Human activity recognition is increasingly used for medical, surveillance and entertainment applications. For better monitoring, these applications require identification of detailed activity like sitting on chair/floor, brisk/slow walking, running, etc. This paper proposes a ubiquitous solution to detailed activity recognition through the use of smartphone sensors. Use of smartphones for activity recognition poses challenges such as device independence and various usage behavior in terms of where the smartphone is kept. Only a few works address one or more of these challenges. Consequently, in this paper, we present a detailed activity recognition framework for identifying both static and dynamic activities addressing the above-mentioned challenges. The framework supports cases where (i) dataset contains data from accelerometer; and the (ii) dataset contains data from both accelerometer and gyroscope sensor of smartphones. The framework forms an ensemble of the condition based classifiers to address the variance due to different hardware configuration and usage behavior in terms of where the smartphone is kept (right pants pocket, shirt pockets or right hand). The framework is implemented and tested on real data set collected from 10 users with five different device configurations. It is observed that, with our proposed approach, 94% recognition accuracy can be achieved.
15 November 2020 in Multimedia Tools and Applications
Human activity recognition (HAR) from sensory data is a crucial task for a wide variety of applications. The in-built inertial sensor facilities of commercial smartphones have made the data collection process easier. However, different smartphone configurations exhibit variations in sensor readings for the same activities. Different smartphone holding positions, like in hand, shirt, or trouser pockets, also lead to variations in signal patterns for the same activity. Some recent works have shown that automated feature extraction using deep learning methods can significantly improve activity recognition, although there is a lack of experimentation considering device heterogeneity and different smartphone holding positions. The proposed work addresses this research gap with a two-fold contribution. First, a CNN-based HAR framework is proposed that forms 2-D frequency domain images to capture temporal patterns in the data along with inter-axis spatial features. Second, an ensemble of conditional classifiers has been designed based on CNN that exhibits generality in terms of device configurations and usage behavior. Real life data have been collected for different activities using different devices for experimentation. The proposed ensemble model is found to recognize activities with 94% accuracy even when the training and test devices are different for real datasets.
27 November 2020 in Wireless Personal Communications
Wireless sensor networks consist of low power lightweight sensor nodes that are often placed remotely for applications like wildlife monitoring, rainforest monitoring, military surveillance, forest fire detection etc. Here energy is a critical issue as nodes cannot be recharged and/or replaced frequently. In this paper we propose an energy efficient hierarchical routing protocol for Wireless Sensor Networks to increase network lifetime. Here the cluster head is selected based on the residual energy, distance of the node from the base station, how recently the node was selected as cluster-head etc. Along with this, the proposed solution can also detect malicious nodes in the network and can prevent them from becoming cluster heads. The simulation results show that proposed solution provides good performance in terms of extending network lifetime, uniform selection of nodes as cluster head that eliminates the probability of a single node from draining away its power by its repetitive selection as a cluster head.