In recent years, unmanned aerial vehicles (UAVs) have received growing attention due to security threats issues. Though UAV detection and positioning systems are commonly used in various scenarios, most of the available systems are still suffering from low accuracy and susceptibility to the environment. Therefore, it is still necessary to design a highly accurate, versatile UAV detection and positioning system. In this paper, a UAV detection and positioning system based on multi-dimensional signal features is proposed. The first step of the system is to monitor the communication signal and channel state information (CSI) between the UAV and the controller. Subsequently the signal frequency spectrum (SFS), the wavelet energy entropy (WEE), and the power spectral entropy (PSE) are extracted as features. At the same time, machine learning algorithms combining the above features are applied to detect UAVs. After the UAV is successfully detected, the spatial features such as angle of azimuth (AOA) and angle of elevation (AOE) were extracted for UAV localization based on a super-resolution estimation algorithm. In conclusion, the Wireless Insite (WI) software was employed for long-distance positioning verification while the software-defined radio (SDR) was used for small-scale testing and verification. The experimental results show that the average detection rate of combining multiple features in the test environment is 95.58%, the median accuracy of 2D positioning is 0.76 m, and the median accuracy of 3D positioning is 1.2 m. In the WI simulation environment, the median accuracy of 2D positioning is 1.1 m, and the median accuracy of 3D positioning is 2.35 m.

Human activity recognition has been growing for decades in a variety of technological disciplines. However, in the existing WiFi-based human activity recognition systems, there are the following problems: Firstly, in the processing of channel state information (CSI) data, mainly for the removal of noise in the superimposed signal, there is no effective removal of useless multipath signals; Secondly, the data segmentation algorithm based on the empirical threshold requires manual adjustment of the threshold in different environments, resulting in poor robustness and unstable segmentation; Thirdly, simple learning classification is applied without specific design for CSI data structure and sufficiently abstracting information features. In this paper, a device-free human activity recognition system with a temporal-frequency attention mechanism is proposed, which can be deployed on commercial WiFi devices to identify human’s daily activities. Firstly, the multipath signal affected by the channel change is extracted by using the difference of the propagation delay of different multipath, thereby eliminating the delay and invalid multipath signals that have undergone multiple reflections and refractions. Secondly, a neural network model based on attention mechanism is proposed, which assigns different weights to different characteristics and sequences by imitating the human brain to dedicate more attention to important information. Then, the long short-term memory (LSTM) model is used to learn the correlation features of different dimensions to realize human activity recognition. Finally, the system performance is evaluated in different environments, and the experimental results show that our syetem holds a better performance in both line-of-sight (LOS) and non-line-of-sight (NLOS) than the existing human activity recognition systems.

Indoor localization using WiFi signal parameters is challenging, with encouraging decimeter localization results available with enough line-of-sight coverage and hardware infrastructure. This paper proposes a new 2-dimensional multiple packets based matrix pencil (2D M-MP) method to estimate the Angle of Arrival (AoA) and Time of Flight (ToF) based on WiFi channel state information (CSI). Compared with the conventional parameter estimation algorithms, this method has two advantages. First, 2D M-MP method uses the discrete Fourier transform (DFT) to convert the complex computation into real computation to reduce the computational complexity significantly without losing accuracy. Second, it accumulates multiple CSI packets to improve the parameter estimation accuracy effectively, especially at low values of signal-to-noise-ratio (SNR) environment. To verify the practicability of our proposed 2D M-MP method, we set up a localization system in an actual scenario using commodity WiFi cards which demonstrates that the performance of 2D M-MP method is better than conventional parameter estimation algorithms and can achieve a localization accuracy of 42 cm in indoor hall deployment.

Due to the pervasive deployment of Wi-Fi devices, Wi-Fi-based human intrusion detection has attracted intensive attention in recent years. By analyzing temporal changes of signal caused by human movement, the existing solutions can achieve passive human intrusion detection. However, the detection of intrusion direction, from which the person intrudes the area of interests, is still a challenging task. Considering the influence of intrusion direction on the energy and spatial statistical distributions of multiple reflections, in this study, we propose WIDD, a Wi-Fi-based passive human intrusion direction detection system. Concretely, WIDD first extracts the angle of arrival (AoA) and amplitude distributions of the multipath signal and conducts the Jarque-Bera (JB) test to analyze the normality of the obtained distributions. A novel joint hypothesis testing algorithm is proposed to monitor changes in amplitude and AoA distributions of reflections and realize the human intrusion detection based on the analysis result. At last, it finds the reflection affected by the human intrusion most via the divergence evaluation and extracts the corresponding AoA to realize the intrusion direction detection. The comprehensive experimental evaluation demonstrates that WIDD provides human intrusion direction detection accuracy of about 0.943 and 0.931, under the scenario where the transmitter and receiver are separated by the glass and brick wall, respectively, outperforming the state-of-the-art solutions and shedding bright lights on ubiquitous human detection in practice.

This paper proposes a compact $4 \times 4$ filtering microstrip patch antenna array with Dolph- Chebyshev distribution implemented by multilayer structure. The proposed antenna array features good filtering performance but contains no extra filtering parts in the feeding or the radiating element. The whole antenna array consists of four $1 \times 4$ antenna subarrays, each subarray is composed of four patch antennas which are coupled-fed by a single microstrip line. Inside the desired passband, by properly choosing the coupling area along the microstrip line, the patches are induced in-phase currents therefore achieving the excellent radiation characteristics. On the other hand, the four patches can’t be properly excited outside the passband which would enable radiation suppression performance. In order to implement a $4 \times 4$ filtering microstrip patch antenna array with low sidelobe level (SLL), four subarrays are combined and fed by Dolphy-Chebyshev excitation method. Finally, the proposed antenna array is implemented, fabricated and measured. The maximum measured gain in the passband is 14.7 dBi, and the radiation suppression outside the passband is better than 25 dB.

In the context of integrated sensing and communication, indoor device-free positioning and tracking technology based on channel state information (CSI) has become a key and significant branch. CSI extracted by high-bandwidth chips is not widely used for sensing due to problems of clock asynchrony and phase locking instability. This paper implements passive positioning based on the CSI collected by the bcm4366 Wi-Fi chip. First, through extensive testing and observation, we describe the unstable phase difference phenomenon of the radio frequency channel of the bcm4366 chip. Calibrating phase ambiguity (PA) is implemented by a phase detection and compensation method. Based on this, a three-dimensional (3D) beamspace matrix pencil (BMP) algorithm is proposed to jointly estimate the angle of arrival, time of flight and Doppler frequency shift of CSI dynamic targets. Finally, passive localization is achieved based on joint multi-parameters. Experiments verify that the 3D BMP algorithm is superior to existing algorithms in terms of trade-off estimation accuracy and computational complexity. The positioning scheme based on 3D BMP effectively reduces the positioning error after eliminating PA. In the single-link case, the median localization errors of the activity room and meeting room scenarios reach 0.42m and 0.55m, respectively.