Shape recognition using partial shapes
AssigneeMicrosoft Corporation
InventorsAlexander Kolmykov-Zotov, Sashi Raghupathy, Xin Wang
Life of this patent
Estimating the term…
Abstract
Shape recognition is performed based on determining whether one or more ink strokes is not part of a shape or a partial shape. Ink strokes are divided into segments and the segments analyzed employing a relative angular distance histogram. The histogram analysis yields stable, incremental, and discriminating featurization results. Neural networks may also be employed along with the histogram analysis to determine complete shapes from partial shape entries and autocomplete suggestions provided to users for conversion of the shape into a known object.
Claims20
A method to be executed at least in part in a computing device for shape recognition, the method comprising:
- receiving ink entry comprising at least one stroke in a drawing mode;
- dividing the entire ink entry into segments that each have a predefined length;
- generating a Relative Angular Distance (RAD) histogram based on analyzing distances and angles of consecutive segment pairs;
- determining, based on a RAD histogram pattern, whether the ink entry is one of:
- an ambiguous partial shape, an unambiguous partial shape, a complete shape, and a non-shape;
- and
- upon determination that the ink entry is an unambiguous partial shape or a complete shape providing a suggested completion of the partial shape.
A computing device arranged for shape recognition, the computing device comprising:
- a user application with a user interface for accepting ink entries and displaying parsing results to a user;
- a parsing platform that includes an engine stack with a shape recognition engine, wherein the shape recognition engine includes:
- a featurization block for dividing an entire received ink entry into predefined length segments, generating a RAD histogram based on distances and angles of consecutive segment pairs, and determining whether the ink entry is one of:
- an ambiguous partial shape, an unambiguous partial shape, a complete shape, and a non-shape based on a RAD histogram pattern;
- a machine learning block for determining a complete shape from one of:
- an ambiguous partial shape and an unambiguous partial shape;
- and
- a post processing block that provides a suggested completion of a partial shape to a recognizable complete shape based on a result from the RAD histogram pattern analysis and the machine learning block determination.
A computer-readable storage medium that is not a signal with instructions encoded thereon for shape recognition, the instructions comprising:
- receiving ink entry in a handwriting recognition application set to a drawing mode;
- dividing the ink entry into segments of a predefined length, wherein a number and a size of the segments is adjustable based on one of:
- a system parameter and a user preference;
- generating a Relative Angular Distance (RAD) histogram based on analyzing distances and angles of consecutive segment pairs;
- determining whether the ink entry is an unambiguous partial shape based on:
- employing a machine learning algorithm in addition to generating the RAD histogram to recognize a shape based on one of the ambiguous partial shape and the unambiguous partial shape;
- and
- analyzing a RAD histogram pattern;
- if when the ink entry is determined to be an unambiguous partial shape, providing a suggested completion of the partial shape to a recognizable complete shape.
Dependencies are inferred from the claim text.