machine and deep learning approaches for human activity recognition

Abstract

Human Activity Recognition (HAR) is a domain that has shown great interest in the past years and tills now. The main cause for this is that it can be used in various applications. There exist several devices and sensors that can capture and record activities. In this paper, a survey about the machine learning and deep learning methodologies in HAR is provided with information about the data, filtering methods, feature extraction methods, classification, and different performance measurements. The main aim is to present the machine learning and deep learning methodologies used in HAR. Therefore, the methods that showed the highest performance can be presented and investigated. In addition to this, the survey will cover the types of actions or activities that are predicted. Then, the results obtained from the survey are discussed to explore the most efficient methods in both machine and deep learning for the recognition of HAR. Moreover, the results involves illustrating whether the deep or machine learning methods is better in terms of data size, enhance performance, and number of activities to be recognized. Finally, the conclusions, limitations, and challenges in HAR are presented clearly.