Work

A Systematic Review of various Fusion Techniques for Human Activity Recognition

Year: 2024
Type: review
Abstract: Human Activity Recognition (HAR) has fetched considerable prominence as it plays a critical role in a wide number of applications ranging from healthcare monitoring to human-computer interaction. Gaining accuracy as well as efficiency in the process of representing and recognizing an activity is one of the critical goals in the domain. Apart from developing via technical perspective, utilizing the resources and technicalities in hand to the fullest is another significant criterion to gain accuracy and efficiency in any process. Embedding multiplicity in the sub-tasks via the fusion of multiple sources is one of the options to ensure that the resources enrolled in the task are being utilized effectively and to the fullest. In HAR, fusion could be considered from three perspectives, namely, data fusion, feature fusion and classifier fusion. In this paper, a survey of research work that implemented fusion from any of the three perspectives in the process of recognizing the activity has been generated. Apart from embedding multiplicity via fusion of each criterion on an individual basis, multiplicity could be embedded in the domain via the perspective of number of modes of fusion as well. The review also presents the work that implemented fusion via multiple criteria to optimize the process of recognizing the activity being executed. Section 1 generates an overview of technicalities in hand to represent and recognize the activity and justifies the criteria of embedding multiplicity in the process of activity recognition, Section 2 discusses three modes of fusion to embed multiplicity and gain both accuracy as well as efficiency in the process of HAR and Section 3 gives an overview of open research issues and finally Section 4 justifies the importance of the criteria of fusion.
Language: English
Open Access status: diamond
OpenAlex ID: W7216818092
FWCI: 0.02
Cites: 105
Cited by: 1
Related to: 10