The identification of accurate features is the initial task for benchmarked handwriting recognition. For handwriting recognition, the objective of feature computation is to find those characteristics of a handwritten stroke that depict the class of a stroke and make it separable from the rest of the stroke classes. The present study proposes a feature extraction technique for online handwritten strokes based on a self controlled Ramer-Douglas-Peucker (RDP) algorithm. This novel approach prepares a smaller length feature vector for different shaped online handwritten strokes without preprocessing and without any control parameter to RDP. Thus, it also overcomes the shortcomings of the traditional chain code based feature extraction approach that requires preprocessing of data, and the original RDP algorithm that requires a control parameter as an input to RDP. We further propose a deep learning network of 1-dimensional convolutional neural networks (Conv1Ds) for recognition, which trains in few minutes due to the smaller dimension of the convolution combined with smaller length feature vectors. The proposed approach can be applied to different scripts and different writing styles. The key aim of the present study is to provide a script independent feature extraction technique that is well suited for smaller devices. It improves the recognition over the best reported accuracy in the literature which was achieved using hidden Markov models with directional features, from 87.67% to 95.61% on a Gurmukhi dataset. For Unipen online handwriting datasets the results are at par with the literature.
This is an author-produced, peer-reviewed version of this article. The final, definitive version of this document can be found online at Applied Intelligence, published by Springer. Copyright restrictions may apply. https://doi.org/10.1007/s10489-020-01632-4
Singh, Sukhdeep; Chauhan, Vinod Kumar; and Barney Smith, Elisa H.. (2020). "A Self Controlled RDP Approach for Feature Extraction in Online Handwriting Recognition Using Deep Learning". Applied Intelligence, 50(7), 2093-2104. https://doi.org/10.1007/s10489-020-01632-4