Memristor-Based Hierarchical Attention Network for Multimodal Affective Computing in Mental Health Monitoring

Abstract

We present a circuit design of the hierarchical attention network for multimodal affective computing, which can be used in mental health monitoring. Specifically, a kind of cost-effective memristor is fabricated using the albumen protein, and the corresponding testing performance is conducted to ensure its efficiency and stability. Then, considering the hierarchical mechanism inspired by the human limbic system, the nanoscale memristors arranged in a crossbar array configuration are further applied to construct a compact hierarchical attention network that can perform the multimodal affective computing. Furthermore, based on the wearable technology and flexible electronics technology, a mental health monitoring system with low privacy invasiveness, low energy consumption, and low fabrication cost can be designed. Based on the mapping relationship between the multimodal affective computing and mental health, the mental health state of the current user can be monitored. This study is expected to help achieving the deep integration of neuromorphic electronics and mental health monitoring system, further promoting the development of next-generation consumer healthcare technology in smart city.