Chakraborty, Suchetana

2024, Anirban Das, Navlika Singh, Chakraborty, Suchetana

Next-generation smart city applications, attributed to the power of the Internet of Things (IoT) and Cyber–Physical Systems (CPS), significantly rely on sensing data quality. With an exponential increase in intelligent applications for urban development and enterprises offering sensing-as-a-service these days, it is imperative that a shared sensing infrastructure could thwart the better utilization of resources. However, a shared sensing infrastructure that leverages low-cost sensing devices for a cost-effective solution remains unexplored territory. A significant research effort is still needed to make edge-based data shaping solutions more reliable, feature-rich, and cost-effective while addressing the associated challenges in sharing the sensing infrastructure among multiple collocated services with diverse Quality of Service (QoS) requirements. Towards this, we propose UniPreCIS, a novel edge-based data preprocessing solution that accounts for the inherent characteristics of low-cost ambient sensors and their exhibited measurement dynamics concerning application-specific QoS. UniPreCIS aims to identify and select quality data sources by performing sensor ranking and selection that dynamically adapts to the change in sensor attributes. Finally, multimodal data preprocessing is performed in a unified manner to meet heterogeneous application QoS and, at the same time, reduce the resource consumption footprint for the resource-constrained network edge. We study the effectiveness of UniPreCIS on a real-world testbed deployed on our campus. As observed, the processing time and memory utilization of the stakeholder services have been reduced in the proposed approach while achieving up to 90% accuracy, which is arguably significant compared to state-of-the-art sensing techniques.

2024, Osho, Pranay, Pradeep Kumar, Chakraborty, Suchetana

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.

2024, Avijoy Chakma, Anirban Das, Abu Zaher Md Faridee, Chakraborty, Suchetana, Sandip Chakraborty, Nirmalya Roy

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.

2024, Monik Raj Behera, Chakraborty, Suchetana

Conventional federated learning frameworks suffer from several challenges including performance bottlenecks at the central aggregation server, data bias, poor model convergence, and exposure to model poisoning attacks, and limited trust in the centralized infrastructure. In the current paper, a novel game theory-based approach called 'pFedGame' is proposed for decentralized federated learning, best suitable for temporally dynamic networks. The proposed algorithm works without any centralized server for aggregation and incorporates the problem of vanishing gradients and poor convergence over temporally dynamic topology among federated learning participants. The solution comprises two sequential steps in every federated learning round, for every participant. First, it selects suitable peers for collaboration in federated learning. Secondly, it executes a two-player constant sum cooperative game to reach convergence by applying an optimal federated learning aggregation strategy. Experiments performed to assess the performance of pFedGame in comparison to existing methods in decentralized federated learning have shown promising results with accuracy higher than 70% for heterogeneous data.