Through-Wall Human Pose Estimation Using Radio Signals

Abstract

The figure shows a test example with a single person. It demonstrates that our system tracks the pose as the person enters the room and even when he is fully occluded behind the wall. Top: Images captured by a camera colocated with the radio sensor, and presented here for visual reference. Middle: Keypoint confidence maps extracted from RF signals alone, without any visual input. Bottom: Skeleton parsed from keypoint confidence maps showing that we can use RF signals to estimate the human pose even in the presence of full occlusion.

The Parkinson's disease e‐diary: Developing a clinical and research tool for the digital age

“…Visual analog scales (VASs) are ideal to use for nondichotomous questions in active data collection, can be used by patients with cognitive impairment, and are very sensitive to small intrapersonal changes. 16 20,21 Active data collection should be tailored for motor and nonmotor symptoms. Examples for motor symptoms include spiral drawing, finger tapping, and voice characteristics and for nonmotor symptoms, assessments of visual performance and short-term memory.…”

“…An e‐Diary/tracker would allow tools such as surveys and VASs to be administered regardless of time or place. Advanced hardware components, such as accelerometers, gyroscopes, microphones, radio signals, among other wearable sensors, can provide complementary action‐dependent and action‐independent objective measures . Active data collection should be tailored for motor and nonmotor symptoms.…”

ASR is All You Need: Cross-Modal Distillation for Lip Reading

“…There have been many interesting variations on this idea, such as using a visual recognition network (trained on RGB images) as a teacher for student networks which take depth or optical flow [25], or audio [26] as inputs. More specific examples include using the output of a pre-trained face emotion classifier to train a student network that can recognize emotions in speech [27] or visual recognition of human pose to train a network to recognize pose from radio signals [28]. The closest work to ours is Wei et al [29] who apply cross-modal distillation from ASR for learning audio-visual speech recognition.…”

Person-in-WiFi: Fine-Grained Person Perception Using WiFi

“…These approaches require a minimal spatial resolution of sensor outputs. For instance, 300 × 300 pixels images from cameras [28], depth resolution around 2 cm for radars [55], or 32-beam LiDARs [50,31]. Moreover, camera-based solutions are limited by technical challenges such as clothing, background, lighting and occlusion, and social limitations such as privacy concerns.…”

“…Moreover, camera-based solutions are limited by technical challenges such as clothing, background, lighting and occlusion, and social limitations such as privacy concerns. Radar sensors require dedicated hardware, e.g., RF-Pose [55] and RF-Capture [1] used the Frequency Modulated Continuous Wave (FMCW) technology to produce depth maps, requiring carefully assembled and synchronized 16 + 4 T-shaped antenna array with a broad signal bandwidth (1.78 GHz). High-definition LiDAR sensors are very expensive and power-consuming, therefore are difficult to apply for daily and household use.…”