Manual annotation (MA) of sensor data for Human Activity Recognition (HAR) is labor-intensive, error-prone, and limits scalability. This paper proposes a multi-sensor methodology to automatically generate training labels (aka. soft labels) for HAR systems without human intervention. The approach integrates data from inertial measurement units glued to objects of daily life with the Received Signal Strength Indicator (RSSI) information derived from BLE beacon anchors for estimating both the performed activity and the subject’s location. We validate the quality of the generated soft labels against video-based MA ground truth. Experimental results show that a deep learning model for HAR trained on a Wi-Fi Channel State Information (CSI) dataset annotated with soft labels achieves comparable results with respect to the same model trained on the corresponding manually-annotated dataset.

Indoor and outdoor localization systems are now essential tools for delivering Internet of Things (IoT)-based services across a range of fields, including smart buildings, industrial automation, and healthcare. These systems primarily aim to determine the position of users or objects within a given environment. Due to the abundance of electronic devices in these settings, localization techniques that utilize device interactions, such as fingerprinting, offer a cost-effective and accurate solution. However, the fingerprinting approach involves a time-intensive mapping phase, and training and executing the associated localization algorithm can impose demanding time, energy, and computational resource requirements. In this context, our paper introduces a practical system that eliminates the need for time-consuming environment mapping by leveraging radio signal emitters within the environment. Our approach achieved localization accuracy with an error range between 1.5 $m^2$ (best) - 6.5 $m^2$ (worst). Additionally, we collect human activity data from accelerometer, gyroscope, and magnetometer sensors to develop pattern recognition models for everyday activities.

Falls can have significant and far-reaching effects on various groups, particularly the elderly, workers, and the general population. These effects can impact both physical and psychological well-being, leading to long-term health problems, reduced productivity, and a decreased quality of life. Numerous fall detection systems have been developed to prompt first aid in the event of a fall and reduce its impact on people's lives. However, detecting a fall after it has occurred is insufficient to mitigate its consequences, such as trauma. These effects can be further minimized by activating safety systems (e.g., wearable airbags) during the fall itself—specifically in the pre-impact phase—to reduce the severity of the impact when hitting the ground. Achieving this, however, requires recognizing the fall early enough to provide the necessary time for the safety system to become fully operational before impact. To address this challenge, this paper introduces a novel lightweight convolutional neural network (CNN) designed to detect pre-impact falls. The proposed model overcomes the limitations of current solutions regarding deployability on resource-constrained embedded devices, specifically for controlling the inflation of an airbag jacket. We extensively tested and compared our model, deployed on an STM32F722 microcontroller, against state-of-the-art approaches using two different datasets.

The measurement of breath rate (BR) is essential for comprehensive human health monitoring across a wide range of scenarios. Several studies in the literature have explored the estimation of BR using millimeter wave (mmWave) technology. However, these approaches typically focus on a single subject at a time. To enable multi-person estimation, researchers have often relied on data fusion with camera systems or employed specialized hardware configurations. On the contrary, this paper proposes a methodology that employs only one Frequency Modulated Continuous Wave (FMCW) radar to estimate the BR of multiple subjects stationary in the environment. The proposed methodology includes a pre-processing pipeline to refine the radar-captured signals, followed by frequency-domain analysis to distinguish between subjects. Finally, phase variations in the reflected signals caused by chest movements are analyzed to estimate the BR. Advantages and limitations of the approach are discussed on the basis of an experimental campaign.

In the context of smart environments, crafting remote monitoring systems that are efficient, cost-effective, user-friendly, and respectful of privacy is crucial for many scenarios. Recognizing and tracing individuals via markerless motion capture systems in multi-person settings poses challenges due to obstructions, varying light conditions, and intricate interactions among subjects. In contrast, methods based on data gathered by Inertial Measurement Units (IMUs) located in wearables grapple with other issues, including the precision of the sensors and their optimal placement on the body. We claim that more accurate results can be achieved by mixing Human Pose Estimation (HPE) techniques with information collected by wearables. To do that, we introduce a real-time platform that fuses HPE and IMU data to track and identify people. It exploits a matching model that consists of two synergistic components: the first employs a geometric approach, correlating orientation, acceleration, and velocity readings from the input sources. The second utilizes a Convolutional Neural Network (CNN) to yield a correlation coefficient for each HPE and IMU data pair. The proposed platform achieves promising results in identification and tracking, with an accuracy rate of 96.9%.

