Key points are not available for this paper at this time.
The detection of daily human activities is a key component in modern applications of Internet of Things. In this study, we introduce a hierarchical algorithm for online human activity detection with two levels of feature extraction methods. In the lower level, the algorithm gets sensor data from accelerometer and microphone of user smartphone and extracts the models about the Motion and Environment Detection of user. In the higher level, the algorithm takes as input the combination of the output from these models and extracts the model about the Human Activity Detection. This flexible and modular hierarchical algorithm detects more complex activities (than usually in the state of the art), under broader smartphone configurations (position, orientation). It can be extended with more feature extraction models for different sensors, in additional levels of hierarchy and with different combinations in order to recognise with higher accuracy more specific and sophisticated user activities.
Filios et al. (Tue,) studied this question.