The past decades have witnessed a surge in human fall detection with sensors, cameras, and wireless signals. Among them, Wi-Fi-based fall detection has been one of the most attractive solutions due to the ubiquitous and pervasive deployment of Wi-Fi infrastructures. However, these approaches are still difficult to be put into practical use. To push forward Wi-Fi-based fall detection for wide deployment, three major limitations concerningenvironmental diversity,motion diversity, anduser diversityare required to be resolved. In this paper, we propose FallDar, a Wi-Fi-based deep learning-assisted fall detection system that outperforms state-of-the-art works on the three criteria simultaneously. First, to deal with environmental diversity, FallDar characterizes falls with the speed of the body, which is the most relevant and inherent feature of falling activities, making the system resilient to environmental changes. Second, to deal with motion diversity, FallDar simulates a large amount of fall data of various falling types with a DNN-based generative model. Training with these data, FallDar is endowed the capability of detecting more types of falls. Third, to deal with user diversity, FallDar proposes to incorporate the fall detection network with a user identification network. The network is designed to extract user-independent features, requiring no fall data from new users for system adjustment. We implement FallDar on commercial Wi-Fi devices and conduct experiments in home and office environments for six months. The evaluation results show that FallDar achieves a false alarm rate of 5.7% and a missed alarm rate of 3.4% across all factors, making a fundamental step towards ubiquitous fall detection with Wi-Fi.

Existing device-free localization systems have achieved centimeter-level accuracy and show their potential in a wide range of applications. However, today’s radio-based solutions fail to locate the target in millimeter-level due to their limited bandwidth and sampling rate, which constrains their applications in high-accuracy demand scenarios. We find an opportunity to break the bottleneck of existing radio-based localization systems by reconstructing the accurate signal spectral peak from the discrete samples, without changing either the bandwidth or the sampling rate of the radio hardware. This study proposes milliLoc , a millimeter-level radio-based localization system. We first derive a spectral peak reconstruction algorithm to reduce the ranging error from the previous centimeter-level to millimeter-level. Then, we improve the AoA measurement accuracy by leveraging the signal amplitude information. To ensure the practicality of milliLoc , we further extend our system to handle multi-target situations. We fully implement milliLoc on a commercial mmWave radar. Experiments show that milliLoc achieves a median ranging accuracy of 5.5 mm and decreases the AoA measurement error by 31.2% compared with the baseline. Our system fulfills the accuracy requirements of most application scenarios and can be easily integrated with other existing solutions, shedding light on high-accuracy location-based applications.

With the development of signal processing technology, the ubiquitous Wi-Fi devices open an unprecedented opportunity to solve the challenging human gesture recognition problem by learning motion representations from wireless signals. Wi-Fi-based gesture recognition systems, although yield good performance on specific data domains, are still practically difficult to be used without explicit adaptation efforts to new domains. Various pioneering approaches have been proposed to resolve this contradiction but extra training efforts are still necessary for either data collection or model re-training when new data domains appear. To advance cross-domain recognition and achieve fully zero-effort recognition, we propose Widar3.0, a Wi-Fi-based zero-effort cross-domain gesture recognition system. The key insight of Widar3.0 is to derive and extract domain-independent features of human gestures at the lower signal level, which represent unique kinetic characteristics of gestures and are irrespective of domains. On this basis, we develop a one-fits-all general model that requires only one-time training but can adapt to different data domains. Experiments on various domain factors (i.e. environments, locations, and orientations of persons) demonstrate the accuracy of 92.7% for in-domain recognition and 82.6%-92.4% for cross-domain recognition without model re-training, outperforming the state-of-the-art solutions.

Gait, the walking manner of a person, has been perceived as a physical and behavioral trait for human identification. Compared with cameras and wearable sensors, Wi-Fi-based gait recognition is more attractive because Wi-Fi infrastructure is almost available everywhere and is able to sense passively without the requirement of on-body devices. However, existing Wi-Fi sensing approaches impose strong assumptions of fixed user walking trajectories, sufficient training data, and identification of already known users. In this article, we present GaitSense , a Wi-Fi-based human identification system, to overcome the above unrealistic assumptions. To deal with various walking trajectories and speeds, GaitSense first extracts target specific features that best characterize gait patterns and applies novel normalization algorithms to eliminate gait irrelevant perturbation in signals. On this basis, GaitSense reduces the training efforts in new deployment scenarios by transfer learning and data augmentation techniques. GaitSense also enables a distinct feature of illegal user identification by anomaly detection, making the system readily available for real-world deployment. Our implementation and evaluation with commodity Wi-Fi devices demonstrate a consistent identification accuracy across various deployment scenarios with little training samples, pushing the limit of gait recognition with Wi-Fi signals.

