Research On Optimization Method Of English Handwriting Recognition Technology

Handwriting recognition technology has been studied for a long time,and the recognition accuracy has been gradually improved,but so far there are still many unsolved problems in this topic,especially in some practical application scenarios.Writing struck out text is inevitable in the free form handwriting manuscript,which caused great reduction of recognition accuracy and increased the difficulty of recognition,reduced the recognition accuracy.Even though,a few research on struck out text detection exist,which trying to improve the accuracy of handwriting recognition by adding a struck out text filter module,the time and resource for training and inference are significantly increased.The module leads to a sharp drop in recognition efficiency,resulting in contradictions that are difficult to reconcile.In order to solve the recognition accuracy reduction problem caused by struck out text in practical applications without reducing the recognition efficiency,considering to combine the traditional handwriting recognition module and struck out text detection module to splitting normal handwriting word images and struck out word images.The recognition result of struck out texts are marked with special characters which can distinguish from normal words.This method integrates the detecting struck out text into the recognition module,the cost of time and resource of which is reduced so as to meet the needs of practical applications.To further improve the recognition accuracy,consider adding a post-processing module in the traditional recognition process.The confidence-based post-processing method has certain advantages,but these methods have a high rejection rates to get a high recognition accuracy.The high reject rate means high labor cost and low processing efficiency in actual application scenarios.A dictionary based confidence calculation method got the same recognition accuracy with the lower rejection rate.The experimental results show that the method for detecting struck out text proposed in this topic improves the recognition accuracy of handwriting system and ensures the recognition efficiency.At the same time,the rejection rate is reduced during the postprocessing process,and then the post-processing cost is reduced.It proves that the method in this study has certain theoretical research value and high practical application value.