Indoor falls often lead to fatalities due to delayed assistance. Current approaches to detecting indoor falls, such as cameras and wearables, intrude on privacy and are inconvenient. Radar-based device-free sensing has a limited range and requires dense deployment, leading to overhead costs. WiFi-based solutions, while promising, are currently either environment-dependent or insufficiently tested. In this work, we propose a fusion approach that leverages signal processing techniques to extract environment-independent features from the Channel State Information (CSI) in commercial WiFi devices. We then use a neural network to detect differentiating patterns from these features. Our lightweight LSTM network, with just 21,000 parameters, has been tested on 2,400 fall events from over 25 volunteers in 5 environments. It has also undergone 21 months of false alarm testing in 6 diverse settings. The system achieves a 94.1% detection rate and fewer than 5 false alarms per month in single-person homes.

Radio-frequency (RF)-based high-resolution human imaging is an emerging area of research fueled by the increasing availability of RF-radar devices. Even though existing works achieve accurate human body reconstruction for pose estimation purposes, human identification with imaging has not been feasible due to its limited resolution. In this work, we present high-resolution neural network (HRNet), a deep neural network based on conditional generative adversarial network architecture, to achieve high-resolution human silhouette images, which can be used for human identification. HRNet uses radar spatial spectrum generated using a modified multiple signal classification algorithm as input and is trained with Kinect images as ground truth. We tested our design using a commodity millimeter-wave radar device operating at 60 GHz. Experiments performed with 12 users in three different environments show that our proposed system can reconstruct human images with 4% mean silhouette difference when compared with Kinect images. Moreover, the system achieved an average classification accuracy of 90.6% for 12 users and 95.0% for seven users in unseen environments; thereby proving robustness to environment changes.

Recognizing in-air gestures can enable intelligent human–computer interaction (HCI) applications and facilitate human lives. However, existing sensor/camera-based methods for gesture recognition are either nonubiquitous, intrusive to privacy, or inconvenient to carry around. Contemporary device-free approaches require the person to be in the line of sight and proximity to the sensing device. This article shows that WiFi signals can recognize hand-drawn in-air gestures even when the gesture location is nonline-of-sight/beyond walls to the WiFi transceivers. The proposed GWrite system utilizes the channel state information (CSI) time-series information from commercial WiFi chipsets. GWrite employs a unique approach for performing hand gestures, thus enabling the design of a hand movement model. Using the model and the time-reversal (TR) technique, this work derives a correspondence between the similarity of CSIs and the relative distance moved by the hand. This relation gave rise to unique features, such as the number of segments, angle, and the intersection between segments that can classify a set of gesture shapes consisting of straight-line segments. GWrite achieved an accuracy of 92% on a group of 15 gestures. The proposed approach can be applied to a broader set of gestures, unlike the current systems that function over a limited gesture set.

Heat-stroke and death of children being left alone in a parked car has attracted more and more attentions. As a result, car manufactures start to reward solutions for in-car Child Presence Detection (CPD) system to save lives recently. However, most of the existing works rely on dedicated sensors and only achieve limited accuracy and coverage. This paper presents the first-of-its-kind intelligent CPD system using commodity Wi-Fi. Based on a statistical electromagnetic wave model to fully leverage the information in all the multi-path components, the proposed CPD system mainly consists of a motion target detector to detect a child in awake/motion status, a stationary target detector to detect a sleeping child by extracting breathing rate information, and a transition target detector based on a Naive Bayes Classifier using multipath profiles as features. We build a real-time testbed and show through extensive experiments that the proposed system can achieve ≥ 99.34% detection rate and ≤ 4.38% false alarm rate, regardless of the location and motion status of a child. Built upon 2.4/5GHz Wi-Fi, the proposed system can integrate with the existing in-car Wi-Fi system with no additional hardware and calls for low CPU and memory consumptions, thus promising a practical candidate for CPD applications.

Child presence detection (CPD) is becoming a regulatory requirement for car manufacturers to save children’s lives when they are left alone in unattended vehicles. However, most of the existing solutions require dedicated devices and suffer from limited accuracy and coverage. In this article, we build WiCPD, the first-of-its-kind in-car CPD system using commodity Wi-Fi, which can cover the entire interior of a car with no blind spot. First, we introduce a statistical electromagnetic model which accounts for the impact of motion on all the multipaths inside a car, followed by a motion statistics metric indicating the ambient motion intensity and a signal-to-noise-ratio (SNR) boosting scheme to extract the minute chest movement. Then, we design a unified CPD framework consisting of three target detector modules, including a motion target detector to detect a child in motion/awake, a stationary target detector to detect a stationary/sleeping child, and a transition target detector to detect a sleeping child with sporadic motion who is missed by both the motion and stationary target detectors. We implement a real-time WiCPD system by using commercial Wi-Fi chipsets, deploy it over 20 different cars, and collect data for multiple children aging from 4 to 50 months. The results show that WiCPD can achieve 100% detection rate within 8 s when the child is awake/in-motion and 96.56% detection rate within 20 s for a static/sleeping child. Extensive experiments also demonstrate that WiCPD can be easily deployed in minutes without calibration and enjoys very low CPU and memory consumption, thus promising a practical candidate for CPD applications.

