The recent advancement and development of computer electronic devices has led to the adoption of smart home sensing systems, stimulating the demand for associated products and services. Accordingly, the increasingly large amount of data calls the machine learning (ML) field for automatic recognition of human behaviour. In this work, different deep learning (DL) models that learn to classify human activities were proposed. In particular, the long short-term memory (LSTM) was applied for modelling spatio-temporal sequences acquired by smart home sensors. Experimental results performed on the Center for Advanced Studies in Adaptive Systems datasets show that the proposed LSTM-based approaches outperform existing DL and ML methods, giving superior results compared to the existing literature.

Introduction
In the last few decades, human activity recognition (HAR) has been a lively and challenging research area, due to its applicability to different active and assisted living (AAL) domains, as well as the increasing demand for home automation and convenience services for the elderly [1]. Nowadays, mainly because of the rapid increase in the world’s ageing population [2], HAR has acquired much interest in the field of ambient intelligence and assisted living technologies in smart homes. It is meant to improve the residents’ quality of life with the use of simple and ubiquitous sensors [3]. According to [4], a smart home provides independence and comfort to the residents by using all technological devices interconnected within the network, capable of communicating and learning through the user’s habits, creating an interactive space. In particular, HAR is the most salient process for incorporating ambient intelligence into smart environments. It involves a series of complex modelling, reasoning, and decision-making procedures [5], [6]. The goal of HAR is to detect and then identify simple and complex human activities in real-world settings by processing spatial and temporal information acquired by visual and non-visual sensory data [6], [7]. The adopted sensors may be fused in the environment, connected with its objects, or worn directly by the occupant. Compared to wearable sensors, object or environment sensors are advantageous, as they can give an indirect indication of the occupant’s activities; moreover, they can discriminate similar actions [3], [8]. According to [7], [9], HAR application domains are among the most varied, but they can be enclosed in three macro categories: health-care monitoring applications [10], monitoring and surveillance systems for indoor and outdoor activities [11], and lastly, AAL systems [12] for smart homes.

Read all the article >