Extensive efforts have been devoted to human gesture recognition with radio frequency (RF) signals. However, their performance degrades when applied to novel gesture classes that have never been seen in the training set. To handle unseen gestures, extra efforts are inevitable in terms of data collection and model retraining. In this article, we present XGest, a cross-label gesture recognition system that can accurately recognize gestures outside of the predefined gesture set with zero extra training effort. The key insight of XGest is to build a knowledge transfer framework between different gesture datasets. Specifically, we design a novel deep neural network to embed gestures into a high-dimensional Euclidean space. Several techniques are designed to tackle the spatial resolution limits imposed by RF hardware and the specular reflection effect of RF signals in this model. We implement XGest on a commodity mmWave device, and extensive experiments have demonstrated the significant recognition performance.

Wi-Fi based sensing systems, although sound as being deployed almost everywhere there is Wi-Fi, are still practically difficult to be used without explicit adaptation efforts to new data domains. Various pioneering approaches have been proposed to resolve this contradiction by either translating features between domains or generating domain-independent features at a higher learning level. Still, extra training efforts are necessary in either data collection or model re-training when new data domains appear, limiting their practical usability. To advance cross-domain sensing and achieve fully zero-effort sensing, a domain-independent feature at the lower signal level acts as a key enabler. In this paper, we propose Widar3.0, a Wi-Fi based zero-effort cross-domain gesture recognition system. The key insight of Widar3.0 is to derive and estimate velocity profiles of gestures at the lower signal level, which represent unique kinetic characteristics of gestures and are irrespective of domains. On this basis, we develop a one-fits-all model that requires only one-time training but can adapt to different data domains. We implement this design and conduct comprehensive experiments. The evaluation results show that without re-training and across various domain factors (i.e. environments, locations and orientations of persons), Widar3.0 achieves 92.7% in-domain recognition accuracy and 82.6%-92.4% cross-domain recognition accuracy, outperforming the state-of-the-art solutions. To the best of our knowledge, Widar3.0 is the first zero-effort cross-domain gesture recognition work via Wi-Fi, a fundamental step towards ubiquitous sensing.

Vital signs such as heart rate and heartbeat interval are currently measured by electrocardiograms (ECG) or wearable physiological monitors. These techniques either require contact with the patient's skin or are usually uncomfortable to wear, rendering them too expensive and user-unfriendly for daily monitoring. In this paper, we propose a new noninvasive technology to generate an Acousticcardiogram (ACG) that precisely monitors heartbeats using inaudible acoustic signals. ACG uses only commodity microphones and speakers commonly equipped on ubiquitous off-the-shelf devices, such as smartphones and laptops. By transmitting an acoustic signal and analyzing its reflections off human body, ACG is capable of recognizing the heart rate as well as heartbeat rhythm. We employ frequency-modulated sound signals to separate reflection of heart from that of background motions and breath, and continuously track the phase changes of the acoustic data. To translate these acoustic data into heart and breath rates, we leverage the dual microphone design on COTS mobile devices to suppress direct echo from speaker to microphones, identify heart rate in frequency domain, and adopt an advanced algorithm to extract individual heartbeats. We implement ACG on commercial devices and validate its performance in real environments. Experimental results demonstrate ACG monitors user's heartbeat accurately, with median heart rate estimation error of 0.6 beat per minute (bpm), and median heartbeat interval estimation error of 19 ms.

This paper presents Widar2.0, the first WiFi-based system that enables passive human localization and tracking using a single link on commodity off-the-shelf devices. Previous works based on either specialized or commercial hardware all require multiple links, preventing their wide adoption in scenarios like homes where typically only one single AP is installed. The key insight underlying Widar2.0 to circumvent the use of multiple links is to leverage multi-dimensional signal parameters from one single link. To this end, we build a unified model accounting for Angle-of-Arrival, Time-of-Flight, and Doppler shifts together and devise an efficient algorithm for their joint estimation. We then design a pipeline to translate the erroneous raw parameters into precise locations, which first finds parameters corresponding to the reflections of interests, then refines range estimates, and ultimately outputs target locations. Our implementation and evaluation on commodity WiFi devices demonstrate that Widar2.0 achieves better or comparable performance to state-of-the-art localization systems, which either use specialized hardwares or require 2 to 40 Wi-Fi links.

Crowdsourced mobile network access (CMNA), in which mobile users can share their Internet access with others, is a promising paradigm for addressing users' increasing needs for ubiquitous connectivity and alleviating cellular network congestion. In this paper, we study the operator-assisted CMNA model, in which a mobile virtual network operator (MVNO) incentivizes its subscribers to operate as mobile WiFi hotspots (hosts) through reimbursement and gets revenue from the relayed traffic. Despite of the promising performance, practical strategies for MVNO and hosts have not been studied yet. Existing works usually assume both MVNO and hosts can obtain complete information, and ignore the accompanied overhead in backhaul and privacy threats to users. Such assumptions are unrealistic in practice. To address this issue, we first systematically characterize the revenue loss for both MVNO and hosts with incomplete market information. Based on the analysis, we propose a novel partial cooperation strategy (PCS) to enable appropriate information exchange between MVNO and hosts with little overhead. With adaptive reimbursement and subtle information control, our PCS efficiently improves MVNO's revenue at equilibrium, and also satisfies the hosts' rationality. Through extensive evaluation on data from the real world, we demonstrate our PCS can improve MVNO's revenue by 23% at equilibrium, compared with the results without PCS.