Device-free (passive) gesture recognition offers an enormous potential to simplify Human-Computer Interaction (HCI) in future smart environments. WIFI-based gesture recognition approaches have attained acclaim amongst others due to the omnipresence, privacy-preservation, and broad coverage of WIFI. However, there is no universal solution built on off-the-shelf devices that can accommodate an expandable set of gestures in a through-the-wall setting. In this work, we propose such a gesture recognition system that can recover information about the actual trajectory of the hand movement allowing an expandable set of gestures. Further, we leverage the rich multipath in a through-the-wall setting to develop a statistical model for the channel variations induced by a hand gesture. This model is used to derive a correspondence between the relative distance moved by the hand and the Time Reversal Resonating Strength (TRRS) decay. Based on this relation and the geometry of the gesture shape, we design feature extraction modules to enable gesture classification. We built a prototype of the proposed system on off-the-shelf WIFI devices and achieved a classification accuracy of 87% on a set of 6 uppercase English alphabets.

In the era of pervasively connected and sensed Internet of Things, many of our interactions with machines have been shifted from conventional computer keyboards and mouses to hand gestures and writing in the air. While gesture recognition and handwriting recognition have been well studied, many new methods are being investigated to enable pervasive handwriting tracking. Most of the existing handwriting tracking systems either require cameras and handheld sensors or involve dedicated hardware restricting user convenience and the scale of usage. In this article, we present mmWrite, the first high-precision passive handwriting tracking system using a single commodity millimeter-wave (mmWave) radio. Leveraging the short wavelength and large bandwidth of 60-GHz signals and the radar-like capabilities enabled by the large phased array, mmWrite transforms any flat region into an interactive writing surface that supports handwriting tracking at millimeter accuracy. MmWrite employs an end-to-end pipeline of signal processing to enhance the range and spatial resolution limited by the hardware, boost the coverage, and suppress interference from backgrounds and irrelevant objects. We implement and evaluate mmWrite on a commodity 60-GHz device. The experimental results show that mmWrite can track a finger/pen with a median error of 2.8 mm and thus can reproduce handwritten characters as small as 1 cm × 1 cm, with a coverage of up to 8 m2supported. With minimal infrastructure needed, mmWrite promises ubiquitous handwriting tracking for new applications in the field of human-computer interactions.

Gesture recognition using wireless sensing opened a plethora of applications in the field of human-computer interaction. However, most existing works are not robust without requiring wearables or tedious training/calibration. In this work, we propose WiGRep, a time reversal based gesture recognition approach using Wi-Fi, which can recognize different gestures by counting the number of repeating gesture segments. Built upon the time reversal phenomenon in RF transmission, the Time Reversal Resonating Strength (TRRS) is used to detect repeating patterns in a gesture. A robust low-complexity algorithm is proposed to accommodate possible variations of gestures and indoor environments. The main advantages of WiGRep are that it is calibration-free and location and environment independent. Experiments performed in both line of sight and non-line-of-sight scenarios demonstrate a detection rate of 99.6% and 99.4%, respectively, for a fixed false alarm rate of 5%.

In the present evolving world, automobiles have become an intelligent electronic machine and are no longer a mere transport medium. In this article, we attempt to make them smarter by introducing the idea of in-car driver authentication using wireless sensing and develop a system that can recognize drivers automatically. The proposed system can recognize human identity by identifying the unique radio biometric information recorded in the channel state information (CSI) through multipath propagation. However, since the environmental information is also captured in the CSI, the performance of radio biometric recognition may be degraded by the changing environment. In this article, we first address the problem of “in-car changing environments” where the existing wireless sensing-based human identification system fails. We build a long-term driver radio biometric database consisting of radio biometrics of seven people collected over a period of two months. We leverage this database to create machine learning models that make the proposed system adaptive to new in-car environments. Second, we study the performance of the in-car driver authentication system with increasing effective bandwidth. We realize an effective bandwidth of 960 MHz by exploiting the multiantenna and frequency diversities in commercial WiFi devices. The performance of the proposed system is shown to improve with increasing effective bandwidth and the long-term experiments demonstrate the feasibility and accuracy of the proposed system. The accuracy achieved in the two-driver scenario is up to 99.13% for the best case.

Automobiles have become an essential part of everyday lives. In this work, we attempt to make them smarter by introducing the idea of in-car driver authentication using wireless sensing. Our aim is to develop a model which can recognize drivers automatically. Firstly, we address the problem of "changing in-car environments", where the existing wireless sensing based human identification system fails. To this end, we build the first in-car driver radio biometric dataset to understand the effect of changing environments on human radio biometrics. This dataset consists of radio biometrics of five people collected over a period of two months. We leverage this dataset-to create machine learning (ML) models that make the proposed system adaptive to new in-car environments. We obtained a maximum accuracy of 99.3% in classifying two drivers and 90.66% accuracy in validating a single driver.