Presenter Information

Samuel DixonFollow

Human Activity Workflow Parsing

Presentation Type

Poster

Abstract

From brushing your teeth to performing brain surgery, every activity has a logical flow of object interactions that make up the larger process. These object interactions can be represented in a work flow and can be further decomposed into primitive actions that are recognizable to a computer. Research has been done in recognizing specific, highly-structured workflows, but there is currently no general way to accurately reconstitute these computer recognizable primitive actions into higher-level activities.

The objective of this research is to show that compact, context-free grammars can be constructed through observation of human activity, and that these grammars can be used to accurately represent human activities in a computer-recognizable way. Parsers will be generated so that these grammars can be used for the recognition of future human activities.

Testing the recognition of human activity in the real world is a time-consuming and expensive task, which is why a virtual world will be constructed to create results that are easily-reproducible. This virtual world will allow for arbitrarily-complex workflows to be represented with an exhaustive set of possible executions. This platform can also be used for future work in human workflow replication.

Start Date

10-5-2018 12:00 PM

End Date

10-5-2018 2:00 PM

Genre/Form

posters

Subjects - Topical (LCSH)

Human activity recognition; Parsing (Computer grammar); Grammar

Type

Event

Format

application/pdf

Language

English

This document is currently not available here.

Share

COinS
 
May 10th, 12:00 PM May 10th, 2:00 PM

Human Activity Workflow Parsing

From brushing your teeth to performing brain surgery, every activity has a logical flow of object interactions that make up the larger process. These object interactions can be represented in a work flow and can be further decomposed into primitive actions that are recognizable to a computer. Research has been done in recognizing specific, highly-structured workflows, but there is currently no general way to accurately reconstitute these computer recognizable primitive actions into higher-level activities.

The objective of this research is to show that compact, context-free grammars can be constructed through observation of human activity, and that these grammars can be used to accurately represent human activities in a computer-recognizable way. Parsers will be generated so that these grammars can be used for the recognition of future human activities.

Testing the recognition of human activity in the real world is a time-consuming and expensive task, which is why a virtual world will be constructed to create results that are easily-reproducible. This virtual world will allow for arbitrarily-complex workflows to be represented with an exhaustive set of possible executions. This platform can also be used for future work in human workflow replication.