An optimized neural Network-based character recognition and relation finding for mathematical expression images

Recognition of Mathematical Expressions (MEs) is a significant issue with numerous practical applications. Due to the range of writing styles and ME forms, ME recognition might be difficult. In many existing models, the exactness of the recognition system is degraded because of the additional noise present in the input, and the characters are wrongly predicted. Therefore, a novel Chimp-based Spiking Neural Recognition (CbSNR) framework was proposed to recognize the character from the handwritten/printed ME images. Initially, the input image datasets collected are noise-filtered in the preprocessing phase, and the features are extracted in the feature extraction process by the tracking function of the Chimp. Preprocessed images are then segmented into individual characters and are recognized in the recognition phase. Moreover, the Chimp position updating function is used to find the relation between the recognized characters. Furthermore, this test process is executed in the Python environment, and the robustness score has been valued in terms of F-score, recall, Accuracy, precision and error rate. The estimated outputs of the presented model were compared with existing approaches to validate the improvement score. The proposed model produced the highest character recognition rate compared to existing techniques.