Depression is a common, burdensome, often recurring mental health disorder that frequently goes undetected and untreated. Mobile phones are ubiquitous have an increasingly large complement of sensors can potentially be useful in monitoring behavioral patterns might indicative depressive symptoms.The objective this study was to explore the detection daily-life markers using mobile phone global positioning systems (GPS) usage sensors, their use identifying symptom severity.A total 40 adult...

A dramatic rise in health-tracking apps for mobile phones has occurred recently. Rich user interfaces make manual logging of users' behaviors easier and more pleasant, sensors tracking effortless. To date, however, feedback technologies have been limited to providing overall statistics, attractive visualization tracked data, or simple tailoring based on age, gender, calorie activity information. There are a lack systems that can perform automated translation behavioral data into specific...

Mobile sensing systems have made significant advances in tracking human behavior. However, the development of personalized mobile health feedback is still its infancy. This paper introduces MyBehavior, a smartphone application that takes novel approach to generate deeply feedback. It combines state-of-the-art behavior with algorithms are used recommendation systems. MyBehavior automatically learns user's physical activity and dietary strategically suggests changes those behaviors for...

Billions of IoT devices will be deployed in the near future, taking advantage faster Internet speed and possibility orders magnitude more endpoints brought by 5G/6G. With growth devices, vast quantities data that may contain users' private information generated. The high communication storage costs, mixed with privacy concerns, increasingly challenge traditional eco-system centralized over-the-cloud learning processing for platforms. Federated (FL) has emerged as most promising alternative...

As a privacy-preserving paradigm for training machine learning (ML) models, federated (FL) has received tremendous attention from both industry and academia. In typical FL scenario, clients exhibit significant heterogeneity in terms of data distribution hardware configurations. Thus, randomly sampling each round may not fully exploit the local updates heterogeneous clients, resulting lower model accuracy, slower convergence rate, degraded fairness, etc. To tackle client problem, various...

In this paper, we present DoppleSleep -- a contactless sleep sensing system that continuously and unobtrusively tracks quality using commercial off-the-shelf radar modules. provides single sensor solution to track sleep-related physical physiological variables including coarse body movements subtle fine-grained chest, heart due breathing heartbeat. By integrating vital signals movement sensing, achieves 89.6% recall with Sleep vs. Wake classification 80.2% REM Non-REM compared EEG-based...

In this paper, we propose BodyBeat, a novel mobile sensing system for capturing and recognizing diverse range of non-speech body sounds in real-life scenarios. Non-speech sounds, such as food intake, breath, laughter, cough contain invaluable information about our dietary behavior, respiratory physiology, affect. The BodyBeat consists custom-built piezoelectric microphone distributed computational framework that utilizes an ARM microcontroller Android smartphone. is designed to capture...

Mobile vision systems such as smartphones, drones, and augmented-reality headsets are revolutionizing our lives. These usually run multiple applications concurrently their available resources at runtime dynamic due to events starting new applications, closing existing application priority changes. In this paper, we present NestDNN, a framework that takes the dynamics of into account enable resource-aware multi-tenant on-device deep learning for mobile systems. NestDNN enables each model...

There is an undeniable communication barrier between deaf people and with normal hearing ability. Although innovations in sign language translation technology aim to tear down this barrier, the majority of existing systems are either intrusive or constrained by resolution ambient lighting conditions. Moreover, these can only perform single-sign ASL rather than sentence-level translation, making them much less useful daily-life scenarios. In work, we fill critical gap presenting DeepASL, a...

Recent advancements in deep neural networks (DNN) enabled various mobile learning applications. However, it is technically challenging to locally train a DNN model due limited data on devices like phones. Federated (FL) distributed machine paradigm which allows for training decentralized residing without breaching privacy. Hence, FL becomes natural choice deploying on-device the across intrinsically statistically heterogeneous (i.e., non-IID distribution) and usually have communication...

Over the past few years, Federated Learning (FL) has become an emerging machine learning technique to tackle data privacy challenges through collaborative training. In algorithm, clients submit a locally trained model, and server aggregates these parameters until convergence. Despite significant efforts that have been made FL in fields like computer vision, audio, natural language processing, applications utilizing multimodal streams remain largely unexplored. It is known broad real-world...

