Signals and Inference Research Group (SiNRG)

Research Areas: Generative models | Inverse problems | NeRFs and wireless imaging | Blackbox optimization

Our group is broadly interested in machine learning and generative models, with a focus on diffusion models, flow matching, and energy models. Under this broad umbrella, we are learning and working on topics like: (1) various inverse problems in which priors are learnt by diffusion models; (2) neural radiance fields (NeRF) and their generalization to other modalities beyond cameras; (3) composition of vision language models for higher robustness and semantic alignment; (4) blackbox optimization on diffusion models using human feedback.

Ongoing Projects:

EchoNeRF: Can NeRFs "see" without lights and cameras, i.e., wireless imaging of environments.
ArrayDPS: Unsupervised audio source separation, dereverberation, and denoising via posterior sampling.
InPose: 3D human pose estimation via sparse sensing and other generative motion tracking projects.
SPIN: Spatial inverse problems where the forward operators are non-differentiable path-planners.
CoCube: Composing concepts in diffusion models through Langevin correction.
XBO: Optimizing for human satisfaction through blackbox optimization.
Earable Computing: Sensing, computing, and communication on a AI-enabled platform.

In the past, our group was active in wireless networking, wireless and audio-based sensing, and applied signal processing.

For MS/PhD applicants: We always have room for strong students in our group. Ideal candidates are those with a balanced background in AI, where "balance" in our group is roughly defined as 60% mathematical and 40% programming.

Publications: ICML | NeurIPS | ICASSP | AAAI . . .

Zero-shot Human Pose Estimation using Diffusion-based Inverse solvers website
Preprint, Sep 2025.

Contrastive Diffusion Guidance for Spatial Inverse Problems website
Preprint, Sep 2025.

CO3: Contrasting Concepts Compose Better website
Preprint, Sep 2025.

Can NeRFs "See" without Cameras? website
Chaitanya Amballa, Sattwik Basu, Yu-Lin Wei, Z. Yang, M. Ergezer, Romit Roy Choudhury
NeurIPS, Dec 2025.

ArrayDPS: Unsupervised Blind Source Separation with a Diffusion Prior website demo
Zhongweiyang Xu, Xulin Fan, Zhong-Qiu Wang, Xilin Jiang, Romit Roy Choudhury
ICML, Jul 2025.

Estimating Multi-chirp Parameters using Curvature-guided Langevin Monte Carlo
Sattwik Basu, Debottam Dutta, Yu-Lin Wei, Romit Roy Choudhury
IEEE ICASSP, Apr 2025.

Kernel Learning for Sample Constrained Black-Box Optimization website demo
Rajalaxmi Rajagopalan, Yu-Lin Wei, Romit Roy Choudhury
AAAI, Feb 2025.

Unsupervised Multi-channel Speech Dereverberation via Diffusion website demo
Yulun Wu, Zhongweiyang Xu, Jianchong Chen, Zhong-Qiu Wang, Romit Roy Choudhury
WASPAA, Oct 2025.

Multi-Source Music Generation with Latent Diffusion website demo
Zhongweiyang Xu, Debottam Dutta, Yu-Lin Wei, Romit Roy Choudhury
NeurIPS: Audio Imagination workshop, Dec 2024.

Sample-Constrained Black Box Optimization for Audio Personalization website demo
Rajalaxmi Rajagopalan, Yu-Lin Wei, Romit Roy Choudhury
AAAI, Feb 2024.

Multi-Channel Speech Enhancement For SPEAR Challenge: A Three Stage Approach
Zhongweiyang Xu, Debottam Dutta, Xulin Fan, Mark Hasegawa-Johnson, Romit Roy Choudhury
SPEAR Challenge (2nd Rank), 2023.

Audio Personalization through Human-in-the-loop Optimization
Rajalaxmi Rajagopalan, Yu-Lin Wei, Romit Roy Choudhury
NeurIPS Workshop on ML for Audio, Dec 2023.

Self-Supervised Speech Denoising with Multi-Modal Data
Yu-Lin Wei, Rajalaxmi Rajagopalan, Bashima Islam, Romit Roy Choudhury
NeurIPS Workshop on ML for Audio, Dec 2023.

