Importance of Activity and Emotion Detection in the Field of Ambient Assisted Living

Technological advancement improves the standard of living for both individuals and society as a whole. But this improvement should also take care of physically challenged and elderly people in our society, which is our foremost responsibility as well. Ambient Assisted Living (AAL) is a new chapter of technological advancement that aims to extend its support for this indigent site of the society. Implementation of AAL requires ambient intelligence sensors that monitor the status of human being and surrounding environment and analyze the data further with AI techniques. Based on the analysis result, the system extends support to the person or sends urgent messages in case of emergency to the concerned persons or health care unit. Assistance for the human being may be necessary either because the person is physically weak to perform daily activities independently due to the aging effect or because the person is mentally unwilling to perform necessary work due to several disturbances from the surroundings. Under both of these circumstances, it is really important to analyze the human's physical and mental status automatically to extend the necessary support as and when required. There are many state-of-the-art methodologies available to diagnose human health condition, but it is really a challenging task to identify the human physical and mental status based on daily behavioral activities and hence provide the necessary support. A number of research works have been performed on this emerging technology for the last two decades to improve living quality. This study aims to summarize different techniques and their outcomes involved in this field of research and also describe an initiative to develop an Inertial Measurement Unit (IMU)-based hardware system for the detection of activities of daily leaving (ADL). The developed hardware model collects acceleration data from a wrist worn tri-axial accelerometer, sends the data to a computer using Wi-Fi protocol. A set of statistical features are extracted from the data after pre- processing it for elimination of noise. This feature set is tested with various machine learning algorithms, among which Decision Tree based method is found to be most effective in detection of ADL. The developed model ensures compatible activity detection accuracy with other state-of-the-art works utilizing deep learning models with multimodal inputs.