US 5,729,629 A · Handwritten symbol recognizer

Abstract

A handwritten symbol recognizer adapted to automatically recognize various handwritten symbols including Kanji character is disclosed. The symbol recognizer includes a data processor coupled to a user interface that receives handwritten input symbols from a user and displays reference symbols determined by the symbol recognizer to correspond to the handwritten symbols. The data processor includes a memory unit which temporarily stores coordinate information representing features of the input symbols as received from the user interface. The data processor includes a symbol analyzer that retrieves the coordinate information from the memory unit and translates the coordinate information for each written feature into a feature code representing one of a predetermined number of feature models stored in the memory. The symbol analyzer also creates a geometric representation of the input symbol for comparison with one or more geometric models stored in the memory unit. The data processor also includes a label comparator adapted to compare the feature codes for the input symbol with feature codes of reference symbols stored in the memory unit and to identify the reference symbol or symbols having feature codes that most closely match the feature codes of the input symbol. The data processor also includes a geometric comparator that compares stored geometric models corresponding to the reference symbols most closely matching the input label with the geometric representation of the input symbol. The geometric comparator uses its comparison to identify the reference symbols whose geometric model or models most closely match the geometric representation of the input symbol. The reference symbol that most closely matches the handwritten input symbol, as determined by the label and geometric comparators, is sent to the user interface which displays the reference symbol to the user.

Claims61

Dependent claims

An automatic handwritten symbol recognizer that recognizes a handwritten input symbol having sequential features written according to a time sequential order, comprising:

a symbol analyzer that receives a time sequence of coordinate information representing the sequential features of the input symbol according to the time sequential order in which the sequential features were written, translates the coordinate information for each written sequential feature into a feature code representing one of a predetermined number of feature models, and produces an input label of feature codes ordered according to the time sequential order in which the features were written, the symbol analyzer also creates a geometric representation of the input symbol, the geometric representation including data representing positions of the features of the input symbol;
a memory storing reference symbol information including a plurality of label models and a plurality of geometric models, each label model including a plurality of feature codes in a predetermined order and corresponding to one of the geometric models, each geometric model storing data representing positions of features of a reference symbol represented by the geometric model;
a label comparator that compares the input label with the label models without regard for positions of the features represented by the feature codes of the input label and identifies a most closely matching subset of the stored label models by identifying the label model or models that most closely match the input label regardless of the positions of the features represented by the feature codes of the input label;
and
a geometric comparator that compares only the geometric models corresponding to the label models of the most closely matching subset with the geometric representation of the input symbol and identifies the geometric model or models that most closely match the geometric representation of the input symbol.

2 more elements

A computerized method of recognizing a handwritten input symbol having features written according to a time sequential order, comprising:

receiving a time sequence of coordinate information representative of the input symbol;
translating the coordinate information into an input label of one or more feature codes ordered according to the time sequential order of the features, each feature code representing one of a predetermined number of feature models;
comparing the input label to a plurality of stored label models without regard for positions of the features represented by the feature codes of the input label, each label model corresponding to a stored geometric model and including a defined sequence of feature codes that are compared to the input label according to the time sequential order in which the features of the input symbol were written, each geometric model including data representing positions of features of a reference symbol represented by the geometric model;
identifying a most closely matching subset of the stored label models by identifying the label model or models most closely matching the input label regardless of the positions of the features represented by the feature codes of the input label;
creating a geometric representation of the input symbol, the geometric representation including data representing positions of features of the input symbol;
comparing only the geometric models corresponding to the label models of the most closely matching subset with the geometric representation of the input symbol;
and
identifying the geometric model or models that most closely match the input symbol geometric representation.

5 more elements

A computerized method of recognizing a handwritten input symbol having features written according to a time sequential order, comprising:

receiving a time sequence of coordinate information representative of the input symbol;
vectorizing the coordinate information into a plurality of vector codes each of which represents a direction taken while the input symbol was being handwritten;
translating the vector codes into an input label of one or more feature codes ordered according to the time sequential order in which the features were written, each feature code representing one of a predetermined number of feature models;
comparing the input label to a plurality of stored label models having the same number of feature codes as the input label during a first pass and comparing the input label with label models having one more or one less feature code during a second pass, each label model corresponding to a geometric model of a written symbol;
and
identifying the input label model or models that most closely match the input label.

3 more elements

An automatic handwritten symbol recognizer that recognizes a handwritten input symbol having features written according to a time sequential order, comprising:

a symbol analyzer that receives a time sequence of coordinate information representing the features of the input symbol and translates the coordinate information for each feature into a feature code representing one of a predetermined number of feature models, the symbol analyzer also creates a geometric representation of the input symbol by saving a subset of the coordinate points of each feature and discarding the remaining coordinate points of the features, the subset saved including an end point of the feature;
a memory storing reference symbol information including a plurality of label models and a plurality of geometric models, each label model including a predetermined sequence of feature codes and corresponding to one of the geometric models;
a label comparator that compares sequentially the feature codes of the input symbol with the feature codes of the label models according to the time sequential order in which the features of the input symbol were handwritten and identifies the label model or models whose feature codes most closely match the feature codes of the input symbol;
and
a geometric comparator that compares geometric models corresponding to the label models whose feature codes most closely match the feature codes of the input symbol with a geometric representation of the input symbol and identifies the geometric model or models that most closely match the geometric representation of the input symbol.

2 more elements

A computer-readable storage medium including executable instructions for controlling a computer to recognize a handwritten input symbol having features written according to a time sequential order, the executable instructions comprising instructions for:

receiving a time sequence of coordinate information representative of the input symbol;
translating the coordinate information into an input label of one or more feature codes ordered according to the time sequential order of the features, each feature code representing one of a predetermined number of feature models;
comparing the input label to a plurality of stored label models without regard for positions of the features represented by the feature codes of the input label, each label model corresponding to a stored geometric model and including a defined sequence of feature codes that are compared to the input label according to the time sequential order in which the features of the input symbol were written, each geometric model including data representing positions of features of a reference symbol represented by the geometric model;
identifying a most closely matching subset of the stored label models by identifying the label model or models most closely matching the input label regardless of the positions of the features represented by the feature codes of the input label;
creating a geometric representation of the input symbol, the geometric representation including data representing positions of features of the input symbol;
comparing only the geometric models corresponding to the label models of the most closely matching subset with the geometric representation of the input symbol;
and
identifying the geometric model or models that most closely match the input symbol geometric representation.

5 more elements

receiving a time sequence of coordinate information representing the features of the input symbol;
translating the coordinate information for each feature into a feature code representing one of a predetermined number of feature models;
creating a geometric representation of the input symbol by saving a subset of the coordinate points of each feature and discarding the remaining coordinate points of the features, the subset saved including an end point of the feature;
storing reference symbol information including a plurality of label models and a plurality of geometric models, each label model including a predetermined sequence of feature codes and corresponding to one of the geometric models,
comparing sequentially the feature codes of the input symbol with the feature codes of the label models according to the time sequential order in which the features of the input symbol were handwritten,
identifying the label model or models whose feature codes most closely match the feature codes of the input symbol;
comparing geometric models corresponding to the label models whose feature codes most closely match the feature codes of the input symbol with the geometric representation of the input symbol;
and
identifying the geometric model or models that most closely match the geometric representation of the input symbol.

6 more elements

Dependencies are inferred from the claim text.

Handwritten symbol recognizer

Life of this patent

Handwritten symbol recognizer

Life of this patent

Abstract

Claims61