**1. Introduction**

In past decades, sedentary behaviour has appropriately received considerable attention in both developed and developing countries due to societal changes. People spend most of their time in sedentary activities, and their metabolic health is compromised due to low levels of energy expenditure (e.g., while sitting watching television, working on a computer in the workplace, using a cellphone, driving automobiles, playing video/board games, reading books and lying on the couch [1]). To address sedentary behaviour, some initial clarification is required about the terminology. We refer to all sitting activities in different contexts with an energy expenditure of ≤1.5 resting metabolic equivalents (METs) as sedentary behaviours [2]. Hence, a person is considered sedentary if s/he spends a large amount of the day in such activities. Sedentary behaviours are associated with chronic disease [3], physiological and psychological problems [4], cardiovascular disease, diabetes [5] and poor sleep [6]. The most

noticeable one is the high risk of being overweight and obesity, which have become serious public health threats worldwide and comprise the second leading cause of preventable death, trailing only tobacco [7]. Therefore, a self-management approach is required to support self-awareness and promote healthy behaviour to reduce the health risks caused by sedentary behaviour. The research community has suggested that new technologies like smartphone alerts of elapsed sedentary time and short breaks during prolonged sitting could be adopted in our daily routines [1,8].

Recently, activity trackers such as Fitbit [9], smartphone apps such as Google Fit [10] and smartwatch activity apps [11] can recognize many user activities. However, these trackers generate a time-series of user activities, but do not make the user aware of the detected unhealthy behaviour. This paper presents our model to detect the sedentary behaviour patterns and create a personal behaviour profile to store collected information. In the future, these profiles may assist practitioners to counsel the users or predict the future based on everyday rhythms of sedentary activity and past sedentary habits. Integrating smartphone technology has great potential to promote healthy behaviours [12]. Users do not have to wear/carry extra gear to monitor and track their daily routine behaviour. The smartphone has various embedded sensors (e.g., accelerometer, audio, WiFi, Global Positioning System (GPS), Bluetooth, gyroscope, magnetometer), high computational power and storage and programmable capabilities, along with wireless communication technologies [13]. Furthermore, it has become an integral part of our daily routines; and one of the best devices to recognize the user's context.

In order to mine the contexts of sedentary behaviour, there is a need to develop a ubiquitous system that can track the sedentary elapsed time accurately with all its minor routines ranging from office work to watching television during leisure time. Previously, we conducted a pilot study on micro-context recognition [14] and visualizing the user behaviour over the web through the Internet. In this paper, we propose a user-centric smartphone-based approach to recognize the context of sedentary behaviour based on the onboard accelerometers and audio sensors of the smartphone. We compute the acceleration and acoustic features over the collected sensory data streams and mine the contexts by applying the non-parametric nearest neighbour classification algorithm [15]. The main contribution of this paper is two-fold: (i) an initial implementation of the proposed framework supporting real-time context identification; (ii) testing and evaluation of the framework, which prove that our application is capable of substantially reducing sedentary behaviour and assisting humans to be active. The aim of the proposed framework is to monitor human sedentary behaviour in a proactive way. Based on the tracked behaviour, users will be able to monitor and manage their daily routines, which may help them to adopt active lifestyle.

The rest of the paper is structured as follows: Related work and the limitations of existing systems are discussed in Section 2. In Section 3, we provide the proposed architecture and its implementation inside the smartphone environment to track sedentary behaviour. In Section 4, we explain our experimental setup and present the obtained results. We provide a detailed discussion in Section 5 and interventions for our experimental study. Finally, the paper concludes with our findings and proposes future work in Section 6.
