Silent, eyes-free text entry remains challenging when speech and conventional touch input are impractical. Prior wearable systems often required custom sensors or limited users to a small vocabulary. We present MorsEar, an IMU-only earable framework that maps near-ear micro-gestures such as taps for dot/dash; slide/pull/circle for space/delete/send into character-level Morse, enabling unrestricted character composition while using a compact lexicon solely for lightweight on-device autocorrect. The result is a low-bandwidth, reduced-exposure communication channel that works eyes-free and voice-free in accessibility scenarios, silent zones, and constrained environments. MorsEar infers words using a physics-aware preprocessing stack and compact CNN feed a tempo-adaptive segmentation with rolling buffers; an on-device decoder with lightweight autocorrect provides real-time feedback entirely on-phone. In a 24-participant study (with four accessibility users) across Silent, Cafe, and Metro, MorsEar achieved CER 7.3% and WER 12.5% → 7.8% (Autocorrect), with median 9.3/9.1/5.8 WPM, respectively. Similar to other accessibility-oriented encodings such as Braille, Morse requires a brief familiarization period to learn the timing and rhythm of dots and dashes; after which, MorsEar shows that commodity earable IMUs can support discreet, low-exposure text entry that scales beyond discrete commands to language-level interaction.

Wrist micro-behaviors during daily activities such as typing, handwriting, cooking, or carrying objects are valuable indicators for early detection of wrist disorders like Carpal Tunnel Syndrome and tendonitis. However, continuous personalized monitoring remains challenging without intrusive setups or hand-crafted rules. We present WristSense, a real-time, wrist-worn sensing system that introduces: (i) a magnetometer-stabilized quaternion fusion pipeline for orientation-agnostic tracking, (ii) a lightweight 1D-CNN + HMM model to distinguish functional gestures from strain-related coping behaviors, and (iii) an inertial tokenization scheme that converts events into structured prompts for a pretrained LLM. This enables zero-shot ergonomic feedback without per-user calibration. Evaluations across 12 participants show accurate posture tracking (< 10° MAE), high gesture recognition (macro F1 = 0.91), and improved usability (SUS = 85.2), with significantly higher user compliance compared to rule-based methods. WristSense demonstrates the potential of combining inertial sensing with LLMs for scalable, personalized ergonomic monitoring and early intervention.

We present SpineSense , an interactive earable-based framework for continuous monitoring of cervical spine posture and discomfort-related behavior in daily life. Leveraging inertial data from commercial off-the-shelf (COTS) earables (e.g., Apple AirPods Pro 2), SpineSense models the cervical vertebral chain (C1–C7) using SLERP-interpolated quaternions to estimate craniovertebral (CV) angles and identify early pain-relief gestures (e.g., neck rubbing, circular head rolls) that are associated with early signs of musculoskeletal fatigue, as informed by clinical observation. A real-time feedback loop delivers posture-based alerts and weekly compliance summaries to support user awareness and long-term posture correction. Central to our design is a clinician-in-the-loop methodology: clinical experts (including orthopedic surgeons and physiotherapists) informed threshold selection (e.g., CV angle cutoffs), interpreted common discomfort behaviors, and iteratively guided the structure of weekly feedback reports. We evaluate the system on 20 participants and achieve low spine angle estimation error (MAE: 0.876° , RMSE: 1.02° , r = 0.95), and discomfort gesture classification accuracy of F 1 = 0.97. Usability studies across diverse activities show high acceptance (SUS = 84.75, NASA-TLX = 32.7, PSSUQ = 2.13), with formal ANOVA tests validating statistically significant improvements over baseline interfaces. Together, our findings establish the feasibility of engineering interactive cervical health systems using COTS earables that support real-time feedback, clinician-informed reporting, and pervasive deployment in naturalistic settings.

Automated dietary monitoring is essential for gaining insights into eating behaviors, especially for managing chronic conditions such as obesity, diabetes, and hypercholesterolemia. Earable-based inertial sensing has been found promising for detecting chewing and eating activities; however, further insights like what, when, and how much is being eaten are crucial information for effective dietary assessment. Therefore, we propose BiteSense, an earable-based system that leverages inertial sensors (IMU) to monitor food intake and classify various food types. Using a hierarchical classification model, the system analyzes masticatory kinematics to detect food states, textures, nutritional value, and cooking methods, ultimately identifying specific foods consumed, as well as estimating food intake amount and meal type. A semi-controlled user study involving 38 participants from diverse backgrounds demonstrated the system’s high accuracy, with an F1 score of 0.86 for detecting the masticatory process using a leave-one-subject-out (LOSO) approach, while exhibiting significant improvement over benchmark algorithms in extensive experiments by 8-12%.

