Author name cluster

Xia Zhang

Possible papers associated with this exact author name in Arrow. This page groups case-insensitive exact name matches and is not a full identity disambiguation profile.

7 papers

2 author rows

JBHI Journal 2026 Journal Article

AI-Driven Health Monitoring: Integrating Transformer and Convolutional Fusion for Stroke Patient Posture Estimation

Xia Zhang
Fangpeng Jin
Jing Hu
Jiang Xu

Rehabilitative exercises are crucial for the motor recovery of stroke patients. Traditional physical therapy involves various techniques and exercise therapies aimed at helping patients regain function and enhance their quality of life. However, effective treatment through this method requires substantial human resources and can be costly. It also tends to be subjective and lacks real-time responsiveness. With the rapid development of AI-generated content technology, to improve the precision and real-time performance of rehabilitative motion posture recognition and to better assist patients in conducting extended rehabilitation exercises independently at home, this paper proposes an AI-driven network architecture for posture estimation in stroke patient rehabilitation that can generate corresponding posture skeleton content by identifying the patient's posture. This generated content feedback can assist in monitoring and treating stroke patients' rehabilitation. The architecture integrates spatial convolutional layers with an improved transformer module, termed ConvTrans (Integration of Convolution and Transformer). In the ConvTrans block, the use of lightweight multi-head self-attention (LMHSA) and inverted residual forward networks (IRFFN) effectively reduces computational costs and enhances processing efficiency. This configuration captures both information pertaining to local and global structure, thereby enhancing the network's representational capabilities. Demonstrating strong performance on three challenging human pose estimation (HPE) datasets, this method offers efficient feedback on posture to aid in the rehabilitation of individuals affected by strokes.

Details DOI

JBHI Journal 2025 Journal Article

An Enhanced Autoencoder-Based Anomaly Detection Model for Time Series Data From Wearable Medical Devices

Zhiyuan Li
Xia Zhang

In today's era of rapidly advancing technologies, such as sensors and the Internet of Things (IoT), and the increasing focus on promoting healthy lifestyles, smart wearable devices play a crucial role in real-time detection and diagnosis of physical health conditions. Through analyzing the multi-featured time series data captured by these devices with multiple sensors, we can uncover hidden diseases and provide timely treatment. Therefore, it is imperative to study an anomaly detection model with robust feature learning and anomaly diagnosis capabilities. To address this need, this paper proposes an enhanced autoencoder-based anomaly detection model for time series data obtained from wearable medical devices. Initially, the model utilizes a convolutional neural network to learn the correlations between multiple features. Subsequently, a long and short-term memory network is employed to capture the sequence correlations, and an multi-head attention mechanism is used to mitigate the performance degradation caused by increasing the sequence length. The residual loss is also used to effectively mitigate the vanishing gradient problem. Finally, the model is evaluated using two widely recognized public datasets: the Heart Disease dataset, which contains information on patients with heart conditions, and the MIMIC dataset, a comprehensive database of de-identified health data related to critical care. The experimental results demonstrate that our model can achieve an accuracy of 95. 37% and 95. 56% on the two datasets, respectively. Compared to the best performing baseline methods, our model improves 8. 6% and 12. 3% on the two datasets, respectively. Overall, our model enables efficient analysis of sequential data, effectively captures long-term dependencies, and significantly improves the success rate of early health diagnosis for individuals.

Details DOI

AIIM Journal 2025 Journal Article

HMEA: A hierarchical medical knowledge graph entity alignment model fusing multi-aspect information

Weiguang Wang
Lijuan Ma
Wei Cai
Haiyan Zhao
Xia Zhang

Medical entity alignment is crucial for the integration and reasoning of medical knowledge, aiming to match semantically equivalent entities across different medical knowledge graphs. Unlike entities in general knowledge graphs, medical entities contain rich multi-aspect information, which not only includes structural and attribute information but also additional information such as ontology and descriptions. However, existing entity alignment methods overlook these additional pieces of information and lack exploration into the fusion of multi-aspect information. This leads to less-than-ideal performance in medical entity alignment. To address the aforementioned issues, in this paper, we propose a hierarchical medical knowledge graph entity alignment method, termed HMEA, which integrates multi-aspect information. Firstly, we represent the medical knowledge graph as a hierarchical heterogeneous graph to model the multi-aspect information of medical entities. Secondly, we design different representation learning methods according to the characteristics of multi-aspect information to obtain vector representations of entities in different dimensions. Subsequently, we devise a two-stage multi-aspect knowledge fusion mechanism to dynamically integrate multi-aspect information, enabling mutual complementarity. Finally, we utilize the fused entity vector representations to guide entity alignment. We compare our approach with state-of-the-art baseline models on ten different types of publicly available datasets and further conduct ablation and parameter analyses. Experimental results validate the effectiveness and robustness of the proposed model. In benchmark tests across all datasets, HMEA outperforms the current state-of-the-art methods significantly.

