Activity Level Assessment Using a Smart Cushion for People with a Sedentary Lifestyle

• Main Authors: Congcong Ma, Wenfeng Li, Raffaele Gravina, Jingjing Cao, Qimeng Li, Giancarlo Fortino
• Format: Article
• Language: English
• Published: MDPI AG 2017-10-01
• Series: Sensors
• Subjects: activity recognition; activity level assessment; smart cushion; body posture analysis model; activity assessment index
• Online Access: https://www.mdpi.com/1424-8220/17/10/2269

As a sedentary lifestyle leads to numerous health problems, it is important to keep constant motivation for a more active lifestyle. A large majority of the worldwide population, such as office workers, long journey vehicle drivers and wheelchair users, spends several hours every day in sedentary activities. The postures that sedentary lifestyle users assume during daily activities hide valuable information that can reveal their wellness and general health condition. Aiming at mining such underlying information, we developed a cushion-based system to assess their activity levels and recognize the activity from the information hidden in sitting postures. By placing the smart cushion on the chair, we can monitor users’ postures and body swings, using the sensors deployed in the cushion. Specifically, we construct a body posture analysis model to recognize sitting behaviors. In addition, we provided a smart cushion that effectively combine pressure and inertial sensors. Finally, we propose a method to assess the activity levels based on the evaluation of the activity assessment index (AAI) in time sliding windows. Activity level assessment can be used to provide statistical results in a defined period and deliver recommendation exercise to the users. For practical implications and actual significance of results, we selected wheelchair users among the participants to our experiments. Features in terms of standard deviation and approximate entropy were compared to recognize the activities and activity levels. The results showed that, using the novel designed smart cushion and the standard deviation features, we are able to achieve an accuracy of (>89%) for activity recognition and (>98%) for activity level recognition.