Owing to its ultra-low power consumption and long-range communication capabilities, LoRa, a prominent Low-Power Wide Area Network (LPWAN) technology, has seen widespread deployment across academic and industrial domains. However, its operation in unlicensed spectrum and reliance on ALOHA-based Medium Access Control (MAC) make it highly susceptible to packet collisions, undermining network reliability and performance. While existing efforts primarily address intra-channel collisions, the effects of cross-channel interference—caused by overlapping transmissions using different bandwidths or spreading factors—remain insufficiently studied, posing challenges to communication quality, scalability, and spectrum efficiency. To bridge this gap, we propose AntScale, a novel collision resolution framework based on antithetical - stationary scaling. Specifically, AntScale introduces two key techniques: (1) it transforms minor time-domain signal distortions into robust frequency-domain features through dual antithetical scaling, enabling effective collision suppression; and (2) it employs a patching algorithm that leverages wide-bandwidth signal characteristics to identify and isolate dominant interference components for targeted decoding. Implementation on a USRP N210 platform demonstrates that AntScale achieves a 2.3× throughput improvement over state-of-the-art methods.

To achieve 2-D localization using only the Time Difference of Arrival (TDoA) measurements, a minimum of three base stations is required. To further reduce the number of base stations needed for positioning, this study proposes, for the first time, a novel method that combines TDoA and relative speed to base stations (RStBSs) to locate the targets, reducing the minimum number of base stations required by one compared to methods that rely solely on TDoA. The proposed method transforms the coordinate system, constructs a system of equations for the target position in the transformed coordinate system, and solves these equations to determine the target position. The algorithm then substitutes the two types of measurements, TDoA and RStBS, into these equations to estimate the target position. The simulation results indicate that the algorithm achieved a positioning accuracy better than 1.0 m (75%) without cumulative error when the standard deviation of the measurement noise was less than 0.3 m. In the two-base-station acoustic indoor positioning experiment, a permanent magnetic block was utilized to determine the initial position of the target, and the TDoA and RStBS measurements were calculated by estimating the audio arrival time (AAT). The experimental results indicate that the target positions estimated by the proposed method closely align with the actual motion trajectory, particularly when fused with conventional pure TDoA localization, thereby confirming its effectiveness in real-time target tracking. Furthermore, integrating this approach with acoustic NLOS identification techniques is expected to significantly enhance the robustness and practical utility of acoustic indoor positioning systems in complex environments.

Recent studies show that inaudible attacks pose a non-negligible security risk to smart speakers. While several countermeasures have been proposed to detect the occurrence of the inaudible attack passively, accurately locating the attack source in 3D free space remains an unresolved challenge. Arrow is designed to bridge this gap by attempting to detect the occurrence of inaudible attacks and determine their localization simultaneously. Instead of relying on dedicated hardware components, Arrow is implemented with the microphone array widely deployed on COTS (Commercial Off-The-Shelf) smart speakers. Throughout the spatial information captured by the microphone array, Arrow establishes a spatial mapping model and derives orientation-related features to pinpoint the location of the attack source. Furthermore, to improve the robustness against co-channel interference, Arrow adopt carefully-modulated ultrasonic waveforms to achieve noise-robust attack detection. Through the above technical mechanism, Arrow can significantly improve the security level of voice assistants on smart speakers with nearly zero deployment cost. We implement a prototype of Arrow and conduct a comprehensive performance evaluation. The results show Arrow can achieve 2.5 ○ and 7 ○ error in DoA estimation for horizontal and vertical angles, respectively.

In industrial monitoring, the accurate detection of subtle mechanical vibrations is essential for implementing effective preventive maintenance strategies and improving equipment longevity. However, current RFID-based sensing technologies are hindered by environmental noise and the inherent limitations of signal wavelengths. To address these issues, we introduce TagVibra, an innovative RFID-based system enhanced by the Differential Phase Amplification (DPA) algorithm. This algorithm harnesses the modulation characteristics of RFID technology along with the harmonic effects of RFID tags to markedly enhance vibration detection accuracy. TagVibra has been implemented on the USRP platform, and experimental results confirm its ability to detect minute vibrations down to 0.3 mm with an average frequency estimation error of less than 0.9 Hz. This breakthrough significantly advances the capabilities for early fault diagnosis in high-precision industrial settings.

Recently, motor imagery decoding technology based on electroencephalogram (EEG) signals has made significant progress. However, there are still challenges in adapting to new sessions, mainly due to changes in data distribution between different sessions and confusion problems caused by similar oscillation patterns in various categories of EEG signals. To address these problems, this paper proposes a debiased domain adaptation framework with minimum class confusion to learn unbiased representations in motor imagery tasks. Specially, unlike the feature alignment and adversarial training methods, we explore the class predictions for domain adaptation, applying the minimum class confusion loss criterion in the target domain to reduce inter-class confusion. This approach aims to alleviate the bias issues inherent in classifiers trained on the source domain for making predictions in the target domain, achieving class-level alignment. Consequently, it enhances the model’s ability to adapt to the data distribution of the target domain. Experimental results on two public EEG datasets (BCI Competition IV datasets IIa and IIb) show that the method for cross-session decoding is significantly improved compared to the baseline, with average classification accuracy reaching 81.01% and 82.92%, respectively.

