Primitive Based Action Representation and Recognition

Sanmohan Baby, Volker Krüger

Research output: Contribution to journalJournal articlepeer-review

7 Citations (Scopus)

Abstract

There has been a recent interest in segmenting action sequences into
  meaningful parts (action primitives) and to model actions on a
  higher level based on these action primitives. Unlike previous works where action primitives are defined
   a-priori and search is made for them later, we present a sequential and statistical
    learning algorithm for
  automatic detection of the action primitives and the action grammar
  based on these primitives.  We model a set of actions using a
  single HMM whose structure is learned incrementally as we observe
  new types.   Actions are modeled with sufficient number of Gaussians which would
  become the states of an HMM for an action. For different actions we
  find the states that are common in the actions which are then treated
  as an action primitive.
Original languageEnglish
Book seriesLecture Notes in Computer Science
Pages (from-to)31-40
Number of pages10
ISSN0302-9743
DOIs
Publication statusPublished - 2009
EventSCIA 2009 - Oslo, Norway
Duration: 15 Jun 200918 Jun 2009
Conference number: 16

Conference

ConferenceSCIA 2009
Number16
Country/TerritoryNorway
CityOslo
Period15/06/200918/06/2009

Fingerprint

Dive into the research topics of 'Primitive Based Action Representation and Recognition'. Together they form a unique fingerprint.

Cite this