# US 2012/0251006 A1 · Feature design for character recognition

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2012/0251006 A1
- Document ID: US20120251006A1
- Status: Published application
- Kind: A1
- Application number: 13/526,236
- Filed: 2012-06-18
- Published: 2012-10-04
- Earliest claimed: 2007-06-29
- Assignees: Microsoft Corporation
- Inventors: Yu Zou; Ming Chang; Shi Han; Dongmei Zhang; Jian Wang
- Web page: https://patentgrep.com/patent/US20120251006A1

## CPC

- G06V 30/347: {Sampling; Contour coding; Stroke extraction}
- G06V 30/287: {of Kanji, Hiragana or Katakana characters}
- G06V: IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING

## Abstract

An exemplary method for online character recognition of characters includes acquiring time sequential, online ink data for a handwritten character, conditioning the ink data to produce conditioned ink data where the conditioned ink data includes information as to writing sequence of the handwritten character and extracting features from the conditioned ink data where the features include a tangent feature, a curvature feature, a local length feature, a connection point feature and an imaginary stroke feature. Such a method may determine neighborhoods for ink data and extract features for each neighborhood. An exemplary character recognition system may use various exemplary methods for training and character recognition.

## Claims

### Claim 1

A method comprising:
under control of one or more processors configured with executable instructions:
acquiring time sequential ink data for a character;
conditioning the time sequential ink data to separate ink trajectories of real strokes and imaginary strokes into a plurality of ink frames, the conditioning comprising generating a plurality of contiguous time sequential frames for each imaginary stroke;
extracting features from each ink frame of the conditioned ink data; and
recognizing the character based on the extracted features using a character recognition model.

### Claim 2 (depends on 1)

The method as recited in claim 1, wherein the acquiring comprises collecting the time sequential ink data for the character online as a user writes the character on a recording surface using a writing implement.

### Claim 3 (depends on 1)

The method as recited in claim 1, wherein each of the imaginary strokes comprises an imaginary linear movement connecting two neighboring real strokes.

### Claim 4 (depends on 1)

The method as recited in claim 1, wherein the conditioning further comprises re-sampling the time sequential ink data to convert the time sequential ink data into uniform sampled data, the sampled data enabling the character recognition model to perform character recognition independent of a speed of writing the character.

### Claim 5 (depends on 4)

The method as recited in claim 4, wherein the re-sampling further comprises filtering redundant data from the time sequential ink data.

### Claim 6 (depends on 1)

The method as recited in claim 1, wherein the character recognition model comprises a Bayesian model or a Hidden Markov Model.

### Claim 7 (depends on 1)

The method as recited in claim 1, further comprising analyzing each ink frame based on one or more predetermined features, the one or more predetermined features comprising a tangent feature, a curvature feature, a local length feature, a connection point feature and/or an imaginary stroke feature.

### Claim 8 (depends on 7)

The method as recited in claim 7, wherein the tangent feature is used for assessing the character as a link of direction movements and capturing a writing direction for each ink frame.

### Claim 9 (depends on 7)

The method as recited in claim 7, wherein the curvature feature is used for recognizing a state transition from a previous writing direction to a next writing direction for the character.

### Claim 10 (depends on 7)

The method as recited in claim 7, wherein the local length feature is used for characterizing a total trajectory length of a neighborhood of an ink frame of the plurality of ink frames.

### Claim 11 (depends on 7)

The method as recited in claim 7, wherein the connection point feature is used for determining one or more connection points for a neighborhood, the one or more connection points comprises at least one of a “X” connection point or a “T” connection point.

### Claim 12

A character recognition system comprising:
a processing unit;
memory coupled to the processing unit;
a feature extraction module stored in the memory and executable by the processing unit to recognize a handwritten character by using a character recognition model and to extract features from ink data of the handwritten character, the features being extracted comprising at least one of a tangent feature, a curvature feature, a local length feature, a connection point feature, or an imaginary stroke feature.

### Claim 13 (depends on 12)

The system as recited in claim 12, further comprising:
an acquisition module that uniformly samples the ink data; and
a frame module that analyzes the features of the ink data by breaking the ink data into a plurality of frames.

### Claim 14 (depends on 12)

The system as recited in claim 12, wherein the ink data comprises time sequential ink data, and wherein the system further comprises an acquisition module that re-samples the time sequential ink data to convert the time sequential ink data into uniform sampled data, the sampled data enabling the character recognition model to perform character recognition independent of a speed of writing the handwritten character.

### Claim 15 (depends on 14)

The system as recited in claim 14, wherein the acquisition module is further configured to filter redundant data from the time sequential ink data.

### Claim 16 (depends on 12)

The system as recited in claim 12, wherein the character recognition model comprises a Bayesian model or a Hidden Markov Model.

### Claim 17 (depends on 12)

The system as recited in claim 12, the system being configured to be integrated into a computing device.

### Claim 18

One or more computer storage media encoded with executable instructions that, when executed by one or more processors, configure the one or more processors to perform acts comprising:
acquiring time sequential ink data for a character;
conditioning the time sequential ink data to separate ink trajectories of real strokes and imaginary strokes into a plurality of ink frames, the conditioning comprising generating a plurality of contiguous time sequential frames for each imaginary stroke;
extracting features from each ink frame of the conditioned ink data; and
recognizing the character based on the extracted features using a character recognition model.

### Claim 19 (depends on 18)

The one or more computer storage media as recited in claim 18, wherein the character recognition model comprises a Bayesian model or a Hidden Markov Model.

### Claim 20 (depends on 18)

The one or more computer storage media as recited in claim 18, wherein the conditioning further comprises re-sampling the time sequential ink data to convert the time sequential ink data into uniform sampled data, the sampled data enabling the character recognition model to perform character recognition independent of a speed of writing the character.
