PocketVib, the first cost-effective smartphone-based laser speckle vibrometry system for multi-point kHz-frequency micro-vibration sensing, using built-in ToF and rolling shutter cameras.
CUHK-X, a large-scale multimodal dataset and benchmarks for human activity recognition, understanding, and reasoning, with tens of thousands of samples of 40 actions from 30 participants across seven sensor modalities.
HoloMobile, the first end-to-end mobile avatar system that reconstructs photorealistic dynamic human avatars from monocular smartphone video and streams them for real-time viewing on mobile devices.
EventEye, the first system that integrates event cameras into roadside infrastructure to enhance a vehicle’s real-time perception by fusing high-frequency event data with the vehicle’s LiDAR.
CoHear, a collaborative system that leverages a network of earphones to model and enhance speech at the conversation level, using verbal and non-verbal cues to form groups and extract target speech.
ContextAgent, the first context-aware proactive agent that incorporates wearable sensory contexts so LLM agents can predict when help is needed and call tools unobtrusively.
AquaScan, the first scanning sonar-based underwater sensing system for human activity monitoring. To
overcome the low frame rate due to the sonar’s physical
limitation, we propose a novel scanning strategy and apply
an image reconstruction method to accelerate the scanning
speed without compromising the performance of motion
detection
MyoTrainer is the first vision-based muscle-aware motion feedback system that uses explicit muscle-aware motion analysis and domain-specific expert knowledge to provide corrective guidance on muscle engagement and movement execution.
SmartLiDAR, an advanced LiDAR system that enhances scanning efficiency and performance by adaptively optimizing scan focus through an intelligent, softwaredefined micro-mirror controller. Unlike traditional systems, SmartLiDAR dynamically adjusts its scan pattern based on environmental characteristics and application-specific requirements, concentrating sample points on key objects without increasing power consumption or scan time.
SCX, a stateless KV-cache encoding framework for confidential Transformer serving that encodes the intermediate key-value cache with user-controlled keys, so the cloud can neither recover the input nor complete next-token prediction on its own.