The ubiquitousness of smart and wearable devices with integrated acoustic sensors in modern human lives presents tremendous opportunities for recognizing human activities in our living spaces through ML-driven applications. However, their adoption is often hindered by the requirement of large amounts of labeled data during the model training phase. Integration of contextual metadata has the potential to alleviate this since the nature of these meta-data is often less dynamic (e.g. cleaning dishes, and cooking both can happen in the kitchen context) and can often be annotated in a less tedious manner (a sensor always placed in the kitchen). However, most models do not have good provisions for the integration of such meta-data information. Often, the additional metadata is leveraged in the form of multi-task learning with sub-optimal outcomes. On the other hand, reliably recognizing distinct in-home activities with similar acoustic patterns (e.g. chopping, hammering, knife sharpening) poses another set of challenges. To mitigate these challenges, we first show in our preliminary study that the room acoustics properties such as reverberation, room materials, and background noise leave a discernible fingerprint in the audio samples to recognize the room context and proposed AcouDL as a unified framework to exploit room context information to improve activity recognition performance. Our proposed self-supervision-based approach first learns the context features of the activities by leveraging a large amount of unlabeled data using a contrastive learning mechanism and then incorporates this feature induced with a novel attention mechanism into the activity classification pipeline to improve the activity recognition performance. Extensive evaluation of AcouDL on three datasets containing a wide range of activities shows that such an efficient feature fusion-mechanism enables the incorporation of metadata that helps to better recognition of the activities under challenging classification scenarios with 0.7-3.5% macro F1 score improvement over the baselines.

Identifying risky driving behavior is crucial for early hazard detection, encouraging safer driving practices, and minimizing accident risks. Driving patterns, characterized by sensory data, can be used to classify risky behavior. However, effective classification into risk categories relies on supervised learning methods that require labeled data. The challenge lies in the high cost and difficulty of obtaining accurate groundtruth labels for these signatures. As a result, most datasets lack risk labels. Additionally, because risky incidents are infrequent during regular driving, the dataset collected from studies becomes imbalanced, containing fewer instances of risky events. This imbalance poses a significant challenge, as it biases the model towards the majority class, increasing the likelihood of costly misclassifications where risky instances are incorrectly identified as safe. To address this, we propose a three-stage method. First, we identify driving events indicative of risky behavior from the trajectory data and mathematically formulate them as potential risk indicators. Using these indicators, we then employ clustering to assign appropriate risk labels to the data. Finally, we tackle the class imbalance problem using a cost-sensitive LSTM model that combines a custom loss function with LSTM architecture to prioritize accurately classifying risky instances. Our method outperforms other state-of-the-art approaches with high accuracy, precision, F1 score, and recall of 98.13%, 95.2%, 96.6%, and 96.35%, respectively, effectively managing an imbalanced dataset.

Frequent disruption in network connectivity is a major challenge in offering good quality of experience to the users of smart mobile applications. Various apps like Zoom, Google Meet, Skype, etc. for online video calling have become indispensable overnight due to this new paradigm shift in home-bound remote work culture driven by the Covid19 pandemic. However, the performance of these apps is tightly bound to the current network conditions. Under limited network coverage and low bandwidth, the data frames suffer from delay, jitter, and loss resulting in degraded quality of experience for the users. In this paper, we propose a server-less peer-to-peer architecture for the video conferencing apps with in-built adaptive compression techniques. The proposed architecture enables video streaming at a very low data rate just to offer a smooth streaming experience under poor connectivity. Video is compressed by ASCII encoding on the basis of the contrast factor of each pixel in the frame. It has been observed that only 29 KBPS bandwidth is sufficient to conduct video conferencing.

Augmented Reality (AR) and Virtual Reality (VR) games are some of the emerging use cases of 5G in the area of ultra-Reliable and Low Latency Communications (uRLLC). A multiplayer AR/VR game broadly consists of compute-intensive tasks which convert the raw data generated from sensory sources such as wearables, smartphones, etc., to action data such as location, orientation, intention, etc., and services that process the action data. Services generate a common response to all players by taking action data as input. The total response time must be as low as 20 milliseconds for a good user experience and to prevent motion sickness. While considering these aspects, the multiplayer game must be scalable, and users should be able to move. Multi-access edge computing (MEC) helps to improve performance by partially/fully offloading such tasks from mobile devices and latency-sensitive services from the cloud to a server at the edge called the MEC host. We propose, for the first time, an online mobility-aware heuristic in a Multi-access Edge Computing Network (MEN) to reduce the response time, specifically the Game Frame Time (GFT), consistently, for an improved Quality of Experience (QoE), for such games. This is done by jointly offloading tasks and placing services, and migrating both whenever required. Additionally, for improved response, the network is partitioned into regions, and a service instance is placed on a MEC host, called the Region Coordinator (RC), in each region, in a decentralized manner. When a new player joins, an old player leaves, or old players move, the number of players and their mobility patterns change in a particular region. This may require allocating or moving tasks from one MEC host to another and migrating services to a new RC. While tasks and services are migrated, the associated state and data must be moved to the destination MEC host. Our experiments demonstrate that the standard deviation for the mean GFT is 0 ms in the best case and 9.26 ms in the worst case, providing a uniform user experience, even when mobility is as high as 50% (it means 50% of the players are moving). When there is mobility, the GFT increases by 28.29% in the best case and 37.18% in the worst case, compared to a no-mobility scenario. We also demonstrate that, given computing power, there is a tradeoff between responsiveness and GFT.

