Human activity recognition in cognitive environments using sequential ELM

Human activity recognition (HAR) and Extreme Learning Machines (ELM) are emerging fields of research. HAR investigates the behavioural attributes of humans and integrates that to an electronic system. An ELM is a fast learning algorithm, and overcomes the fundamental issue of slow training-error convergence that other algorithms such as the back propagation algorithm suffer. In this paper, we present the blend of the two fields by classifying the behavioural attributes of humans using Artificial Neural Networks (ANN) trained by Sequential Extreme Learning Algorithm (SELA). The algorithm is efficacious with a remarkable accuracy despite circumventing the vital job of pre-processing and feature extraction from signals that have been acquired from sensors.