Writer Identification Using Graphemes

Handwriting is one of the behavioural biometric techniques for person identification. Writer identification is a field of study that focuses on determining the authorship of a given text. This can be done through various techniques and methods, often involving linguistic analysis, machine learning, and computational approaches. Here are some key aspects and methods related to writer identification. Advanced machine learning algorithms are commonly used in writer identification. These algorithms can be trained on a dataset of known authors to learn patterns and characteristics specific to each writer. Support vector machines, decision trees, and neural networks are often employed. The present work is based on handwritten structural primitive called graphemes. The dataset used is IAM offline English database. First of all images of handwritings are segmented into graphemes and then graphemes are represented as projection profile, i.e. each black pixel positions of the obtained segmentedcurve is recorded. Then dictionary is learnt from train-set images of handwriting using k-means clustering. Now using dictionary, writer feature vectors are generated and stored with writer label.