RESUMEN

To optimize the utilization and analysis of tables, it is essential to recognize and understand their semantics comprehensively. This requirement is especially critical given that many tables lack explicit annotations, necessitating the identification of column types and inter-column relationships. Such identification can significantly augment data quality, streamline data integration, and support data analysis and mining. Current table annotation models often address each subtask independently, which may result in the neglect of constraints and contextual information, causing relational ambiguities and inference errors. To address this issue, we propose a unified multi-task learning framework capable of concurrently handling multiple tasks within a single model, including column named entity recognition, column type identification, and inter-column relationship detection. By integrating these tasks, the framework exploits their interrelations, facilitating the exchange of shallow features and the sharing of representations. Their cooperation enables each task to leverage insights from the others, thereby improving the performance of individual subtasks and enhancing the model's overall generalization capabilities. Notably, our model is designed to employ only the internal information of tabular data, avoiding reliance on external context or knowledge graphs. This design ensures robust performance even with limited input information. Extensive experiments demonstrate the superior performance of our model across various tasks, validating the effectiveness of unified multi-task learning framework in the recognition and comprehension of table semantics.

RESUMEN

Deep Learning Automated Patient-Specific Quality Assurance (PSQA) aims to reduce clinical resource requirements. It is vital to ensure the safety and effectiveness of radiation therapy by predicting the dose difference metric (Gamma passing rate) and its distribution. However, current research overlooks uncertainty quantification in model predictions, limiting their trustworthiness in real clinical environments. This paper proposes a Multi-granularity Uncertainty Quantification (MGUQ) framework. A Bayesian framework that quantifies uncertainties at multiple granularities for multi-task PSQA, specifically Gamma Passing Rate (GPR) prediction and Dose Difference Prediction (DDP), integrates visualization-based interactive components. Using Bayesian theory, we derive a comprehensive multi-granularity loss function that comprises granularity-specific loss and coherence loss components. Additionally, we proposed Multi-granularity Prior Networks, a dual-stream network architecture, to infer the distributions of DDP (modeled as t-distributions) and GPR (modeled as Gaussian distributions) under specific statistical assumptions. Comprehensive evaluations are conducted on a dataset from ''Peeking Union Medical College Hospital'', and results show that our proposed method achieves a minimum MAE loss of 0.864 with a 2%/3 mm criterion and realizes the uncertainty visualization of dose difference. Further, it also achieves 100% Clinical Accuracy (CA) with a workload of 67.2%. Experiments demonstrate that the proposed framework can enhance the trustworthiness of deep learning applications in PSQA.

Asunto(s)

RESUMEN

Re-authentication continuously checks to see if a user is authorized during a whole usage session, enhancing secrecy capabilities for computational devices, especially against insider attacks. However, it is challenging to design a reliable re-authentication scheme with accuracy, transparency and robustness. Specifically, the approaches of using biometric features (e.g., fingerprint, iris) are often accurate in identifying users but not transparent to them due to the need for user cooperation. On the other hand, while the approaches exploiting behavior features (e.g., touch-screen gesture, movement) are often transparent in use, their applications suffer from low accuracy and robustness as behavior information collected is subjective and may change frequently over different use situations and even user's motion. In this paper, we propose BioTouch, a reliable re-authentication scheme that satisfies all the above requirements. First, BioTouch utilizes multiple features (finger capacitance and touching behavior) to identify the user for better accuracy. Second, BioTouch automatically works during user operation on capacitive-touch devices, achieving transparency without the need for manual assistance. Finally, by applying finger bio-capacitance, BioTouch is also robust to various conditions, as this feature is determined by the user's physical characteristics and will not change by different user positions and motions. We implement BioTouch for proof-of-concept and conduct comprehensive evaluations. The results show that BioTouch can flag 98% of anomalous behaviors within ten touching operations and achieve up to 99.84% accuracy during usage.

Asunto(s)

RESUMEN

The rapid development of Internet of Things (IoT) applications calls for light-weight IoT sensor nodes with both low-power consumption and excellent task execution efficiency. However, in the existing system framework, designers must make trade-offs between these two. In this paper, we propose an "edge-to-end integration" design paradigm, Butterfly, which assists sensor nodes to perform sensing tasks more efficiently with lower power consumption through their (high-performance) network infrastructures (i.e., a gateway). On the one hand, to optimize the power consumption, Butterfly offloads the energy-intensive computational tasks from the nodes to the gateway with only microwatt-level power budget, thereby eliminating the power-consuming Microcontroller (MCU) from the node. On the other hand, we address three issues facing the optimization of task execution efficiency. To start with, we buffer the frequently used instructions and data to minimize the volume of data transmitted on the downlink. Furthermore, based on our investigation on typical sensing data structures, we present a novel last-bit transmission and packaging mechanism to reduce the data amount on the uplink. Finally, we design a task prediction mechanism on the gateway to support efficient scheduling of concurrent tasks on multiple MCU-free Butterfly nodes. The experiment results show that Butterfly can speed up the task rate by 4.91 times and reduce the power consumption of each node by 94.3%, compared to the benchmarks. In addition, Butterfly nodes have natural security advantages (e.g., anti-capture) as they offload the control function with all application information up to the gateway.

