Article Preview
TopIntroduction
Embedded Sensors are ubiquitous and are becoming sophisticated by nature. This has been changing people’s daily life and has opened the doors for many interesting data mining applications. Human activity recognition (HAR) is a research domain behind many applications on smartphone such as health monitoring, fall detection, context-aware mobile applications, human survey system, home automation, etc. The HAR systems consist to identify the actions being carried out by a person given a set of observations of him/herself and the surrounding environment. Recognition can be accomplished by exploiting the information retrieved from various sources such as environmental sensors (Fahim, et al., 2013), body-worn sensors (Helbostad, et al., 2017; Liang et al., 2018) or the smartphone sensors (Shoaib et al., 2015; Siirtola et al., 2012; Bayat et al., 2014).
Recent developments in sensing technology have led to wireless sensor networks which provide a non-intrusive, privacy friendly and easy to install solution to in-home monitoring (Van Kasteren et al., 2010). Sensors used are generally contact switches to measure open-close states of doors and cupboards; pressure mats to measure sitting on couch or lying in bed; mercury contacts for movement of objects such as drawers; passive infrared sensors to detect motion in a specific area and float sensors to measure the toilet being flushed.
Automatic activity recognition systems aim to capture the state of the user and its environment by exploiting heterogeneous sensors attached to the subject’s body.The such systems can monitor and keep track of the activities of daily living (ADL), learn from them and assist us in making decisions. Such assistive technologies can be of immense use for remote health care (Manirabona et al., 2018; Nimkar et al., 2019), for elderly people, the disabled, and those with special needs.
Human physical activities range from simple full body motor activities like walking, sitting and standing to complex motor activities such as jogging and climbing, the recognition of which plays an important role in many applications such as human-computer interaction and surveillance (Anguita et al., 2013; Liang et al., 2018). Performance in these activities can also be important indicators for patients recovering from newly acquired disability or people who are at risk of decline, due to aging factor. For e.g. a person staying away from his/her elderly parent can monitor their daily activities by providing an alarm if there is a change in the regular pattern or an early alarm of a health care emergency.
Some approaches have adapted dedicated motion sensors in different parts of the body, such as the waist, wrist, chest and thighs achieving good classification performance (Shoaib et al., 2014). These sensors are usually un-comfortable for the common user and do not provide a long-term solution for activity monitoring (e.g. sensor repositioning after dressing (Bao et al., 2004). Activity recognition using wearable body motion sensors has attracted more interests for decades (Lara et al., 2013).