Wearable Structured Mental-Sensing-Graph Measurement

Abstract

The emotional assessment under internet of things (IoT) architecture can support researchers to establish the relationships between human social and physiological signals and emotions. In this paper, a wearable emotion sensing system is developed under narrow band internet of things (NB-IoT) wireless communication technology. The wearable sensing device integrates social linked sensors including voice, activity, and heart rate. Using this system, a dating experiment is set up to investigate multimodal factors of perception. In particular, the multimodal data are fused in a graph structure, and this further leads to a graph convolutional neural networks model for emotion evaluation and a mentalsensing-graph intelligent interpreter. Different types of mentalsensing-graphs are fused during the training stage, and the model achieves a verification accuracy of 0.93. The intrinsic relationships among the multimodal data have been captured by the subgraphs which have star-shaped structures, and the center of the subgraphs are mostly audio node. The obtained results show that the attractiveness perception of the male participants in dating is more aligned to language communication. The results also reveal that when the male participants date highly attractive women during the experiments, a significant correlation is observed between the multimodal features and the attractiveness perception levels.

Automatic Depression Detection Among Higher Education Students Based on DeepFM

“…Detection and diagnosis of depression early is a crucial step for proper treatment [2]. Recently, automatic objective assessment methods to assist mental health detection and diagnosis have been widely explored [3] [4] [5] [6].…”