Link estimation is a fundamental component of forwarding protocols in wireless sensor networks. In low power forwarding, however, the asynchronous nature of widely adopted duty-cycled radio control brings new challenges to achieve accurate and real-time estimation. First, the repeatedly transmitted frames (called wake-up frame) increase the complexity of accurate statistic, especially with bursty channel contention and coexistent interference. Second, frequent update of every link status will soon exhaust the limited energy supply. In this paper, we propose meter, which is a distributed wake-up frame counter. Meter takes the opportunities of link overhearing to update link status in real time. Furthermore, meter does not only depend on counting the successfully decoded wake-up frames, but also counts the corrupted ones by exploiting the feasibility of ZigBee identification based on short-term sequence of the received signal strength. We implement meter in TinyOS and further evaluate the performance through extensive experiments on indoor and outdoor test beds. The results demonstrate that meter can significantly improve the performance of the state-of-the-art link estimation scheme.

Link estimation is a fundamental component of forwarding protocols in wireless sensor networks. In low power forwarding, however, the asynchronous nature of widely adopted duty-cycled radio control brings new challenges to achieve accurate and real- time estimation. First, the repeatedly transmitted frames (called wake-up frame) increase the complexity of accurate statistic, especially with bursty channel contention and coexistent interference. Second, frequent update of every link status exhausts the limited energy supply due to long duration of beacon broadcast. In this paper, we propose meter (Distributed Frame Counter), which takes the opportunities of link overhearing to update link status in real time. Furthermore, meter does not only depend on counting the successfully decoded wake-up frames, but also counts the corrupted ones by exploiting the feasibility of ZigBee identification based on short-term sequence of the received signal strength. We implement meter in TinyOS and further evaluate the performance through extensive experiments on indoor and outdoor testbeds. The results demonstrate that meter can significantly improve the performance of the state-of-the-art link estimation schemes.

Offloading network traffic from cellular to WiFi is widely used to reduce energy consumption since WiFi is assumed to have lower power consumption than cellular. However, we find that WiFi link quality may vary significantly under user mobility. Consequently, the energy efficiency of WiFi varies and sometimes becomes even worse than that of cellular. Therefore, widely used WiFi offloading may not be beneficial or even incurs more energy consumption. To address this issue, we propose Furion, an energy efficient WiFi offloading scheme that exploits beneficial WiFi links on smartphones. Towards such a goal, we investigate the relationship between energy efficiency and link quality. Accordingly, we propose a practical probabilistic model to predict WiFi energy efficiency based on the dynamics of link quality. We further extend the method to different environments by exploiting contextual factors in the prediction model to improve the accuracy. Based on the model, we design an adaptive offloading scheme to optimize the energy efficiency of WiFi offloading, while also guaranteeing user experience. We have implemented Furion on the Android platform and conduct extensive real-world experiments. The results demonstrate that Furion achieves 34.13% improvement in energy efficiency compared with the state-of-the- arts.

Driven by the proliferation of mobile applications, the conflict between data communication requirement and limited battery capacity is becoming sharp on modern smartphones. Offloading mobile traffic from cellular to WiFi is widely recognized as a viable solution to improve the energy efficiency. However, through extensive field experiments, we find WiFi offloading is not always energy efficient and even consumes more energy than cellular network due to link quality variation. In addition, we also observe that practical data transmission deadline requirement and link utilization allows scheduling of data traffic to time periods with good link quality. Accordingly, we propose Q-offload, the first attempt towards energy efficient WiFi offloading with link dynamics. In Q-offload, we propose an iterative framework to achieve energy efficient WiFi offloading by exploiting good link quality while not affecting user experience. We evaluate the performance of Q-offload through both trace-driven analysis and real-world experiments. The results show that it can achieve 33.5%~55.7% energy efficiency improvement, compared with state-of-the-arts under different conditions.

Smartphones nowadays are installed with diverse applications, each of which consumes energy and bandwidth. As more and more applications are crowded into a smart- phone, they cause serious problems with regard to battery life and bandwidth utilization. Existing proposals to tackle such challenges usually resort to two ways: avoiding energy- consuming network activities or improving communication efficiency in terms of power consumption. Those approaches either affect the smartphone users' experience, or offer little benefit in prolonging the battery life. Motivated by insightful understanding of users' habit, we in this paper propose a novel approach to orchestrate network activities of smartphone applications, based on user's habit. We implement our approach on smartphones as a middleware service called NetMaster. The performance evaluation with real traces shows that NetMaster reduces energy consumption of network activities by 77.8% in average and increases network bandwidth utilization by over 200%. The user experience is surprisingly well preserved. The chance of undesired interrupt during normal usage is less than 1%.