Soil monitoring plays an essential role in agricultural systems. Rather than deploying sensors' antennas above the ground, burying them soil is attractive way to retain a non-intrusive aboveground space. Low Power Wide-Area Network (LPWAN) has shown its long-distance and low-power features for Internet-of-Things (IoT) communication, presenting potential of extending underground cross-soil communication over wide area, which however not been investigated before. The variation conditions...

The clinical assessment of severity depressive symptoms is commonly performed with standardized self-report questionnaires, most notably the patient health questionnaire (PHQ-9), which are usually administered in a clinic. These questionnaires evaluate that stable over time. Ecologic

Video cameras have been deployed at scale today. Driven by the breakthrough in deep learning (DL), organizations that these start to use DL-based techniques for live video analytics. Although existing systems aim optimize analytics from a variety of perspectives, they are agnostic workload dynamics real-world deployments. In this work, we present Distream, distributed system based on smart camera-edge cluster architecture, is able adapt achieve low-latency, high-throughput, and scalable The...

Most cross-device federated learning (FL) studies focus on the model-homogeneous setting where global server model and local client models are identical. However, such constraint not only excludes low-end clients who would otherwise make unique contributions to training but also restrains from large due on-device resource bottlenecks. In this work, we propose FedRolex, a partial (PT)-based approach that enables model-heterogeneous FL can train larger than largest model. At its core, FedRolex...

This paper introduces ChirpTransformer, a versatile LoRa encoding framework that harnesses broad chirp features to dynamically modulate data, enhancing network coverage, throughput, and energy efficiency. Unlike the standard encoder offers only single configurable feature, our four distinct features, expanding spectrum of methods available for data modulation. To implement these on commercial off-the-shelf (COTS) nodes, we utilize combination software design hardware interrupt....

In the U.S., people spend approximately 90 percent of their time indoors. Unfortunately, indoor air quality (IAQ) may be two to five times worse than outdoors, and is often overlooked. Existing IAQ monitoring technologies focus on measurements visualization. However, lack information about pollution sources as well seriousness makes feel powerless frustrated, resulting in ignorance polluted at homes. this work, we fill critical gap by presenting AirSense, an intelligent home-based sensing...

Federated learning (FL) has gained substantial attention in recent years due to data privacy concerns related the pervasiveness of consumer devices that continuously collect from users. While a number FL benchmarks have been developed facilitate research, none them include audio and audio-related tasks. In this paper, we fill critical gap by introducing new benchmark for tasks which refer as FedAudio. FedAudio includes four representative commonly used datasets three important are well...

The scarcity of data presents a critical obstacle to the efficacy medical visionlanguage pre-training (VLP). A potential solution lies in combination datasets from various language communities. Nevertheless, main challenge stems complexity integrating diverse syntax and semantics, language-specific terminology, culture-specific implicit knowledge. Therefore, one crucial aspect consider is presence community bias caused by different languages. This paper novel framework named Unifying...

In cross-device Federated Learning (FL) environments, scaling synchronous FL methods is challenging as stragglers hinder the training process. Moreover, availability of each client to join highly variable over time due system heterogeneities and intermittent connectivity. Recent asynchronous (e.g., FedBuff [22]) have been proposed overcome these issues by allowing slower users continue their work on local based stale models contribute aggregation when ready. However, we show empirically that...

In the domain of cardiovascular healthcare, Electrocardiogram (ECG) serves as a critical, non-invasive diagnostic tool. Although recent strides in self-supervised learning (SSL) have been promising for ECG representation learning, these techniques often require annotated samples and struggle with classes not present fine-tuning stages. To address limitations, we introduce ECG-Text Pre-training (ETP), an innovative framework designed to learn cross-modal representations that link signals...

Informational support and nurturant are two basic types of social offered in online health communities. This study identifies the Q uitStop forum brings insights to exchange patterns user behaviors with content analysis network analysis. Motivated by information behavior, this defines describe exchange: initiated invited exchange. It is found that users a longer quitting time tend actively give support, recent quitters shorter abstinent likely seek receive support. also finds givers...