Shueng-Han Gary Chan

Counting Wi-Fi devices within access point (AP) coverage by listening to their probe request frames (PRFs) is a well-established research problem, fundamental to many Internet of Things (IoT) applications such as crowd management and public transportation scheduling. While commendable counting accuracy has been reported, existing approaches fall short in deployment convenience due to their reliance on machine learning, which necessitates (1) extensive data collection and training efforts for system setup, and (2) specialized model fine-tuning for operational maintenance. We propose RateCount, an accurate, lightweight, and learning-free counting approach to lower deployment costs. RateCount employs a provably unbiased closed-form expression to estimate the device count based on the rate at which APs receive PRFs, along with an error model to compute the estimation variance. We also demonstrate its application in people counting by incorporating a device-to-person calibration scheme. Through extensive real-world experiments conducted at multiple sites spanning a wide range of counts, we show that RateCount, without any deployment costs for machine learning, achieves comparable counting accuracy to the state-of-the-art (SOTA) learning-based device counting and improves previous people counting schemes by a large margin.

Background: Dementia increases the risk of individuals getting lost due to cognitive decline, impacting daily functioning and heightening caregiver worry. Traditional search methods are often time-consuming and stressful, whereas GPS-based technologies face limitations such as battery dependency. A crowdsourcing Internet of Things (IoT) technology using energy-efficient Bluetooth Low Energy (BLE) offers a potential solution to locate missing individuals with dementia more effectively by harnessing the power of the crowd and fostering a caring and inclusive community. Objective: This study aimed to evaluate the effectiveness of a BLE-based privacy-preserving crowdsourcing IoT system consisting of a BLE tag and an Android and iOS app in improving lost-related behavior and psychological well-being by facilitating searches, after-care arrangements, and reducing caregiver worry, as well as to assess its usability among caregivers of individuals with dementia in Hong Kong. Methods: A single-arm, prospective observational study was conducted from November 2020 to October 2023. Caregivers (N=1034) of individuals with dementia used a staff-assisted crowdsourcing IoT technology comprising a BLE tag, mobile app sensor, and location cloud server. Outcomes included search strategies, post–getting lost care arrangements, caregiver worry and distress (10-point scale), and usability (modified Quebec User Evaluation of Satisfaction with Assistive Technology 2.0 survey). Data were collected at 6- and 12-month follow-ups and analyzed using generalized estimating equations and linear mixed models. Results: Of the 1034 participants, 143 (13.82%) reported lost episodes, with 51 (35.7%) using BLE tags for searches. Worry about future lost episodes decreased significantly over time ( P =.008), especially among BLE tag users ( P =.04). There was an association between BLE tag use and adoption of proactive search strategies (eg, going out to search: adjusted odds ratio 2.78, 95% CI 1.33-5.82; P =.007) and preventative measures (eg, IoT devices or CCTV: adjusted odds ratio 2.92, 95% CI 1.61-5.29; P <.001). Usability satisfaction was high for design and data security, whereas approximately half of the participants (309/707, 43.7%) were satisfied with accuracy. Conclusions: The BLE crowdsourcing system may reduce caregiver worry and encourage proactive search behaviors, although accuracy depends on broader community adoption. Integration into dementia care plans could enhance safety and autonomy. Further research with a randomized controlled trial design is needed to confirm these findings.

The delivery of goods to numerous indoor stores poses significant safety risks, with heavy, high-stacked packages on delivery trolleys posing a potential hazard to passersby. This paper reports our experiences of developing and operating DeMo, a practical system for real-time monitoring of indoor delivery. DeMo employs sensors attached to trolleys, utilizing Inertial Measurement Unit (IMU) and Bluetooth Low Energy (BLE) readings to detect delivery violations, such as speeding and the use of non-designated delivery paths, and ensure accurate matching of each delivery to its intended destination store. Unlike typical indoor localization applications, DeMo addresses unique challenges, including sensor placement and the complex electromagnetic characteristics encountered in underground settings. Specifically, DeMo adapts the classical logarithmic radio signal model to facilitate fingerprint-free localization, significantly reducing deployment and maintenance costs. DeMo has been operating since May 2020, covering more than 200 shops with 74,537 deliveries (6193.2 km) across 12 subway stations in Hong Kong. DeMo’s 4-year operation witnessed a significant violation rate drop, from 19% (May 2020) to 0.9% (Mar 2024).

