11 May 2023 in IEEE Transactions on Instrumentation and Measurement
Mental workload (MWL) assessment is crucial in fatigue evaluation applications to avoid potential health problems or serious accidents. This paper proposes an MWL recognition approach developed with a topological investigation of EEG-based brain functional connectivity (FC) network. In this work, the graph filtration-based features are extracted to reveal the brain state variations utilizing the persistent homology technique from the topological data analysis area. Three public open benchmark datasets are used to test and verify the recognition ability of the proposed method, which are developed with the MWL assessment experiments of Simultaneous Capacity (SIMKAP) test tasks, arithmetic calculations, and Multi-Attribute Task Battery II (MATB-II) MWL tasks. The experimental results show that the proposed topological FC network analysis scheme shows excellent distinguishing ability in brain state recognition, comparable to or better than the state-of-art results with similar settings. This work is the first investigation of EEG-based MWL evaluation with the persistent homology analysis of multivariate time series. The proposed topological features are effective and robust brain states indicators, providing an alternative feature in designing novel brain-computer interface systems.
3 May 2024 in IEEE Transactions on Consumer Electronics
The convergence of hardware, materials, sensors, and artificial intelligence brought a new class of consumer portable devices with advanced computing and connectivity capabilities, serving diverse aspects such as health monitoring, fitness tracking, and entertainment. These devices handle sensitive personal information and necessitate adequate security protection. Inertial measurement units (IMU) commonly integrated into smart wearables can be utilized to derive gait-based personal signatures. Consequently, ongoing research is focused on gait-based security, aiming to develop robust recognition methods that can adapt to various real-world conditions in terms of on-body IMU location and orientation, terrain type, and sensor signal utilisation. The present study addresses these challenges through phase-space analysis and topological data analysis. The proposed method has undergone validation using both a custom Gait-TNDA dataset and a public Gait-MIUS dataset, achieving mean accuracies surpassing 96%. Furthermore, its performance was compared with established deep learning methods. The results affirm that topological descriptors effectively capture gait dynamics, suggesting potential applications in the next generation of consumer wearables.
1 April 2025 in IEEE Transactions on Geoscience and Remote Sensing
Remote sensing change detection (RSCD) holds significant research value in remote sensing image processing. In recent years, many researchers have achieved remarkable results in RSCD tasks using methods based on CNNs or Transformers. Considering the limited receptive field of CNN models and the high computational cost of Transformers, many researchers have combined the two approaches, yielding promising results. However, most current RSCD-based models focus solely on change and temporal information, overlooking their complementary relationship. Additionally, some multi-scale feature fusion methods emphasize enhancing individual scales while neglecting the correlations between different scales. To address the above issues, we propose a multi-scale complementary enhanced context fusion network (MCECF). The network first introduces a global-local context aggregation module (GLCAM) to capture global-local context information while extracting multi-level feature maps. Subsequently, a complementary enhancement difference module (CEDM) is employed to complementarily aggregate the captured change and temporal information of bi-temporal remote sensing image features. To fully leverage the correlations between multi-scale features, a progressive decoder comprising a supervised spatial attention (SSA) mechanism and a multi-scale complementary enhanced fusion module (MCEFM) was developed. Moreover, to tackle the disparity between changed and unchanged regions, a dual-branch dynamic attention fusion module (DAFM) was designed to enhance the model’s adaptability to diverse scenarios. We conducted comparative experiments on five RSCD datasets against nine SOTA methods, and the results confirmed the effectiveness of the proposed MCECF in RSCD tasks. Our code will be made available at https://github.com/kakuqikaduo/MCECF.
26 December 2025 in Knowledge-Based Systems
11 March 2026 in Medical Image Analysis
21 February 2026 in Pattern Recognition