Intent Recognition Using Recurrent Neural Networks on Vital Sign Data: A Machine Learning Approach

Abstract

The growing importance of technology in daily life has led to a focus on making robots think like humans to enhance the integration of humans and robots in Cyber-Physical Systems (CPS). Cognitive science and psychology offer important knowledge and tools for integrating human-like learning processes into robots. The challenge is to enhance robots with prior knowledge and information, rather than starting the learning process from scratch. The goal of this research is to enable efficient interaction and co-existence of humans, robots, and other agents in CPS. This paper presents a review of the current academic literature on identifying human intentions and feeding robots for their effectiveness when interacting with humans. As a new contribution, this paper also proposes a state-of-the-art solution for human intent recognition studies and focuses our research roadmap on emotion recognition using Vital Signs including electroencephalography (EEG) data (signals) to understand the intent of human action using deep learning techniques. The research also compares the prediction performance of recurrent neural networks (RNN) with other algorithms. Understanding humans’ intent using vital signs for effective co-existence of humans in the cyber physical system and how to identify the intent of the agent and ensure that it aligns with the context of the given task or goal based on immediate perceptible visual attributes and dynamic properties (the perception of movement, gaze, vocalization, and emotional state.)