Method for Recognition of the Physical Activity of Human Being Using a Wearable Accelerometer

Abstract

Companies are interested in retaining workers healthy, productive, and satisfied while cutting health-care and insurance costs. Using a computer at work can cause back, neck and shoulder pain, eyestrain, and overuse injuries of human hands and wrists. It is possible to reduce these risks with better posture and good habits, such as taking rest breaks. During these breaks computer users should be encouraged to stand, stretch, and move around. For people who forget about a break or truly are focused on their direct work need help from special equipment for evaluation of real physical activity of computer user. Method for recording accelerometer data from moving human as he or she performs daily activities and for identification of type, duration and intensity of movements by using wearable wireless sensing system is presented in this paper. The extraction of orientation independent acceleration data has positive effect on recognition accuracy of k-nearest neighbour classification scheme used for classification task. The recognition accuracy of algorithm is 78.9% and these results are better than accuracy obtained from raw accelerometer data. The method presented is simple, exhibited good performance and does not require significant computational recourses.

Using Accelerometer and GPS Data for Real-Life Physical Activity Type Detection

“…Although the 3D accelerometer is the most common and informative sensor for PA type detection, it is challenging to accurately detect real-life activity types using only a single 3D accelerometer [5][6][7]. Researchers have extensively examined the usefulness of complementing accelerometer-based PA measures with additional sensors such as gyroscope, magnetometer, barometer and heart rate [8][9][10][11] or using multiple accelerometer devices on different body locations to improve the activity recognition [5,12]. However, these solutions entail mounting more devices on a person's body or rendering data analysis more complex due to dealing with different sensors featuring different data formats and sampling rates.…”

A survey on wearable sensor modality centred human activity recognition in health care

“…The development of wearable devices, such as smart watches, smartphones, wristbands, smart clothes, makes it feasible to acquire data from the ubiquitous equipment and provide continuous monitoring of human activities (Adaskevicius, 2014, Filippoupolitis, et al, 2017, Hassan, et al, 2018. Data-driven-based WSHAR systems share basically a similar procedure, as shown in Fig.2.…”

“…Smart clothes can embed more sensors, especially physical sensors, to achieve a diverse function compared with smartphones or smart watches, especially for long term monitoring applications (Adaskevicius, 2014). For instance, smart shirts are designed to monitor precise cardiac, respiratory, sleep and other daily activities, which incorporate heart rate and ECG sensors (Hexoshin, 2018).…”

“…Limited sensor types for different applications Smart clothes Adaskevicius, 2014Hexoshin, 2018Lorussi, et al, 2016 More sensors embedded, long term monitoring, the relative movement between the body parts and the sensors, etc.…”

“…WSHAR can function in a relatively large space while the wearer is moving. Currently, smartphone, smartwatch, smart clothes, and other specifically-designed devices are the mainstream products embedded wearable technologies in HAR (Hassan et al, 2018, Filippoupolitis et al, 2017, Adaskevicius, 2014. Generally, placing more sensors on multiple body parts (e.g., head, wrists, waist, legs, feet) can benefit improving the performance and robustness of WSHAR (Laudanski, et al, 2015, Gao, et al, 2014, Chernbumroong, et al, 2014.…”