Chao-Chun Chen

The “Hospital at Home” initiative transforms medical service automation through modern technologies. This letter revisits remote physiotherapy, allowing convalescents to record exercises using mobile devices from arbitrary angles. To address this, we propose a physiotherapy video matching method that accurately aligns movements from unconstrained viewpoints. The task is formulated as an optimization problem and solved using a modular pipeline. We introduce the Angle-of-Limb-based Posture Structure (ALPS) and the Camera-Angle-Free (CAFE) transformation to counter camera-angle differences. We also develop the Three-phase ALPS Matching Algorithm (TALMA) for matching movements between mentor and convalescent videos. Real-world experiments show our method outperforms existing solutions in both precision and practicality, with a time deviation of less than 0.07 seconds from expert annotations.

In many industries, intelligent mobile unmanned vehicles (MUVs) are transforming automation and enhancing business competitiveness. A key technology enabling precise task execution is monocular camera-based ranging, which estimates distances to relevant targets. However, existing methods often evaluate targets in isolation, lacking spatial and temporal consistency across frames. To address this, we propose the Consistent and Accurate Ranging Estimation (CARE) framework, which delivers reliable measurements by capturing consistent spatial relationships between targets. CARE integrates two key components: 2D-sense features and a hierarchical Consistent Ranging Network (CRN). The 2D-sense features fuse object detection with inferred depth to provide rich semantic features that enhance input quality, even for tiny objects. The CRN further ensures accurate and consistent ranging through customized network architecture and loss function design. Real-world experiments using data on tomato and rose flowers to simulate tiny-flower pollination demonstrate the effectiveness of our proposed method. The CARE framework outperforms existing approaches in both accuracy and consistency. The CARE model and the test dataset are available at: https://github.com/NCKU-CIoTLab/CARE-for-MUV/.

This study presents CoD-CAST+, a distributed clustering algorithm tailored for large-scale and high-dimensional data analysis. Built upon the CoD-CAST-GPU architecture, CoD-CAST+ preserves the core idea of on-demand affinity computation while introducing a hierarchical acceleration strategy that integrates multi-threading, GPU parallelism, and inter-node coordination. A visualization module is embedded to enhance interpretability, and the system design is guided by a systematic review of clustering algorithms developed over the past decade. Experimental results on a dataset containing 2,048,000 gene records demonstrate that CoD-CAST-GPU achieves a 12.44× speedup over the original CAST algorithm. Building on this, CoD-CAST+(Local) and the fully distributed CoD-CAST+ further extend the performance, with the final system achieving a 110.33× total speedup. Additionally, CoD-CAST+ is the first in the CAST family to support result visualization, making it an innovative and scalable solution for clustering analysis in ultra-large-scale data environments.

Single-image haze removal algorithms restore clear images from hazy inputs but often struggle with high computational complexity and halos or artifacts from patch-based priors. We propose a novel non-local patch-based haze removal algorithm that estimates atmospheric light and scene transmission using the minimum image channel. This approach reduces reliance on patch-based priors, ensuring efficient and high-quality dehazing. An adaptive gain, based on the ratio of Imin and Imax, adjusts the minimum channel to refine atmospheric light, scene depth, and transmission estimates, addressing inaccuracies in traditional methods. Histogram equalization and image multiplication enhance scene radiance, improving contrast, color accuracy, and naturalness under challenging haze conditions. Subjective evaluations confirm our method achieves the highest MOS, consistently rated most visually appealing. In addition, objective metrics show our method outperforms state-of-the-art techniques, achieving better FADE, NIQE, SSEQ, BRISQUE, and PIQE scores by 2% to 13%, indicating more natural colors, minimal distortion, and superior restoration quality. Additionally, our method demonstrates the lowest computational cost and processes 8K images up to 21 times faster than the second-fastest method, CAP. This breakthrough makes the algorithm ideal for real-time applications in consumer electronics such as smartphones and digital cameras, delivering high-quality, haze-free imagery with minimal computational cost.

Virtual Metrology (VM) technology can convert offline sampling inspection into online and realtime total inspection. As the processes of high-tech industries (semiconductor or TFT-LCD) are getting more sophisticated, higher VM prediction accuracy is demanded. With regard to this requirement, the advanced Convolutional-Neural-Networks (CNN) based VM system (denoted as Advanced AVMCNN) was proposed and verified to significantly enhance the overall prediction accuracy. Nevertheless, two issues need to be addressed to enhance the accuracy of the Advanced AVMCNN System: 1) rare and imbalanced collected metrology values lead to poor prediction accuracy of the extreme values, and 2) the model can only be updated when sufficient metrology values are collected. To tackle these problems, the KeepImportant-Samples (KIS) Scheme for the Advanced AVMCNN System is proposed in this paper with consideration of data balance. The experiments reveal that the proposed KIS Scheme can effectively enhance the prediction performance of the Advanced AVMCNN System on the extreme values.