Radon-Based Text and Script-Independent Gender Detection Using Symbolic Dynamic Filtering

In this paper an automated system based on feature extraction of new techniques is presented to detect the gender from the scanned images (off-line) handwriting samples. In order to show the difference between examples of handwriting, in the first step Radon transform is taken from the handwritten image, and then each handwriting sample features are extracted using symbolic dynamic filtering. Training and classification of extracted features from the samples are carried out by the multi-layer perceptron neural network. At the end, to determine the effectiveness of the proposed method, experiments are carried out on the Multi Script Handwritten Database (MSHD). In addition, two new challenges of text and script-independent gender detection are explored. Experiences show that the proposed method improves the detection rate compared to the previous works such as fractals, chain codes and textures. The best detection rate is able to achieve accuracy of 84.9% in experiences.