Knowing a pedestrian’s conveyor state of “elevator,” “escalator,” or “neither” is fundamental to many applications such as indoor navigation and people flow management. Previous studies on classifying the conveyor state often rely on specially designed body-worn sensors or make strong assumptions on pedestrian behaviors, which greatly strangles their deployability. To overcome this, we study the classification problem under arbitrary pedestrian behaviors using the inertial navigation system (INS) of the commonly available smartphones (including accelerometer, gyroscope, and magnetometer). This problem is challenging, because the INS signals of the conveyor states are entangled by the arbitrary and diverse pedestrian behaviors. We propose ELESON, a novel and lightweight deep-learning approach that uses phone INS to classify a pedestrian to elevator, escalator, or neither. Using causal decomposition and adversarial learning, ELESON extracts the motion and magnetic features of conveyor state independent of pedestrian behavior, based on which it estimates the state confidence by means of an evidential classifier. We curate a large and diverse dataset with 36,420 instances of pedestrians randomly taking elevators and escalators under arbitrary unknown behaviors. Our extensive experiments show that ELESON is robust against pedestrian behavior, achieving a high accuracy of over 0.9 in F1 score, strong confidence discriminability of 0.81 in AUROC (Area Under the Receiver Operating Characteristics), and low computational and memory requirements fit for common smartphone deployment.

WiFi received signal strength (RSS) environment evolves over time due to the movement of access points (APs), AP power adjustment, installation and removal of APs, etc. We study how to effectively update an existing database of fingerprints, defined as the RSS values of APs at designated locations, using a batch of newly collected unlabelled (possibly crowdsourced) WiFi signals. Prior art either estimates the locations of the new signals without updating the existing fingerprints or filters out the new APs without sufficiently embracing their features. To address that, we propose GUFU, a novel effective graph-based approach to update WiFi fingerprints using unlabelled signals with possibly new APs. Based on the observation that similar signal vectors likely imply physical proximity, GUFU employs a graph neural network (GNN) and a link prediction algorithm to retrain an incremental network given the new signals and APs. After the retraining, it then updates the signal vectors at the designated locations. Through extensive experiments in four large representative sites, GUFU is shown to achieve remarkably higher fingerprint adaptivity as compared with other state-of-the-art approaches, with error reduction of 21.4% and 29.8% in RSS values and location prediction, respectively.

A user's movement path can be precisely and concisely described as a concatenation of straight lines having the user's turns as their end points. Learning such a path description or representation from inertial measurement unit (IMU) sensors enables various mobile and IoT applications, as it allows efficient processing of the movement path data. It is, however, non-trivial to learn a succinct yet accurate path description from IMU sensor readings in the mobile device of a moving user on the fly due to the dynamically changing behaviors and the technical difficulty in detecting the user's turns. We propose PATHLIT, a novel online path description learning system based on IMU signals. PATHLIT learns position vectors of a user from IMU sensor readings by our custom-made self-attention network model. Once each position vector is learned, PATHLIT also decides whether or not to take it as a part of the resulting path description by our efficient online algorithm developed under the minimum description length principle, which essentially detects the user's turns along the path. We conduct extensive experiments on two large datasets. The experiment results show that PATHLIT achieves superior performance over state-of-the-art algorithms by up to 50% in absolute trajectory error using only 15% of trajectory data points.

Due to its promising results, density map regression has been widely employed for image-based crowd counting. The approach, however, often suffers from severe performance degradation when tested on data from unseen scenarios, the so-called "domain shift" problem. To address the problem, we investigate in this work single domain generalization (SDG) for crowd counting. The existing SDG approaches are mainly for image classification and segmentation, and can hardly be extended to our case due to its regression nature and label ambiguity (i.e., ambiguous pixel-level ground truths). We propose MPCount, a novel effective SDG approach even for narrow source distribution. MPCount stores diverse density values for density map regression and reconstructs domain-invariant features by means of only one memory bank, a content error mask and attention consistency loss. By partitioning the image into grids, it employs patch-wise classification as an auxiliary task to mitigate label ambiguity. Through extensive experiments on different datasets, MPCount is shown to significantly improve counting accuracy compared to the state of the art under diverse scenarios unobserved in the training data characterized by narrow source distribution. Code is available at https://github.com/Shimmer93/MPCount.