Asunto(s)

RESUMEN

Recently, classification from compressed physiological signals in compressed sensing has been successfully applied to cardiovascular disease monitoring. However, in real-time wearable electrocardiogram (ECG) monitoring, it is very difficult to directly obtain the heartbeats information from compressed ECG signals. Thus arrhythmia classification from compressed ECG signals has to be handled in fixed-length segments instead of individual heartbeats. An inevitable issue is that a fixed-length ECG segment may contain multiple different types of arrhythmia. As a result, it is not appropriate to represent the multi-type real arrhythmia with a single label. In this paper, we first introduce multiple labels into fixed-length compressed ECG segments to challenge the arrhythmia classification issue. Then, we propose a deep learning model, which can directly classify multiple different types of arrhythmia from fixed-length compressed ECG segments with the advantages of low time cost for data processing and relatively high classification accuracy at a high compression ratio. Experimental results on the MIT-BIH arrhythmia database show that the exact match rate of our proposed method has reached 96.03% at CR(Compression Ratio)=70%, 94.99% at CR=80% and 93.19% at CR=90%.

Asunto(s)

RESUMEN

In this paper, we propose subcarrier allocation based cooperative spectrum sharing protocol for OFDM relaying networks with wireless energy harvesting. In the proposed protocol, the cognitive relay node utilizes different subcarriers to forward the primary information to obtain the spectrum access for the cognitive system transmission. The primary system consists of two parts, a primary transmitter (PT) and primary receiver (PR), and cognitive system includes a cognitive source node (CSN), cognitive destination node (CDN) and cognitive relay node (CRN). In the first phase, CRN splits a fraction of the power received from the PT and CSN transmission to decode information, while the remaining power is used for energy harvesting. Then CRN uses disjoint subcarriers to forward the signals of PT and CSN by utilizing the harvested energy in the second phase. Three parameters which consist of power splitting ratio, power allocation and subcarriers allocation are optimized in our algorithm to maximize the cognitive transmission rate with the constraint of primary target transmission rate. Numerical and simulation results are shown to give useful insights into the proposed cooperative spectrum sharing protocol, and we also found that various system parameters have a great effect for the simulation results.

RESUMEN

Location information plays a key role in pervasive computing and application, especially indoor location-based service, even though a mass of systems have been proposed, an accurate and practical indoor localization system remains unsettled. To tackle this issue, in this paper, we present a new localization scheme, SITE, combining acoustic Signals and Images to achieve accurate and robust indoor locaTion sErvice. Relying on a pre-deployed platform of acoustic sources with different frequencies, using proactively generated Doppler effect signals, SITE could track relative directions between the phone and the sources. Given m (m≥5) relative directions, SITE can use the angle differences to compute a set of locations corresponding to different subsets of sources. Then, based on a key observation-while the simultaneously estimated locations using different sets of acoustic anchors are within a small circle, the results converge to a point near the true location-SITE proposes a decision scheme that confirms whether these locations satisfy the demand of localization accuracy and can be used to search the user's location. If not, SITE utilizes VSFM(Visual Structure from Motion) technique to achieve a set of relative locations using some images captured by the phone's camera. By exploiting the synergy between the set of relative locations and the set of initial locations computed by relative directions, an optimal transformation relationship is obtained and applied to refine the initial calculated results. The refined result will be regarded as the user's location. In the evaluation, we implemented a prototype and deployed a real platform of acoustic sources in different scenarios. Experimental results show that SITE has excellent performance of localization accuracy, robustness and feasibility in practical application.

RESUMEN

Wearable telemonitoring of electrocardiogram (ECG) based on wireless body Area networks (WBAN) is a promising approach in next-generation patient-centric telecardiology solutions. In order to guarantee long-term effective operation of monitoring systems, the power consumption of the sensors must be strictly limited. Compressed sensing (CS) is an effective method to alleviate this problem. However, ECG signals in WBAN are usually non-sparse, and most traditional compressed sensing recovery algorithms have difficulty recovering non-sparse signals. In this paper, we proposed a fast and robust non-sparse signal recovery algorithm for wearable ECG telemonitoring. In the proposed algorithm, the alternating direction method of multipliers (ADMM) is used to accelerate the speed of block sparse Bayesian learning (BSBL) framework. We used the famous MIT-BIH Arrhythmia Database, MIT-BIH Long-Term ECG Database and ECG datasets collected in our practical wearable ECG telemonitoring system to verify the performance of the proposed algorithm. The experimental results show that the proposed algorithm can directly recover ECG signals with a satisfactory accuracy in a time domain without a dictionary matrix. Due to acceleration by ADMM, the proposed algorithm has a fast speed, and also it is robust for different ECG datasets. These results suggest that the proposed algorithm is very promising for wearable ECG telemonitoring.

Asunto(s)