Recent works have shown that inaudible signals (at ultrasound frequencies) can become audible to the microphone by exploiting the nonlinear effects. With a well-designed inaudible signal, an adversary can control Amazon Echo and Google Homelike devices in people’s rooms silently and remotely. A voice command like “Alexa, open the door“ can be a serious treat. Although recent works design various methods against such inaudible attacks, one important issue remains open: there is no clear solution to locate the attack source accurately. Obviously, the only way to completely eliminate such inaudible threats is to locate and remove the attack source. This paper is an attempt to close this gap. We propose Arrow, an effective method to help users locate the ultrasound attack source in 3D space indoors. Arrow establishes the relationship between inaudible signals and the recorded sounds of the microphone, and then explores the architecture of the embedded microphone array on smart speaker for extracting a 3D direction-specific signature. By learning such directional signature, Arrow can accurately estimate the spatial orientation of the inaudible attack source and help users to locate and remove it. We implement a prototype of Arrow and conduct comprehensive experiments to validate its performance. The results show Arrow can achieve 2.5° and 7° error in DoA(Direction of Arrival) estimation for horizontal and vertical angles, respectively.

Vibration is a normal reaction that occurs during the operation of machinery and is very common in industrial systems. How to turn fine-grained vibration perception into visualization, and further predict mechanical failures and reduce property losses based on visual vibration information, which has aroused our thinking. In this article, the phase information generated by the tag is processed and analyzed, and MFD is proposed, a real-time vibration monitoring and fault-sensing discrimination system. MFD extracts phase information from the original RF signal and converts it into a Markov transition map by introducing White Gaussian Noise and a low-pass filter for denoising. To accurately predict the failure of machinery, a deep and machine learning model is introduced to calculate the accuracy of failure analysis, realizing real-time monitoring and fault judgment. The test results show that the average recognition accuracy of vibration can reach 96.07%, and the average recognition accuracy of forward rotation, reverse rotation, oil spill, and screw loosening of motor equipment during long-term operation can reach 98.53%, 99.44%, 97.87%, and 99.91%, respectively, with high robustness.

Radio Frequency IDentification (RFID) is emerging as a vital technology of the Internet of Things (IoT). Billions of RFID tags have been deployed to locate daily objects such as equipment, pharmaceuticals, vehicles, and so on. Unlike previous solutions that focus on localizing tagged objects in the world coordinate system in reference to reader antennas, this work exploits a system, called RFCamera, that can identify and locate RFID-tagged objects in images with pixel dimensions. Our core insight is that an image is a visual AoA profile in terms of lights, which is resulted from the pinhole camera model. Similarly, we generate an RF image derived from the AoA profile of a tag using the same pinhole model as the camera. Consequently, the locations of visual entities corresponding to tagged objects are highlighted by comparing two types of images. To this end, we customized a camera system equipped with a pair of rotatable reader antennas. Our experimental evaluation demonstrates that RFCamera enables a mean error of 5.7∘ and 2.9∘ at azimuth and elevation angle estimation, respectively. It can locate a visual entity with a mean error of 51 pixels (i.e., ≈1.3 cm at 96 dpi) in a 640× 480 image.

Conventional vibration sensing systems, equipped with specific sensors (e.g., accelerometer) and communication modules, are either expensive or cumbersome to deploy. Recently research community revisits this classic topic by taking advantage of off-the-shelf RFIDs. However, limited by low reading rate and long wavelength, current RFID based solutions can only sense low-frequency (e.g., below 100 Hz) mechanical vibrations with larger amplitude (e.g., >5 mm). To address the issue, this work presents TagSound, an RFID-based vibration sensing system that explores a tag's harmonic backscattering to recover high-frequency and tiny mechanical vibrations accurately. The key innovations are in two aspects: harmonics based sensingand a newrecovery scheme. We implement TagSound with USRP platforms. Our comprehensive evaluation shows (i) TagSound can achieve a mean error of 0.37 Hz when detecting vibrations at frequencies below 100 Hz, and a mean error of 4.2 Hz even when the vibration frequency is up to 2500 Hz. (ii) TagSound can achieve a Hz-level frequency estimation even when the vibration amplitude is only 2 mm.

In recent years, considerable effort has been recently exerted to explore the high-precision RF-tracking systems indoors to satisfy various real-world demands. However, such systems are tailored for a particular type of device (e.g., RFID, WSN or Wi-Fi). With the rapid development of the Internet of Things (IoT), various new wireless protocols (e.g., LoRa, Sigfox, and NB-IoT) have been proposed to accommodate different demands. The coexistence of multiple types of IoT devices forces users to deploy multiple tracking systems in a warehouse or a smart home where various IoT devices are running, which causes huge additional costs in installation and maintenance. To address this issue, this work presents iArk, which is a general-purpose tracking platform for all types of IoT devices working at the ultra high frequency band. Our innovation lies in the design of the "K+1"-model hardware, the protocol free middleware, and the multipath resistant learnware. By the virtue of decoupling from wireless protocols, iArk also allows researchers to concentrate on developing a new tracking algorithm without considering the protocol diversity. To date, the platform can support five mainstream types of IoT devices (i.e., NB-IoT, LoRa, RFID, Sigfox and Zigbee) and is scalable to other types with minimal effort.

