Similarity Estimation for Large-Scale Human Action Video Data on Spark

Weihua Xu,Young-Koo Lee,Mostafijur Akhond,Kifayat Khan,Md Uddin,Md Hossain,Batjargal Dolgorsuren

doi:10.3390/app8050778

Abstract

The amount of human action video data is increasing rapidly due to the growth of multimedia data, which increases the problem of how to process the large number of human action videos efficiently. Therefore, we devise a novel approach for human action similarity estimation in the distributed environment. The efficiency of human action similarity estimation depends on feature descriptors. Existing feature descriptors such as Local Binary Pattern and Local Ternary Pattern can only extract texture information but cannot obtain the object shape information. To resolve this, we introduce a new feature descriptor, namely Edge based Local Pattern descriptor (ELP). ELP can extract object shape information besides texture information and ELP can also deal with intensity fluctuations. Moreover, we explore Apache Spark to perform feature extraction in the distributed environment. Finally, we present an empirical scalability evaluation of the task of extracting features from video datasets.

Highlights

A large amount of human action video data has sharply increased due to the expansion of multimedia devices, the Internet and human activities
In order to validate the performance of the proposed feature descriptor, we have computed the mean average precision for the proposed method
We observed that MAP (Mean average precision) of horse back riding is the highest (0.89)

Summary

Introduction

A large amount of human action video data has sharply increased due to the expansion of multimedia devices, the Internet and human activities. Similarity estimation of human action video data has been connected with various human activity applications such as similarity measure to recognize human activity [1] and similarity measure for matching correspondence [2,3]. Dancing games such as Let’s Dance with Mel B [4] and Just Dance [5]. Zhu et al [1] constructed a similarity model based on two different activity feature vectors using acceleration sensors of smart-phones to collect data, and compared feature similarities of extensive location to recognize human activities. Suppose we have one reference video on the database and one user performed

Objectives

Results

Discussion

Conclusion