Using Machine Learning for Person Identification through Physical Activities

In this paper, the concept of utilizing machine learning algorithms for person identification through physical activity is proposed. Many previous machine learning research articles focused on building models to identify physical activities using a sensor fusion input. Nevertheless, there has been no focus on building models that can identify the activity performer as well. This paper will demonstrate that machine learning can be applied not only for the identification of physical activities but also for the identification of the activity performer as well. The paper will present the achieved accuracies for the person identification through physical activities using different machine learning algorithms. Additionally, a novel multi-label shared deep neural network (DNN) is proposed for identifying both the physical activity and the activity performer simultaneously. The proposed design allows for a single training/re-training which is advantageous over having to train two separate DNNs. Moreover, it is 30% smaller compared to a design that consists of two separate DNNs for identifying the physical activity and the activity performer.