Wireless sensor networks have grown significant attentions among researchers for providing a flexible and low-cost framework to design an architecture for Intelligent Transport Systems. The inherent challenges in distribution and management of sensor networks along the road require an application-specific protocol support for the network connectivity, the sensing coverage, the reliable data forwarding, and the network lifetime improvement. This paper introduces the concept of k-strip length coverage along the road, which ensures a better sensing coverage for the detection of moving vehicles compared with the conventional barrier coverage and full area coverage, in terms of the availability of sufficient information for statistical processing and the number of sensors required to be active. To extend the network lifetime, every sensor follows a sleep-wakeup schedule maintaining the network connectivity and the k-strip length coverage. This scheduling problem is modeled as a graph optimization, the NP-hardness of which motivates to design a centralized heuristic, providing an approximate solution. As a sensor network is inherently distributed in nature, properties of the centralized heuristic are explored to design a per-node solution based on local information. Performance of the proposed scheme is analyzed through simulation results.

Data gathering or converge cast is one of the most popular applications of road side sensor network where the data sensed from the road are accumulated in the road side gateways or sinks for traffic monitoring purpose. The required delay sensitivity and reliability of the application as well as the scarcity of sensor resources make the task challenging. In this paper, a novel tree based data gathering scheme has been proposed exploiting the strip like structure of the road network. Sensor nodes are distributed in several virtual blocks along the road and a converge cast tree is constructed selecting one active node from each block. Implementation of efficient scheduling assures both the coverage and critical power savings of sensor nodes. The network connectivity is guaranteed throughout by the proposed tree maintenance module that handles the sensor node joining and leaving events. Simulation results show that the tree maintenance overhead in terms of both delay and control message communication is nominal.

Sensor networks deployed for the border area monitoring requires a high degree of reliability for the data gathering in spite of any arbitrary node or sink failures. This paper proposes RelBAS, a robust data gathering scheme specially designed for the border area network to provide a guaranteed delivery of sensory data. The proposed protocol aims to find out multiple node-disjoint paths to multiple sinks so that the disconnectivity in one path due to a node failure does not disrupt the delivery of data to the sink. The forwarding path selection at every node in RelBAS is based on the combination of three parameters - the hop-count, the residual energy and the number of children for for parent of the corresponding tree. This helps in adapting the protocol to the application requirement depending on the delay, energy efficiency and data aggregation. Moreover, RelBAS is capable of detecting an affected zone due to multiple node failures. The effectiveness of the proposed scheme has been analyzed using the simulation results.

The lifetime of sensor network depends on the efficient utilization of resource-constrained sensor nodes. Several MAC protocols like DMAC and its variants have been proposed to save critical sensor resources through sleep-wakeup scheduling over data gathering tree. For applications where data aggregation is not possible, the sleep duration decreases gradually from the leaves to the root of the data gathering tree. This results early failure of sensor nodes near the sink, and affects network connectivity and coverage. Deploying redundant sensors can solve this problem where a faulty node is replaced by a redundant node to maintain network connectivity and coverage. However, the amount of redundancy depends on the node failure pattern, and thus more number of redundant nodes required to be deployed near the sink. This paper proposes a gradient based sensor deployment scheme for energy-efficient data gathering exploring the trade-off among connectivity, coverage, fault-tolerance and redundancy. The density of deployment is estimated based on the distance of a node from the sink while dealing with connectivity, coverage and fault-tolerance. The effectiveness of the proposed scheme has been analyzed both theoretically and with the help of simulation.

IEEE 802.11s provides Hybrid Wireless Mesh Protocol (HWMP) to find out the forwarding path in a mesh network based on mesh peering and MAC layer scheduling information. However, both proactive and reactive modes of HWMP perform poorly for multi-radio mesh network because of inefficient radio selection, time-varying channel conditions and interference among the radios. This paper proposes an improved opportunistic path selection protocol over HWMP for multi-radio support that goes beyond the traditional routing mechanisms, operates either in proactive, reactive or hybrid mode. The efficiency of the proposed scheme is analyzed using simulation results.

Event driven data gathering or convergecast through sensor nodes requires efficient and correct delivery of data at the sink. A tree rooted at the sink is an ideal topology for data gathering which utilizes sensor resources properly. Resource constrained sensor nodes are highly prone to sudden crash. A set of algorithms, proposed in this paper, builds a data gathering tree rooted at the sink. The tree eventually becomes a Breadth First Search (BFS) tree where each node maintains the shortest distance in hop-count to the root to reduce the routing delay and power consumption. The data gathering tree is repaired locally within a constant round of message transmissions after any random node fails. Simulation result shows that the repairing delay is very less in average, and the proposed scheme can repair from arbitrary node failure using constant number of message passing.