create knotted handwriting. In the used dataset, for each pixel p in the text, we consider 20-dimensional features related to tortuosity. 10-dimensional Probability density function represents the length of the longest line segment which traverses p and completed within the text and 10-dimensional Probability density function denotes the direction of the largest line portion.
This feature can measure the tangential direction of central axis of text. Here it uses a Probability density function of 10 dimensions.
This attribute is usually accepted in forensic science examination which studies the curvature as discriminating feature. It uses a Probability density function of 100 dimensions which represent the values of curvature at the outline pixels.
4. Chain code:
Chain codes can be generated by scanning the outline of the text and assigning a number to each pixel according to its location with respect to the previous pixel. For each pixel, we can consider eight possible directions and consider the location with respect to the previous 1,2,3 and 4 pixels.
5. Edge direction:
Edge-based directional features give a detailed distribution of directions and can also be applied at several sizes by positioning a window centered at each contour pixel and counting the occurrences of each direction. This feature has been computed from size 1 (which PDF size is 4) to size 10 (which PDF size is 40).
We will use K-Nearest neighborhood, L1 Regularized Logistic Regression, Decision tree, Random Forests and many new algorithms to evaluate the image and predict the gender of the User.
For the above mentioned purpose, we will use all the above attributes and will generate some new attribute to enhance the efficiency of the existing system.