Handwriting model training method, handwriting character recognition method and device, apparatus and medium

Abstract

The invention discloses a handwriting model training method, a handwriting character recognition method and device, an apparatus and a medium. The handwriting model training method comprises the following steps of: obtaining a standard Chinese character training sample, and batching the standard Chinese character training sample according to a preset batch; obtaining the Chinese character recognition model by updating the network parameters with time-dependent back propagation algorithm and training the batched Chinese character training samples with bi-directional long-short time memory neural network; obtaining and using non-standard Chinese character training samples, and training to obtain the adjusted Chinese handwritten character recognition model; obtaining a Chinese character sample to be tested and adopting the same to obtain an error character training sample; based on the time-dependent back propagation algorithm of batch gradient descent, updating the network parameters ofChinese handwritten character recognition model by training samples of error characters, and obtaining the target Chinese handwritten character recognition model. A target Chinese handwritten character recognition model with high recognition rate can be obtained by using that handwritten model train method.

Description

technical field [0001] The invention relates to the field of Chinese character recognition, in particular to a handwriting model training method, a handwriting recognition method, a device, equipment and a medium. Background technique [0002] Traditional handwritten character recognition methods mostly include steps such as binarization processing, character segmentation, feature extraction, and support vector machine for recognition. When traditional handwritten character recognition methods are used to recognize rough non-standard characters (handwritten Chinese characters), the accuracy of recognition is high. The accuracy is not high, making its recognition effect unsatisfactory. Traditional handwritten character recognition methods can only recognize standard characters to a large extent, and the accuracy rate is low when recognizing various handwritten characters in real life. Contents of the invention [0003] Embodiments of the present invention provide a handwri...

AI Technical Summary

Problems solved by technology

[0003] The embodiment of the present invention provides a handwriting model training method, d

Images