Abstract:Most approaches to human action recognition is limited due to the use of simple action datasets under controlled environments or focus on excessively localized features without sufficiently exploring the spatio-temporal information. This paper proposed a framework for recognizing realistic human actions. Specifically, a new action representation is proposed based on computing a rich set of descriptors from keypoint trajectories. To obtain efficient and compact representations for actions, we develop a feature fusion method to combine spatial-temporal local motion descriptors by the movement of the camera which is detected by the distribution of spatio-temporal interest points in the clips. A new topic model called Markov Semantic Model is proposed for semantic feature selection which relies on the different kinds of dependencies between words produced by "syntactic" and "semantic" constraints. The informative features are selected collaboratively based on the different types of dependencies between words produced by short range and long range constraints. Building on the nonlinear SVMs, we validate this proposed hierarchical framework on several realistic action datasets.

A Semantic Feature for Human Motion Retrieval

A Human Motion Retrieval Algorithm Based on Logical Similarity

Text-Like motion representation for human motion retrieval

Human Motion Retrieval Using Topic Model.

A Human Motion Feature Based on Semi-Supervised Learning of GMM

Motion Retrieval Based On Movement Notation Language

Motion retrieval based on movement notation language: Motion Capture and Retrieval

An efficient algorithm for content-based human motion retrieval

Performance-Driven Motion Retrieval and its Usability Evaluation

Motion Retrieval Method Based on Geometric Feature Coding

Sketch-based Human Motion Retrieval Via Selected 2D Geometric Posture Descriptor

Vector Space Modeling and Retrieval of Human Motion Capture Data

Using feature models to present the semantic of human motion image

Introducing Semantic Information into Motion Graph

Content-Based human motion retrieval with automatic transition

Human motion retrieval based on freehand sketch

3D Human Motion Retrieval Based on Human Hierarchical Index Structure.

Video retrieval using VQ-based global motion features

Spatio-temporal Semantic Features for Human Action Recognition.

Study of Motion Feature Representation and Extraction Techniques for Video Semantic Analysis

A Perceptual User Interface For Motion Retrieval