Handwritten formula recognition, Incremental learning, Feedback mechanism, Convolutional neural network
Abstract
Handwritten formula recognition systems have significant application value in education, scientific research, and smart note applications. However, due to the complexity and irregularity of handwritten formulas, improving the accuracy and efficiency of recognition has been a challenging research hotspot. This paper focuses on enhancing handwritten formula recognition methods and proposes a feedback system based on incremental learning, which improves recognition accuracy and enables user personalization. The article utilizes CNN convolutional neural networks for model training, and a regularization-based incremental learning system is used to enhance the model. The proposed algorithm is validated using the mathematical formula data provided by the Competition Organization for Recognition of Handwritten Mathematical Expressions (CROHME) and user experience feedback. Results show the feasibility of the proposed method.
