Recently, Human Activity Recognition (HAR) has been a popular research field due to wide spread of sensor devices. Embedded sensors in smartwatch and smartphone enabled applications to use sensors in activity recognition with challenges for example, support of elderly’s daily life . In the aim of recognizing and analyzing human activity many approaches have been implemented in researches. Most articles published on human activity recognition used a multi -sensors based methods where a number of sensors were tied on different positions on a human body which are not suitable for many users. Currently, a smartphone and smart watch device combine different types of sensors which present a new area for analysis of human attitude. This paper presents a review on methodologies applied to solve problems related to human activity recognition that use the equipped sensors in smartphone and smartwatch with the employ of Machine Learning and the advance of deep learning approaches. The literature is summarized from four aspects: sensors types, applications, Machine Learning (ML) and Deep Learning (DL) models, results and challenges.
The wide use of smartphones and later smartwatches equipped with a set of sensors such as location, motion, and direction blaze the trail for researchers to better recognize human activity. However, researches on using inertial or motion sensors (i.e., accelerometer, gyroscope) for human activity recognition (HAR) has intensified and reside a great confrontation to be faced. Lately, many deep learning methods have been suggested to improve the human activity classification and discrimination performance to reach an optimal accuracy. Therefore, this paper applies a Convolutional eXtreme Gradient Boosting (ConvXGBoost), which combines Convolutional Neural Network (CNN) represented by AlexNet to learn the input features automatically, followed by XGBoost decision tree used to predict the class label and thereof recognize the performed activity. Human activities are collected from sensors as time series data. Therefore, we suggested using one-dimensional AlexNet (1D AlexNet) model instead of 2D. The AlexNet model is compiled with two optimizers Adam and Stochastic Gradient Descent (SGD) which are applied consecutively. The suggested architecture was trained and evaluated on the “WISDM Smartphone and Smartwatch Activity and Biometric Dataset” that consists of raw data for eighteen activities recorded from phone and watch. The experiments revealed that using multi optimizer with a convolutional neural network improved the accuracy of recognition by 5%. Moreover, a proposed ConvXGBoost model outperformed the performance of other models works with the dataset as mentioned above with an overall accuracy of 98-99% depends on the device used.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.