Using WiFi's Channel State Information for human activity recognition—referred to as WiFi sensing—has attracted considerable attention. But despite this interest and many publications over a decade, WiFi sensing has not yet found its way into practice because of a lack of robustness of the inference results. In this paper, we quantitatively show that even “microchanges” in the environment can significantly impact WiFi signals, and potentially alter the ML inference results. We therefore argue that new training and inference techniques might be necessary for mainstream adoption of WiFi sensing.

In the context of smart buildings and smart cities, the design of low-cost and privacy-aware solutions for recognizing the presence of humans and their activities is becoming of great interest. Existing solutions exploiting wearables and video-based systems have several drawbacks, such as high cost, low usability, poor portability, and privacy-related issues. Consequently, more ubiquitous and accessible solutions, such as WiFi sensing, became the focus of attention. However, at the current state-of-the-art, WiFi sensing is subject to low accuracy and poor generalization, primarily affected by environmental factors, such as humidity and temperature variations, and furniture position changes. Such is-sues are partially solved at the cost of complex data preprocessing pipelines. In this paper, we present a highly accurate, resource-efficient deep learning-based occupancy detection solution, which is resilient to variations in humidity and temperature. The approach is tested on an extensive benchmark, where people are free to move and the furniture layout does change. In addition, based on a consolidated algorithm of explainable AI, we quantify the importance of the WiFi signal w.r.t. humidity and temperature for the proposed approach. Notably, humidity and temperature can indeed be predicted based on WiFi signals; this promotes the expressivity of the WiFi signal and at the same time the need for a non-linear model to properly deal with it.

Obesity is a real health emergency with significant consequences for individuals and society, reducing expectations and quality of life or even the death for 2.8 million people each year. Being a chronic condition with limited pharmacological options, weight loss therapies are mainly based on dietary and cognitive-behavioral interventions delivered at specialist public and/or private centers. In order to promote and facilitate weight loss, this paper proposes a virtual coaching system to motivate and guide the patients during the therapy and to support the clinicians in monitoring its effectiveness. The system has three components: a) the clinician's web application, which is used to monitor and modify the therapy for each patient; b) a mobile application for patients, used to log nutrition intakes and physical activities, and to consult motivational documents/videos; and c) a cloud server, which collects data and implements smart monitoring features. The platform is adopted in a clinical trial involving 120 patients with obesity (Body Mass Index 2> 30) that has recently started.

Over the last decade, Human Activity Recognition (HAR) has become a vibrant research field in various applications scenarios, ranging from sports, healthcare and well-being to smart cities, smart homes, and industry, mainly due to the widespread availability of devices as smartphones, smartwatches, and wearables. A key ingredient for sophisticated HAR systems is represented by the availability of high-quality datasets. These are generally gathered by dedicated Body Area Networks (BANs), and further elaborated through machine learning and deep learning algorithms. Thus, the BAN design plays a central role in such a context, where the main challenges are related to easiness of use, costs and energy constraints of their components. In this context, our paper presents a highly configurable HAR system, based on a low-cost and easy-to-use BAN. The system includes a CNN-based algorithm validated over a dataset, collected through the proposed BAN, on 12 persons performing 7 different human activities.

Identity recognition is increasingly used to control access to sensitive data, restricted areas in industrial, healthcare, and defense settings, as well as in consumer electronics. To this end, existing approaches are typically based on collecting and analyzing biometric data and imply severe privacy con-cerns. Particularly when cameras are involved, users might even reject or dismiss an identity recognition system. Furthermore, iris or fingerprint scanners, cameras, microphones, etc., imply installation and maintenance costs and require the user's active participation in the recognition procedure. This paper proposes a non-intrusive identity recognition system based on analyzing WiFi's Channel State Information (CSI). We show that CSI data attenuated by a person's body and typical movements allows for a reliable identification - even in a sitting posture. We further propose a lightweight deep learning algorithm trained using CSI data, which we implemented and evaluated on an embedded platform (i.e., a Raspberry Pi 4B). Our results obtained using real-world experiments suggest a high accuracy in recognizing people's identity, with a specificity of 98% and a sensitivity of 99%, while requiring a low training effort and negligible cost.