Electronic handwriting processor with convolutional neural networks

Patent metadata

Field	Value
Publication number	US-10949660-B2
Application number	US-201815905690-A
Country	US
Kind code	B2
Filing date	Feb 26, 2018
Priority date	Apr 10, 2017
Publication date	Mar 16, 2021
Grant date	Mar 16, 2021

Abstract

An improved machine learning system is provided. For example, a content management server may provide real-time analysis of a user's handwriting to assess the user's knowledge of a language, including using a convolution neural network method. The convolution neural network method may be executed to normalize at least some identified strokes in the user's handwritten user input. Normalization may be performed by translating a window comprising a subset of pixels in a digital representation of the handwritten user input amongst a plurality of pixels in the digital representation.

First claim

What is claimed is:

1. A computer-implemented method for automated handwriting assessment comprising:
generating a digital representation of a handwritten user input;
identifying strokes in the digital representation of the handwritten user input;
storing the identified strokes;
executing a heuristic segmentation method to correct a slant of at least some of the identified strokes using linear regression;
executing a convolution neural network model to normalize the at least some of the identified strokes by performing a convolution by translating a window comprising a subset of pixels in the digital representation of the handwritten user input amongst a plurality of pixels in the digital representation;
segmenting the digital representation of the handwritten user input into normalized digital strokes;
determining a number of characters in a correct response word;
determining at least one character associated with the normalized digital strokes, wherein the at least one character includes a number of characters equal to the number of characters in the correct response word; and
generating a word hypothesis by applying a model to the at least one character.

2. The computer-implemented method of claim 1 , further comprising:
identifying a length of the identified strokes; and
generating an analysis set from the identified strokes, wherein the analysis set excludes strokes having a length below a threshold value, wherein the at least some of the identified strokes comprise the strokes in the analysis set.

3. The computer-implemented method of claim 1 , wherein correcting the at least some of the identified strokes comprises estimating the slant of the at least some of the identified strokes.

4. The computer-implemented method of claim 1 , wherein the at least some of the identified strokes are normalized with respect to a top left point of a part of the digital representation of the handwritten user input.

5. The computer-implemented method of claim 4 , wherein the part of the digital representation of the handwritten user input comprises a word.

6. The computer-implemented method of claim 1 , wherein the normalizing the at least some of the identified strokes comprises matching a size of the identified strokes to a predefined symbol and padding the identified strokes with a blank space.

7. The computer-implemented method of claim 1 , wherein normalizing the at least some of the identified strokes comprises converting the identified strokes into a standard image.

8. The computer-implemented method of claim 7 , wherein the standard image comprises a 32×32 gray image.

9. A computer device including a content management system for improving distribution and conversion of academic data, the content management system comprising:
one or more processors; and
one or more memories coupled with the one or more processors, wherein the one or more processors and the one or more memories are configured to:
generate a digital representation of a handwritten user input;
identify strokes in the digital representation of the handwritten user input;
store the identified strokes;
execute a heuristic segmentation method to correct a slant of at least some of the identified strokes using linear regression;
execute a convolution neural network model to normalize the at least some of the identified strokes by performing a convolution by translating a window comprising a subset of pixels in the digital representation of the handwritten user input amongst a plurality of pixels in the digital representation;
segment the digital representation of the handwritten user input into normalized digital strokes;
determine a number of characters in a correct response word;
determine at least one character associated with the normalized digital strokes, wherein the at least one character includes a number of characters equal to the number of characters in the correct response word; and
generate a word hypothesis by applying a model to the at least one character.

10. The computer device of claim 9 , wherein the one or more memories are further configured to:
identify a length of the identified strokes; and
generate an analysis set from the identified strokes, wherein the analysis set excludes strokes having a length below a threshold value, wherein the at least some of the identified strokes comprise the strokes in the analysis set.

11. The computer device of claim 9 , wherein correcting the at least some of the identified strokes comprises estimating a slant of the at least some of the identified strokes.

12. One or more non-transitory computer-readable storage media device collectively storing computer-executable instructions that, when executed by one or more computer systems, configure the one or more computer systems to collectively perform operations comprising:
generating a digital representation of a handwritten user input;
identifying strokes in the digital representation of the handwritten user input;
storing the identified strokes;
executing a heuristic segmentation method to correct a slant of at least some of the identified strokes using linear regression;
executing a convolution neural network model to normalize the at least some of the identified strokes by performing a convolution by translating a window comprising a subset of pixels in the digital representation of the handwritten user input amongst a plurality of pixels in the digital representation;
segmenting the digital representation of the handwritten user input into normalized digital strokes;
determining a number of characters in a correct response word;
determining at least one character associated with the normalized digital strokes, wherein the at least one character includes a number of characters equal to the number of characters in the correct response word; and
generating a word hypothesis by applying a model to the at least one character.

13. The one or more non-transitory computer-readable storage media device of claim 12 , wherein the operations further comprise:
identifying a length of the identified strokes; and
generating an analysis set from the identified strokes, wherein the analysis set excludes strokes having a length below a threshold value, wherein the at least some of the identified strokes comprise the strokes in the analysis set.

14. The one or more non-transitory computer-readable storage media device of claim 12 , wherein correcting the at least some of the identified strokes comprises estimating a slant of the at least some of the identified strokes.

15. The one or more non-transitory computer-readable storage media device of claim 12 , wherein the at least some of the identified strokes are normalized with respect to a top left point of a part of the digital representation of the handwritten user input.

Assignees

• Pearson Education Inc

Inventors

• Wang Zhaodong

Classifications

• G09B5/02 Primary

  with visual presentation of the material to be studied, e.g. using film strip · CPC title

• G06V30/18057

  Integrating the filters into a hierarchical structure, e.g. convolutional neural networks [CNN] · CPC title

• G06V30/166

  Normalisation of pattern dimensions · CPC title

• G06V30/347

  Sampling; Contour coding; Stroke extraction · CPC title

• G06N3/045

  Combinations of networks · CPC title