Selected work

Representative Papers

Real-Time Automatic License Plate Recognition Using YOLOv8, SORT Tracking, and Temporal Data Interpolation

Jun 03, 2026

This study addresses the challenges of low license plate recognition accuracy and fragmented tracking trajectories in dynamic traffic surveillance, caused by abrupt illumination changes, extreme viewing angles, high-speed motion, and occlusions. To tackle these issues, the authors propose a five-stage end-to-end pipeline that leverages YOLOv8-nano for joint vehicle and license plate detection, integrates the SORT algorithm for multi-object tracking, and introduces an innovative offline temporal bounding box interpolation mechanism to recover broken trajectories. Furthermore, the framework incorporates location-guided OCR (EasyOCR) fused with a license plate syntax validation module to enhance recognition accuracy and spatiotemporal consistency under complex conditions. Experimental results demonstrate that the proposed approach significantly improves both the continuity of license plate tracking and the robustness of recognition in highly challenging dynamic scenarios.

0 citationsRead paper

SasMamba: A Lightweight Structure-Aware Stride State Space Model for 3D Human Pose Estimation

Nov 12, 2025

Existing state-space model (SSM)-based methods for 3D human pose estimation rely on manual flattening of 2D pose sequences into unstructured time series, which destroys skeletal topology and conflates spatiotemporal features. To address this, we propose a structure-aware stride-wise SSM architecture: (1) a structure-aware spatiotemporal convolution explicitly models local joint interactions and global pose constraints; (2) a stride-wise multi-scale scanning strategy decouples and jointly captures spatiotemporal dependencies while preserving linear complexity; and (3) a lightweight Mamba backbone enables efficient global modeling. Our approach achieves state-of-the-art performance on benchmarks including Human3.6M, while reducing parameter count by over 40% compared to leading hybrid models. It is the first to unify structural fidelity—preserving anatomical relationships—with computational efficiency in SSM-based 3D pose estimation.

0 citationsRead paper

Physics-Informed Neural Networks for Speech Production

Nov 01, 2025

This work addresses key challenges in speech production modeling—namely, non-differentiability and gradient vanishing induced by vocal fold collisions, and the unknown fundamental period of self-sustained oscillation. To this end, we propose a physics-informed neural network (PINN) framework tailored for the glottis–vocal-tract coupled system. Methodologically: (i) a differentiable approximation function is introduced to model vocal fold collisions; (ii) the oscillation period is treated as a learnable parameter to accommodate inter-subject variability; and (iii) hard constraints are enforced to directly embed the physical coupling between glottal airflow and vocal-tract acoustics, eliminating error accumulation inherent in soft-constraint formulations. Experiments demonstrate that the unified architecture jointly infers glottal flow rate, vocal fold vibration dynamics, and subglottal pressure. Moreover, it supports both forward speech synthesis and inverse physiological parameter estimation, significantly improving physical consistency and cross-speaker generalization.

0 citationsRead paper

A Sequential Computation Algorithm for the Center of the Smallest Enclosing Ball

May 31, 2025

This paper addresses the efficient computation of the center ( Q^* ) of the smallest enclosing ball (SEB) of ( n ) points in ( mathbb{R}^d ), with particular relevance to machine learning applications such as support vector data description (SVDD). We propose a novel sequential algorithm inspired by the Arimoto–Blahut iterative framework from information theory, which directly computes the barycentric coordinate representation of ( Q^* ) with respect to the input point set. To our knowledge, this is the first application of the Arimoto–Blahut paradigm to SEB computation. The algorithm employs a compact, coordinate-only recursive update rule that uniformly handles both cases—with and without equidistant points—without requiring geometric preprocessing or active-set identification. It exhibits theoretical convergence guarantees and numerical stability, with computational complexity ( O(kappa n^2 log(1/varepsilon)) ), where ( kappa ) denotes the number of iterations. Empirical evaluation confirms high accuracy, controlled runtime, exceptional simplicity of implementation, and straightforward extensibility.

0 citationsRead paper

HGMamba: Enhancing 3D Human Pose Estimation with a HyperGCN-Mamba Network

Apr 09, 2025

To address insufficient accuracy in 3D human pose estimation from realistic 2D pose inputs, this paper proposes Dual-Stream HGMamba: a novel framework that constructs multi-granularity human hypergraphs via Hyper-GCN to explicitly model high-order local joint dependencies, and integrates Shuffle Mamba—a state-space model-based temporal scanning module—to efficiently capture global spatiotemporal dynamics. It represents the first deep integration of hypergraph modeling with state-space models, enabling configurable trade-offs between accuracy and inference speed through scalable model variants. Extensive experiments demonstrate state-of-the-art performance on Human3.6M and MPI-INF-3DHP, achieving P1 errors of 38.65 mm and 14.33 mm, respectively. The code and pretrained models are publicly released.

