Li Yi Associate Professor
Hongo Campus
Graduate SchoolGraduate School of Engineering - Electrical Engineering and Information Systems
Departmentー
Ubiquitous Information Environment Technology Field
Artificial Intelligence (AI)
Electron device/Electronic equipment
Measurement engineering
Deep Learning and Generative AI
System engineering
Development of a Next-Generation Sensing Platform Beyond Human Perception through the Fusion of Terahertz Waves and Artificial Intelligence
Terahertz (THz) waves possess unique properties of penetration and high resolution, enabling the acquisition of physical information — including see-through imaging, dielectric characteristics, and molecular-level dynamic changes — that is difficult to obtain with conventional visible-light cameras. In this research, we are working to establish a next-generation THz sensing platform toward practical implementation, through the integration of THz devices, measurement technologies, and physics-based models with artificial intelligence.
Research field 1
Integrated Terahertz Sensor Based on Resonant Tunneling Diodes (RTDs)
We will develop compact and integrated terahertz sensors based on resonant tunneling diodes (RTDs). By leveraging the oscillation and detection capabilities of RTDs, we aim to simplify conventional large-scale and complex THz measurement systems, thereby realizing practical on-chip THz sensing. Furthermore, through co-design of device characteristics and signal processing, we will advance the construction of a high-sensitivity, high-speed, and stable platform for physical property measurements.
Research field 2
Physical Property Extraction Based on Physics-Informed Neural Networks (PINNs)
We will develop a physical property extraction method based on Physics-Informed Neural Networks (PINNs) by embedding electromagnetic phenomena — including the propagation, scattering, and absorption of terahertz (THz) waves — into AI as physics-based models. By accurately estimating dielectric properties and related parameters from the amplitude and phase information contained in spatiotemporal THz signals, we aim to achieve stable and practical THz sensing analysis.
Research field 3
AI-Driven Information Fusion of THz and Visible-Light Data
We will develop a method for fusing, via AI, the physical information obtained from terahertz (THz) waves — including see-through imaging and dielectric properties — with the shape, color, and texture information contained in visible-light images. By complementarily integrating these heterogeneous sensor modalities, we aim to realize a multimodal sensing platform capable of highly reliable and high-accuracy recognition and measurement.