Radio Frequency IDentification (RFID) is emerging as a vital technology of the Internet of Things. Billions of RFID tags have been deployed to locate daily objects such as equipment, pharmaceuticals, and vehicles, and so on. Unlike previous solutions that focus on localizing tagged objects in the world coordinate system in reference to reader antennas, this work exploits a system, called RFCamera, that can identify and locate RFID-tagged objects in images with pixel dimensions. Many applications would benefit from RFCamera. For instance, the RF-aware image annotation system is able to generate rich annotations for RFID-tagged entities in images at the pixel level for the deep learning; the RF-aware auto-focus allows surveillance camera to exactly focalize the burglar who carries the stolen tagged-property out of a crowd. Our core insight is that an image is a visual AoA profile in terms of lights, which is resulted from the pinhole camera model. Similarly, we generate an RF image, derived from the AoA profile of a tag using the same pinhole model as the camera. Consequently, the locations of visual entities corresponding to tagged objects are highlighted by comparing two types of images. To this end, we customized a camera system equipped with a pair of rotatable reader antennas. Our experimental evaluation demonstrates that RFCamera enables a mean error of 5.7° and 2.9° at azimuth and elevation angle estimation. It can locate a visual entity with a mean error of 51 pixels (i.e., ≈ 1.3 cm at 96 dpi) in a 640 × 480 image.

Conventional vibration sensing systems, equipped with specific sensors (e.g., accelerometer) and communication modules, are either expensive or cumbersome in deployment. In recent years, the community revisits this classic topic by taking advantage of off-the-shelf RFIDs. However, limited by lower reading rate and larger wavelength, current RFID based solutions can only sense low-frequency (e.g. below 100Hz) mechanical vibrations with larger amplitude (e.g. (>) 5mm). To address this issue, this work presents TagSound, an RFID-based vibration sensing system that explores a tag's harmonic backscattering to recover high-frequency and tiny mechanical vibrations accurately. The key innovations are in two aspects: harmonics based sensing and a new recovery scheme. We implement TagSound with USRP platforms. Our comprehensive evaluation shows TagSound can achieve a mean error of 0.37 Hz when detecting vibrations at frequencies below 100Hz, and a mean error of 4.2 Hz even when the vibration frequency is up to 2500Hz.

This paper presents WordRecorder, an efficient and accurate handwriting recognition system that identifies words using acoustic signals generated by pens and paper, thus enabling ubiquitous handwriting recognition. To achieve this, we carefully craft a new deep-learning based acoustic sensing framework with three major components, i.e., segmentation, classification, and word suggestion. First, we design a dual-window approach to segment the raw acoustic signal into a series of words and letters by exploiting subtle acoustic signal features of handwriting. Then we integrate a set of simple yet effective signal processing techniques to further refine raw acoustic signals into normalized spectrograms which are suitable for deep-learning classification. After that, we customize a deep neural network that is suitable for smart devices. Finally, we incorporate a word suggestion module to enhance the recognition performance. Our framework achieves both computation efficiency and desirable classification accuracy simultaneously. We prototype our design using off-the-shelf smartwatches and conduct extensive evaluations. Our results demonstrate that WordRecorder robustly archives 81% accuracy rate for trained users, and 75% for users without training, across a range of different environment, users, and writing habits.

Acoustic sensing has brought forth the advances of prosperous applications such as gesture recognition. Specifically, ambient acoustic sensing has drawn many contentions due to the ease of use property. Unfortunately, the inherent ambient noise is the major reason for unstable gesture recognition. In this work, we propose “SoundWrite II”, which is an improved version of our previously designed system. Compared with our previous design, we utilize the two threshold values to identify the effective signals from the original noisy input, and leverage the MFCC (Mel frequency cepstral coefficient) to extract the stable features from different gestures. These enhancements could effectively improve the noise tolerant performance for previous design. Implementation on the Android system has realized the real time processing of the feature extraction and gesture recognition. Extensive evaluations have validated our design, where the noise tolerant property is fully tested under different experimental settings and the recognition accuracy could be 91% with 7 typical gestures.

In this work, we leverage the reflection property explored in our experimental results and propose an efficient and effective algorithm for the reconstruction of the layout. The proposed algorithm identifies the reflective path, which is caused by obstacle, and evaluated by the difference of Angle of Arrival (AoA) at each transceiver deployed at the room. Then with the knowledge of AoA corresponding to the affected reflective path, the reflection points can be calculated. These reflection points could be used to outline the surface of obstacle. Moreover, we can estimate the shape, size, and location of the obstacle by using AlphaShape algorithm. The simulation results prove the effectiveness of our algorithm.