RoSS: Rotation-induced Aliasing for Audio Source Separation website demo
Hyungjoo Seo, Sahil Bhandary Karnoor, Romit Roy Choudhury
ICRA, Jan 2023.

Learning to Separate Voices by Spatial Regions website demo
Zhongweiyang Xu, Romit Roy Choudhury
ICML, Jul 2022.

PoseKernelLifter: Metric Lifting of 3D Human Pose Using Sound
Zhijian Yang, Xiaoran Fan, Volkan Isler, Hyun Soo Park
IEEE/CVF CVPR, Jun 2022.

Inferring Facing Direction from Voice Signals
Yu-Lin Wei, Rui Li, Abhinav Mehrotra, Romit Roy Choudhury, Nic Lane
arXiv, Sep 2021.

Personalizing Head Related Transfer Functions for Earables slides
Zhijian Yang, Romit Roy Choudhury
ACM SIGCOMM, Aug 2021.

Angle-of-arrival (AOA) Factorization in Multipath Channels
Yu-Lin Wei, Romit Roy Choudhury
IEEE ICASSP, June 2021.

Earable Computing: A New Area to Think About website
Romit Roy Choudhury
ACM HotMobile, Feb 2021.

Voice Localization Using Nearby Wall Reflections slides
Sheng Shen, Daguan Chen, Yu-Lin Wei, Zhijian Yang, Romit Roy Choudhury
ACM MobiCom, September 2020.

Ear-AR: Indoor Acoustic Augmented Reality on Earphones slides
Zhijian Yang, Yu-Lin Wei, Sheng Shen, Romit Roy Choudhury
ACM MobiCom, September 2020.

EarSense: Earphones as a Teeth Activity Sensor slides
Jay Prakash, Zhijian Yang, Yu-Lin Wei, Haitham Hassanieh, Romit Roy Choudhury
ACM MobiCom, September 2020.

Many-to-Many Beam Alignment in Millimeter Wave Networks slides
Suraj Jog, Jiaming Wang, Junfeng Guan, Thomas Moon, Haitham Hassanieh, Romit Roy Choudhury
NSDI, February 2019.

Ghostbuster: Detecting the Presence of Hidden Eavesdroppers slides
Anadi Chaman, Jiaming Wang, Jiachen Sun, Haitham Hassanieh, Romit Roy Choudhury
ACM MobiCom, October 2018.

Closing the Gaps in Inertial Motion Tracking slides
Sheng Shen, Mahanth Gowda, Romit Roy Choudhury
ACM MobiCom, October 2018.

MUTE: Bringing IoT to Noise Cancellation slides video
Sheng Shen, Nirupam Roy, Junfeng Guan, Haitham Hassanieh, Romit Roy Choudhury
ACM SIGCOMM, August 2018.

LiquID: A Wireless Liquid IDentifier slides
Ashutosh Dhekne, Mahanth Gowda, Yixuan Zhao, Haitham Hassanieh, Romit Roy Choudhury
ACM MobiSys, June 2018.

Inaudible Voice Commands: The Long-Range Attack and Defense slides video
Nirupam Roy, Sheng Shen, Haitham Hassanieh, Romit Roy Choudhury
NSDI, April 2018.

If WiFi APs Could Move: A Measurement Study
Ashutosh Dhekne, Mahanth Gowda, Romit Roy Choudhury, Srihari Nelakuditi
IEEE Transactions on Mobile Computing, 2018.

Technical perspective: Is your WiFi a sensor?
Romit Roy Choudhury
Communications of the ACM, 2018.

BackDoor: Making Microphones Hear Inaudible Sounds slides video
Nirupam Roy, Haitham Hassanieh, Romit Roy Choudhury
ACM MobiSys, June 2017.
Best Paper Award

Bringing IoT to Sports Analytics slides
Mahanth Gowda, Ashutosh Dhekne, Sheng Shen, Romit Roy Choudhury, Xue Yang, Lei Yang, A. Essanian, S. Golwalkar
NSDI, March 2017.

Extending Cell Tower Coverage through Drones slides
Ashutosh Dhekne, Mahanth Gowda, Romit Roy Choudhury
ACM HotMobile, February 2017.