Projects | Yi-Chao Chen

research

Circuit Intelligence -- LMM for PCB Design

A long-term research and translation direction for LMM-powered PCB design. We build multimodal PCB data, circuit-specific reasoning agents, and design-to-manufacturing validation loops that transform requirements, datasheets, IC footprints, schematics, layouts, BOMs, Gerbers, and production rules into structured, verifiable design assets.

Metasurface Sensing -- Programmable Physical Channels

A long-term research direction on using passive and adaptive metasurfaces as programmable physical channels. Instead of treating waves as fixed inputs to algorithms, this work co-designs surfaces, signals, optimization, and learning so commodity acoustic, RF, microwave, mmWave, GNSS, and wearable systems can sense, localize, communicate, image, and monitor health beyond their native hardware limits.

Unlocking Bluetooth Direction-Finding for All Devices -- Bridging the Gap

Bridge represents a pivotal step towards universal indoor localization. By elegantly solving compatibility issues without altering existing devices, Bridge democratizes the high-precision capabilities of BLE 5.1 direction finding, making advanced localization widely accessible. Its innovative nesting packets, precise AoA estimation via overhearing, and scalable deployment position it as a significant advancement in Bluetooth-based technologies. Bridge exemplifies a compelling vision where legacy devices seamlessly integrate into modern localization infrastructures, offering industries and consumers enhanced accuracy, reduced costs, and effortless upgrades to their localization capabilitie.

MIMSID -- Revolutionizing mmWave Imaging with AI and Metasurface Technology

MIMSID is a novel technique for high-resolution millimeter-wave (mmWave) imaging. It utilizes a compact, off-the-shelf mmWave module and a specially designed passive metasurface, eliminating the need for mechanical movement and offering a portable solution. The system tackles the limitations of current mmWave hardware, such as a restricted number of antennas and the reliance on compressive sensing algorithms that assume data sparsity. By integrating an optimized metasurface with a diffusion-based neural network (Sig2Img Diffusion), MIMSID directly transforms mmWave signals into high-quality images. This approach not only enhances image resolution but also provides a robust alternative to traditional methods. Extensive experiments demonstrate MIMSID's ability to achieve sub-centimeter resolution imaging, significantly outperforming existing techniques.

Whispering Spaces -- Create Sound Out of Thin Air for a New Reality

Imagine walking through a museum, and instead of looking down at a map, a voice seems to emanate from the direction of the exhibit you're seeking, guiding you with a simple "Follow me!". This isn't science fiction; it's the world of **acoustic augmented reality (AAR)**, and a new technology called **VISAR** is making it possible without headphones or special gadgets. By cleverly using the way sound behaves in air, VISAR can create **virtual sound spots**—focused areas of sound—that seem to appear out of nowhere, offering a more natural and immersive way to interact with our surroundings. This breakthrough paves the way for intuitive navigation, especially for the visually impaired, and opens up exciting possibilities for how we experience information and entertainment, transforming everyday spaces into interactive soundscapes.

Multi-directional Speaker Using Air Nonlinearity

We propose a novel multi-directional speaker using a parametric array to generate multiple highly directional sound beams. The system leverages air nonlinearity to reproduce audible sound from ultrasounds. It overcomes the challenges in traditional parametric arrays in terms of transducer size and focused wavefront, by using a special cell structure that connects ultrasonic transducers with optimal interspace to redirect an approximate omnidirectional wavefront. An optimization-based algorithm is developed to minimize unintended leakages. To further improve sound quality, we propose a nonlinear distortion reduction scheme.

Optical Image Stabilization -- Unlocking Hidden Potentials of Your Smartphone Camera

Optical Image Stabilization (OIS), initially designed to reduce camera shake and enhance photo clarity, can now revolutionize how we interact with everyday technology. By controlling OIS in innovative ways, we unlock capabilities such as precise depth sensing and image super-resolution, significantly enhancing smartphone photography and enabling new applications without additional hardware.

Nonlinearity of Spatial Frequency in Light and Its Applications

We model the nonlinear interaction between Color Filter Array and screens, and build applications inlcluding the novel optical encryption method for QR codes on top of our model.

Magic Finger Input System

We propose a training-free text input system called MagicInput that support multiple languages using acoustic-based 1D finger tracking technology. An artificial dataset (called TrackMNIST) simulated from MNISTseries datasets is utilized to satisfy the various users' writing characteristics.

Enabling Multi-priority Coding in LED-based Optical Camera Communications

We propose a novel hierarchical coding scheme named as OnionCode to support dynamic range of channel capacity in one-to-many OCC scenario.

User Identification based on Leakage Current

We developed a behavior-irrelevant user identification system applicable to laptops with a metal casing based on leakage current.

Handwriting Recognition System based on Vibration Signal

We presented a novel handwriting recognition system based on vibration signals detected by the built-in accelerometer of smart phones.

Enhancing QRCode Security by Fingerprinting Screens

In ScreenID, PWM frequency of screens is exploited as the unique screen fingerprint and used to enhance the security of a QR code by identifying its authenticity.

Utilizing Acoustic-based Imaging for Issuing Contact-Free Silent Speech Commands

A silent speech interface based on acoustic signals.

Near-filed communication via EM side-channel

A novel near-filed communication that utilizes EMI signals emitted from CPUs to transmit data and magnetometers on mobile devices to receive data.

MagThief

MagThief outlines a novel monitoring system to enable the theft of sensitive app usage information without the need for user permissions on the mobile devices.