Gender Classification and Writer Identification System based on Handwriting in Gurumukhi Script

Gender Classification and Writer Identification system are the challenging applications of artificial intelligence and machine learning and widely helpful in forensic, criminal, and suspected investigations. The proposed system is based on behavioral biometric science. Physiological and behavioral biometric traits are the two traits of biometric modality. The paper proposed a novel move in direction of the Gurumukhi (Punjabi) script using multiple feature extraction techniques and hybridization of classification algorithms. The dataset for the experimental evaluation consists of 200 writers with 100 males and 100 females. Two feature extraction methods namely, Intersection and Open Endpoint based feature extraction method and Curve fitting-based feature extraction are considered in this work. For classification, various classifiers namely, Support Vector Machine (SVM), Multi-Layered Perceptron (MLP), K-Neural Network (NN), Random forest, and hybridization of these classifiers are used for both the identification of writer and classification of gender based on the handwriting sample. It has been reported that the maximum gender classification accuracy of 90.57% is reported with curve fitting-based features and hybridization of classifiers. And for writer identification, an accuracy of 87.76% is reported with curve fitting-based features and hybridization of classifiers. The authors also revealed performance evaluation by calculating metrics such as True Positive Rate (TPR) and False Positive Rate (FPR). Regarding future perspective, authors also directed the researchers of handwriting-based communities, to explore gender classification for other Indic scripts and also to utilize handwriting modality for the development of many utilitarian applications such as age, nationality, autopsy, mood, left or right-handedness or nationality from the handwriting modality.