Details DOI

TIST Journal 2025 Journal Article

Integrated Hybrid Transformer and Multi-Receptive Feature Extraction Mechanism for Electrocardiogram Denoising Using Score-Based Diffusion Model

Baofeng Zhu
Wanjun Cheng
Xia Zhang
Jiren Liu

Electrocardiogram (ECG) is the foundation of the analysis of cardiac disease. In the hospital clinical ECG diagnostic scenarios, when doctors analyze ECG signals or when an ECG intelligent diagnostic system is used, there might be strong noises like baseline wander or muscle artifact in the ECG signals due to the unstable state of the subjects, and such interferences are usually difficult to be filtered out by traditional filters, which can lead to serious errors in the subsequent signal analysis. To solve this problem, we propose a novel network which integrates hybrid transformer and multi-receptive feature extraction mechanism into score-based diffusion model. We used score-based diffusion model to reconstruct the clean ECG signals from noisy ones. The experiment was conducted on the QT Database and the MIT-BIH Noise Stress Test Database to verify the feasibility of our method. Baseline methods are used for comparison. The evaluation results show that our method can achieve an outstanding performance on four distance-based evaluation metrics by at least 26% overall improvement in the comparison with the best baseline method. The study demonstrates that the signal denoising and reconstruction method based on the self-designed score-based diffusion model can effectively remove the interferences in the ECG signals, thereby facilitating the subsequent diagnosis in real-world situation. It also has huge potential for establishing the ECG intelligent analysis system.

Details DOI

AAAI Conference 2024 Conference Paper

CoLAL: Co-learning Active Learning for Text Classification

Linh Le
Genghong Zhao
Xia Zhang
Guido Zuccon
Gianluca Demartini

In the machine learning field, the challenge of effectively learning with limited data has become increasingly crucial. Active Learning (AL) algorithms play a significant role in this by enhancing model performance. We introduce a novel AL algorithm, termed Co-learning (CoLAL), designed to select the most diverse and representative samples within a training dataset. This approach utilizes noisy labels and predictions made by the primary model on unlabeled data. By leveraging a probabilistic graphical model, we combine two multi-class classifiers into a binary one. This classifier determines if both the main and the peer models agree on a prediction. If they do, the unlabeled sample is assumed to be easy to classify and is thus not beneficial to increase the target model's performance. We prioritize data that represents the unlabeled set without overlapping decision boundaries. The discrepancies between these boundaries can be estimated by the probability that two models result in the same prediction. Through theoretical analysis and experimental validation, we reveal that the integration of noisy labels into the peer model effectively identifies target model's potential inaccuracies. We evaluated the CoLAL method across seven benchmark datasets: four text datasets (AGNews, DBPedia, PubMed, SST-2) and text-based state-of-the-art (SOTA) baselines, and three image datasets (CIFAR100, MNIST, OpenML-155) and computer vision SOTA baselines. The results show that our CoLAL method significantly outperforms existing SOTA in text-based AL, and is competitive with SOTA image-based AL techniques.

PDF Details DOI

ICRA Conference 2011 Conference Paper

Inhibitory connections between neural oscillators for a robotic suit

Xia Zhang
Minoru Hashimoto

A new synchronization-based control is proposed for a robotic suit which is designed for walking assistance. Neural oscillators are connected to each joint of the robotic suit to synchronize suit's movement with human user's movement (outer-synchronization). At the same time, mutual inhibition is incorporated between neural oscillators at each left and right joint of the suit to help to maintain a human-gait-like anti-phase relationship (inner-inhibition). We have developed a 2-DOF robotic suit, which consists of two actuators fixed to the places where the human hip joints are. Each actuator has a built-in torque sensor, which helps to measure the mutual joint torque generated by once there is any difference between the movement of a user and that of the suit. The mutual joint torque is used as input to neural oscillators, which control each actuator of the robotic suit through outer-synchronization with the mutual joint torque and inner-inhibition. The inhibitory weight, used to adjust the inhibitory strength between neural oscillators, is designed properly using series of simulations. We have conducted walking experiments to show the validity of our proposal for walking assistance of the robotic suit with mutual inhibition between neural oscillators.

Details

ICRA Conference 2009 Conference Paper

SBC for motion assist using neural oscillator

Xia Zhang
Minoru Hashimoto

In this paper we propose a framework for synchronization based control (SBC) using neural oscillators for motion assist. A neural oscillator is used to accomplish synchronization and entrainment between periodic motions by the human and robot. The mutual joint torque between the human and robot is used as an external input signal for the neural oscillator, which generates the desired trajectory of a robot joint angle, so that the robot motion synchronizes with the external mutual joint torque. The validity and feasibility of the proposed method is examined from three points of view. The first is whether synchronization of action between human and robot can be realized. The second is whether the assist effect can be obtained, and the third is whether the proposed method has an acceptable level of usability for the user. We explored those three points of view by conducting computer simulations on a human-motion assist system and experiments with a joint torque sensing robot suit.

Details