Handwritten Chinese grass-style phrase identification method irrelevant to rotation

Abstract

The invention provides a handwritten Chinese grass-style phrase identification method irrelevant to rotation. The identification method comprises the following steps: (1) carrying out rotation direction correction on handwritten phrase input and the rotation direction correction adopts a rotation direction correction method based on gravity center balancing to ensure that the center of each word of the phrase is positioned on a horizontal line; (2) extracting directional characteristics of the entire phrase based on elastic grids to ensure that the extracted characteristic vectors are in Gaussian distribution; and (3) carrying out classifying identification on the entire phrase by means of a corrected quadratic decision function classifier to acquire an identification result.

Description

technical field [0001] The invention belongs to the technical field of using computer pattern recognition to recognize user's handwritten Chinese, in particular to a rotation-independent Chinese cursive phrase recognition method. technical background [0002] Handwritten Chinese character recognition generally refers to the recognition technology that the user writes Chinese characters through a handwriting input device (such as: tablet, touch screen, mouse, etc.), and at the same time, the computer converts the Chinese character writing trajectory collected by the handwriting input device into the corresponding Chinese character machine internal code. It has been widely used in real-time handwriting input, automatic input of Chinese electronic data processing, Chinese text compression, office automation and computer-aided teaching, and has brought huge economic and social benefits. Traditional handwriting recognition technology usually adopts single-character input method, ...

AI Technical Summary

Problems solved by technology

[0007] The above methods all have different degrees of defects. For example, the first method needs to click the button, which is time-consuming and laborious and not natural enough. The second method needs to pause for a while after writing. Although it is not laborious, it is time-consuming and directly affects the speed of input. , the third and fourth methods need to move a long distance after writing a character before starting to write the next character, which is not in line with people's natural writing habits

The handwriting method of multi-character continuous cursive script is the most natural and fastest Chinese handwriting input method for people, but the above methods obviously cannot solve the problem of multi-character continuous stroke input

[0008] Aiming at the inconvenience of single character input, handwritten phrase recognition has attracted more and more attention. At present, the commonly used phrase recognition method is to use the phrase segmentation method to divide the phrase into individual characters, and then use the single character recognition technology to identify each Segmented single characters are recognized to obtain the final recognition result

[0009] However, the biggest defect of this method is that the result of phrase segmentation has a very important impact on the final recognition result. If an error occurs in a certain segmentation path, it will lead to a final recognition error.

Moreover, if the phrase written by the user is inclined at a certain angle when writing, it is prone to errors in the segmentation path. Therefore, in terms of current technological development, the phrase segmentation technology is facing many challenges, especially for inclined handwritten phrases. Recognition has not yet developed a very mature and effective phrase segmentation technology

Images