Publications

PhysioLabXR: A Software Platform for Real-Time Multi-Modal Brain-Computer Interfaces and Extended Reality Experiments

Ziheng Li*, Haowen Wei*, Ziwen Xie, Yunxiang Peng, June Pyo Suh, Steven Feiner, Paul Sajda

Journal of Open Source Software, September 2023.

PhysioLabXR is a Python-based open-source software platform for neuroscience and human-computer interaction (HCI) experiments, enabling real-time and multi-modal physiological data processing. The platform supports a variety of sensors including EEG, fNIRS, and eye trackers, while offering tools for multi-stream visualization, real-time digital signal processing (DSP), and experiment recording. With native support for popular data transfer protocols such as Lab Streaming Layer (LSL) and ZeroMQ (ZMQ), PhysioLabXR facilitates seamless integration and control over experimental pipelines. This tool serves as a foundation for future BCI and HCI experiments, significantly benefiting the research community.

Download Paper | Code | Video

Interactively Assisting Glaucoma Diagnosis with an Expert Knowledge-distilled Vision Transformer

Ziheng Li*, Haowen Wei*, Kuang Sun, David Li, Leyi Cui, Steven Feiner, Kaveri Thakoor

The ACM (Association of Computing Machinery) CHI conference on Human Factors in Computing Systems, 2025

This project enhances glaucoma diagnosis using an expert knowledge-distilled Vision Transformer, providing AI-augmented insights to ophthalmologists. The system integrates deep learning with medical imaging to focus on key diagnostic features in retinal images, interactively highlighting areas of interest for improved diagnosis. Validated with a user study involving 15 ophthalmologists, the tool demonstrates the potential of AI in supporting clinical decision-making and aiming for more accurate glaucoma diagnosis.

Download Paper | Video

Master’s Thesis: From Brain–Computer Interfaces to AI-Enhanced Diagnostics: Developing Cutting-Edge Tools for Medical and Interactive Technologies

Haowen Wei, Steven K. Feiner, Paul Sajda, Kaveri Thakoor

Columbia University

My master’s thesis advances brain-computer interfaces (BCI), human-computer interaction (HCI), and extended reality (XR) through three key projects. First, PhysioLabXR, an open-source Python platform, enables real-time, multi-modal BCI and XR experiments, streamlining data processing, visualization, and machine learning. Second, our work on Interactively Assisting Glaucoma Diagnosis employs deep learning to support clinical decision-making, aiming to introduce an AI-based diagnostic tool to CHI 2025. Lastly, the In Search for an Intuitive and Efficient Text-Entry in Mixed Reality project explores innovative text-entry methods in mixed reality for enhanced user interaction. Together, these projects push the boundaries of HCI and BCI research.

Download Paper

Efficient Text-Entry in Mixed Reality: Tap, Gaze & Pinch, SwEYEpe (CHI 2025 Late-Breaking Work)

Haowen Wei*, Ziheng Li*, Xichen He, Ben Yang, Steven Feiner

Preparing for CHI 2025 Late-Breaking Work

This project explores intuitive text-entry methods in mixed reality (MR), combining modalities like tapping, gaze, pinching, and swiping. It offers natural text-entry experiences through multi-modal interaction and user-centric design, aiming to enhance usability and efficiency in MR environments. The system’s effectiveness was assessed through user studies, demonstrating the potential of more intuitive text-entry solutions in MR.

IndexPen: Two-Finger Text Input with Millimeter-Wave Radar.

Wei, Haowen*, Ziheng Li*, Alexander D. Galvan, Zhuoran Su, Xiao Zhang, Kaveh Pahlavan, and Erin T. Solovey.

Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies 6, no. 2 (2022): 1-39.

In this paper, we introduce IndexPen introduces a novel interaction technique for touch-free text input using two-finger in-air micro-gestures. It employs millimeter-wave radar sensing to recognize 30 distinct gestures, corresponding to letters A-Z, along with Space, Backspace, Enter, and a special Activation gesture to avoid unintentional input. The system includes a noise class to distinguish gestures from noise. We detail our system design, RF processing pipeline, classification model, and real-time detection algorithms. With data collected from five participants over ten days, the system achieved 95.89% cross-validation accuracy across 31 classes. Further evaluation with 16 new participants demonstrated the system’s adaptability, with sentence typing accuracy reaching 86.2%.

Download Paper | Dataset | Video

Real-Time LiDAR Point-Cloud Moving Object Segmentation for Autonomous Driving.

Xie, Xing, Haowen Wei, and Yongjie Yang

Sensors 23, no. 1 (2023): 547.

In this paper, we propose a lightweight CNN architecture for LiDAR point-cloud moving object segmentation, targeting real-time autonomous driving applications. The network reduces the computational burden with 66% fewer parameters than the state-of-the-art and achieves real-time processing speeds on GPU and FPGA platforms. Our system achieves 51.3% IoU on the SemanticKITTI dataset and meets the real-time requirements of autonomous vehicles with 32 frames per second (fps) processing on FPGA.

Download Paper | Dataset

Proximity Detection During Epidemics: Direct UWB TOA Versus Machine Learning Based RSSI.

Zhuoran Su, Kaveh Pahlavan, Emmanuel Agu, Haowen Wei

International Journal of Wireless Information Networks, 29(4), 2022.

This study compares the proximity detection performance of UWB TOA and machine learning-based BLE RSSI during epidemics. The results show UWB TOA achieves slightly higher accuracy with less computational complexity, while BLE RSSI requires extensive training for similar results. The study evaluates both technologies in different environments and postures to assess their robustness for social distancing applications.

Download Paper

A Study of Interference Analysis Between mmWave Radars and IEEE 802.11 AD at 60 GHz Bands.

Shiyu Cheng, Kaveh Pahlavan, Haowen Wei, Zhuoran Su, Seyed Reza Zekavat, and Ali Abedi

International Journal of Wireless Information Networks, 29(3), 2022.

This study presents an empirical analysis of mutual interference between IEEE 802.11ad communication and millimeter-wave radar in the 60 GHz band. It examines the impact of interference on radar coverage and precision, and models the effect on packet loss rates in wireless communications.

Download Paper