0 citationsRead paper

Recent publications

Latest Papers

Real-Time Automatic License Plate Recognition Using YOLOv8, SORT Tracking, and Temporal Data Interpolation

Jun 03, 2026

This study addresses the challenges of low license plate recognition accuracy and fragmented tracking trajectories in dynamic traffic surveillance, caused by abrupt illumination changes, extreme viewing angles, high-speed motion, and occlusions. To tackle these issues, the authors propose a five-stage end-to-end pipeline that leverages YOLOv8-nano for joint vehicle and license plate detection, integrates the SORT algorithm for multi-object tracking, and introduces an innovative offline temporal bounding box interpolation mechanism to recover broken trajectories. Furthermore, the framework incorporates location-guided OCR (EasyOCR) fused with a license plate syntax validation module to enhance recognition accuracy and spatiotemporal consistency under complex conditions. Experimental results demonstrate that the proposed approach significantly improves both the continuity of license plate tracking and the robustness of recognition in highly challenging dynamic scenarios.

0 citationsRead paper

SasMamba: A Lightweight Structure-Aware Stride State Space Model for 3D Human Pose Estimation

Nov 12, 2025

Existing state-space model (SSM)-based methods for 3D human pose estimation rely on manual flattening of 2D pose sequences into unstructured time series, which destroys skeletal topology and conflates spatiotemporal features. To address this, we propose a structure-aware stride-wise SSM architecture: (1) a structure-aware spatiotemporal convolution explicitly models local joint interactions and global pose constraints; (2) a stride-wise multi-scale scanning strategy decouples and jointly captures spatiotemporal dependencies while preserving linear complexity; and (3) a lightweight Mamba backbone enables efficient global modeling. Our approach achieves state-of-the-art performance on benchmarks including Human3.6M, while reducing parameter count by over 40% compared to leading hybrid models. It is the first to unify structural fidelity—preserving anatomical relationships—with computational efficiency in SSM-based 3D pose estimation.

0 citationsRead paper

Physics-Informed Neural Networks for Speech Production

Nov 01, 2025

This work addresses key challenges in speech production modeling—namely, non-differentiability and gradient vanishing induced by vocal fold collisions, and the unknown fundamental period of self-sustained oscillation. To this end, we propose a physics-informed neural network (PINN) framework tailored for the glottis–vocal-tract coupled system. Methodologically: (i) a differentiable approximation function is introduced to model vocal fold collisions; (ii) the oscillation period is treated as a learnable parameter to accommodate inter-subject variability; and (iii) hard constraints are enforced to directly embed the physical coupling between glottal airflow and vocal-tract acoustics, eliminating error accumulation inherent in soft-constraint formulations. Experiments demonstrate that the unified architecture jointly infers glottal flow rate, vocal fold vibration dynamics, and subglottal pressure. Moreover, it supports both forward speech synthesis and inverse physiological parameter estimation, significantly improving physical consistency and cross-speaker generalization.

0 citationsRead paper

A Sequential Computation Algorithm for the Center of the Smallest Enclosing Ball

May 31, 2025

This paper addresses the efficient computation of the center ( Q^* ) of the smallest enclosing ball (SEB) of ( n ) points in ( mathbb{R}^d ), with particular relevance to machine learning applications such as support vector data description (SVDD). We propose a novel sequential algorithm inspired by the Arimoto–Blahut iterative framework from information theory, which directly computes the barycentric coordinate representation of ( Q^* ) with respect to the input point set. To our knowledge, this is the first application of the Arimoto–Blahut paradigm to SEB computation. The algorithm employs a compact, coordinate-only recursive update rule that uniformly handles both cases—with and without equidistant points—without requiring geometric preprocessing or active-set identification. It exhibits theoretical convergence guarantees and numerical stability, with computational complexity ( O(kappa n^2 log(1/varepsilon)) ), where ( kappa ) denotes the number of iterations. Empirical evaluation confirms high accuracy, controlled runtime, exceptional simplicity of implementation, and straightforward extensibility.

0 citationsRead paper

HGMamba: Enhancing 3D Human Pose Estimation with a HyperGCN-Mamba Network

Apr 09, 2025

To address insufficient accuracy in 3D human pose estimation from realistic 2D pose inputs, this paper proposes Dual-Stream HGMamba: a novel framework that constructs multi-granularity human hypergraphs via Hyper-GCN to explicitly model high-order local joint dependencies, and integrates Shuffle Mamba—a state-space model-based temporal scanning module—to efficiently capture global spatiotemporal dynamics. It represents the first deep integration of hypergraph modeling with state-space models, enabling configurable trade-offs between accuracy and inference speed through scalable model variants. Extensive experiments demonstrate state-of-the-art performance on Human3.6M and MPI-INF-3DHP, achieving P1 errors of 38.65 mm and 14.33 mm, respectively. The code and pretrained models are publicly released.

0 citationsRead paper