A Method for Rejecting Recognition of Handwritten Characters

A character and opponent technology, applied in the field of handwritten character rejection, can solve the problems of poor recognition effect, difficult feature recognition, low accuracy, etc., to improve accuracy, improve reliability, enhance robustness and stability Effect

Inactive Publication Date: 2020-04-21
林辉
View PDF11 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0005] However, when there is a handwriting of a character that does not belong to the character set supported by the system or is not a meaningful character handwriting input into the recognition algorithm, the recognition algorithm often classifies it as a certain character in the character set with a high confidence score, so that The result is obviously absurd; in the existing handwritten character recognition process, the solution is to add a large number of handwriting samples of characters outside the supported character set and use them as a rejection class to train, but this requires a lot of manual work At the same time, because these character handwriting samples do not have the same characteristics, it is difficult to identify them through algorithms such as machine learning, so the recognition effect is not good, and the recognition accuracy is not high

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • A Method for Rejecting Recognition of Handwritten Characters
  • A Method for Rejecting Recognition of Handwritten Characters

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0085] Such as figure 1 and figure 2As shown, a method for realizing rejection of handwritten characters comprises the following steps:

[0086] a. Collect handwriting data, and perform training processing on the collected handwriting data to obtain training handwriting data;

[0087] b. Establish a stacked RBN neural network, read the training handwriting data in step a, build, train and save the model;

[0088] c. Read the training handwriting data in step a and the model in step b, and calculate the reconstruction error data of the training handwriting data according to the model;

[0089] d. Repeat step c multiple times to obtain the reconstruction error data set R, obtain the confidence interval by calculating the reconstruction error data set R, and save the confidence interval data;

[0090] e. Input new handwritten handwriting data, and calculate the reconstruction error of the new handwritten handwriting data, and determine whether to reject the new handwritten ha...

Embodiment 2

[0094] In this embodiment, on the basis of Embodiment 1, said step a includes the following steps:

[0095] a1, obtain the handwriting point coordinate sequence T1, and obtain the maximum ordinate value h and the maximum abscissa value w of T1;

[0096] a2, according to the maximum ordinate value h and the maximum abscissa value w of the handwriting point coordinate sequence T1, calculate the shrinkage ratio ShrinkageRatio, and according to the scaling ratio ShrinkageRatio the handwriting point is scaled to the horizontal and vertical coordinates that are parameters Len Two-dimensional matrix M;

[0097] a3. When the scaling ratio ShrinkageRatio is greater than 1, fill the gaps between the enlarged handwriting points to obtain coordinate data of multiple two-dimensional matrices M;

[0098] a4. Concatenate the two-dimensional matrix M row by row into a vector Vo whose length is Len*Len.

[0099] This program realizes the data collection, processing and training of handwritte...

Embodiment 3

[0101] In this embodiment, on the basis of Embodiment 2, the step a1 includes the following steps:

[0102] a11, judge whether the presentation form of the handwritten handwriting data is a picture or a series of handwriting coordinate points, if it is a picture, then enter step a12; if it is a series of handwriting coordinate points Tp, then directly enter step a15;

[0103] a12. Determine whether the picture is a color RRG three-channel image or a single-channel grayscale image Io, if it is a color RRG three-channel image, convert it into a single-channel grayscale image Io, and enter step a13; if it is a single-channel grayscale image Io, then directly enter step a13;

[0104] a13, according to the OTSU algorithm, the single-channel grayscale image Io is thresholded to form a binary image Ir; and when the handwriting pixel is white, get Ir=1-Ir;

[0105] a14. According to the tracking algorithm, track a black pixel point of a certain length, continuous, and within a certai...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to view more

PUM

No PUM Login to view more

Abstract

The invention discloses a method for realizing rejection of handwritten characters. First, collect and train handwriting data to obtain training handwriting data; build a stacked RBN neural network, read the training handwriting data, and build, train and save the model; Read the training handwriting data and model, and calculate the reconstruction error data of the training handwriting data; repeat the calculation of the reconstruction error data multiple times, obtain the reconstruction error data set, and obtain and save the reconstruction error data set R by calculating confidence interval; then input new handwritten handwriting data, and determine whether to reject the new handwritten handwriting data by judging whether the reconstruction error of the new handwritten handwriting data is within the confidence interval. The present invention uses the decoding reconstruction loss of the deep belief network as the confidence evaluation of whether the handwriting of the input character belongs to the character type supported by the system and a threshold to judge whether it is rejected; it can effectively reject the handwriting of the character, so that it can be used at a low error rejection rate. Under the premise, it has a high correct rejection rate.

Description

technical field [0001] The invention relates to the field of handwritten character recognition, in particular to a method for realizing rejection of handwritten characters. Background technique [0002] With the development of pattern recognition and artificial intelligence technology, handwritten text machine recognition technology has been effectively applied, and related theories, technologies, and methods have also been greatly developed. Handwritten character recognition technology is the basic function of handwritten text recognition technology. The module has received extensive attention and vigorous research. [0003] The problem of handwritten character recognition is essentially a classification problem. Each recognition system has its supported character set. The algorithm of handwritten character recognition needs to determine which type of character a handwritten character handwriting belongs to in the character set, and in most cases Give a confidence score. ...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to view more

Application Information

Patent Timeline
no application Login to view more
Patent Type & Authority Patents(China)
IPC IPC(8): G06K9/00G06K9/62G06N3/08
CPCG06N3/088G06V30/32G06F18/214
Inventor 林辉
Owner 林辉
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Try Eureka
PatSnap group products

Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic.

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap