Design and Evaluation of a Hybrid Feature Descriptor based Handwritten Character Inference Technique

Handwritten character inference refers to the task of converting the character images into the corresponding digital characters. A hybrid descriptor based character recognition algorithm is proposed in this paper. The novel hybrid descriptor model exhibits a better character inference rate when empirically assessed against the standard Chars74K dataset and the self curated Telugu handwritten character dataset. In order for the computational complexity of the recognition task to reduce, the number of comparisons required to recognize the character is reduced using an effective pruning technique. Experimentation of the proposed technique depicted that the proposed hybrid model recognizes the characters at a better rate compared to the techniques using HMM model, HoG, run length features and transform based feature extraction techniques.