Analysis of Multi-Scale Fractal Dimension to Classify Human Motion

TL;DR

This paper explores the use of 3D multi-scale fractal dimension combined with Fourier transform to classify human motion in videos, aiming to improve automatic human action recognition.

Contribution

It introduces a novel method integrating fractal analysis and Fourier transform for robust human motion classification in video data.

Findings

Different accuracy rates achieved across various databases.

The proposed method shows potential for applications in security, traffic, and healthcare.

First step towards automatic monitoring systems using fractal-based motion analysis.

Abstract

In recent years there has been considerable interest in human action recognition. Several approaches have been developed in order to enhance the automatic video analysis. Although some developments have been achieved by the computer vision community, the properly classification of human motion is still a hard and challenging task. The objective of this study is to investigate the use of 3D multi-scale fractal dimension to recognize motion patterns in videos. In order to develop a robust strategy for human motion classification, we proposed a method where the Fourier transform is used to calculate the derivative in which all data points are deemed. Our results shown that different accuracy rates can be found for different databases. We believe that in specific applications our results are the first step to develop an automatic monitoring system, which can be applied in security systems, traffic monitoring, biology, physical therapy, cardiovascular disease among many others.