Page 24 - sprouting/2025autumn

今期專題 FEATURE

研究種子栽成應用成果
FROM RESEARCH IDEAS TO PRACTICAL IMPACT

五位學生奔赴葡萄牙學習前，已於 Before travelling to Portugal, all five scholars had already begun
澳理大順利開展各自的博士研究項 their doctoral research at MPU, projects carefully designed
目。他們的研究課題著重以個人學 to reflect personal interests while addressing critical gaps in
術興趣為出發點，緊密對接領域內 their fields. Each study embraced interdisciplinary thinking and
的研究缺口，積極融合跨學科視野 focused on practical impact, aiming to turn innovative ideas into
與方法，並高度注重成果的實際應 solutions for real-world challenges.
用價值，致力針對當前的關鍵難題，
Gao Hao explored the intersection of neuroscience and
提出具備操作性的創新解決方案。
software engineering, using eye-tracking and data analytics to
高皜聚焦於神經科學與軟件工程的 study programmers’ cognitive load. “We found that traditional
交叉應用，透過眼動追蹤及數據分 code complexity metrics aren’t absolute; eye revisits often
析，研究程式設計師的認知負荷： reveal cognitive strain more accurately,” he explains. His work
“在研究過程中發現，傳統的程式 produced a new complexity model that reflects real cognitive
碼複雜度指標並非絕對，眼睛回看 effort, paving the way for more human-centred software
程式碼的次數，更能真實反映大腦 design.
的認知負擔。”他開發出反映程式
Chong Chak Fong focused on medical imaging, tackling the
員實際認知負荷的代碼複雜度模型，
scarcity of labelled data in deep multilabel image recognition.
推動以人為本的軟體設計方法。
He developed a suite of post-training algorithms for deep
鍾澤豐專注於醫學影像領域，在“深 learning, significantly improving chest X-ray disease diagnosis
度多標籤圖像識別”中標註數據稀 accuracy. His work provides more precise AI-assisted tools
缺的難題，開發出一系列具普適性 for medical diagnostics, advancing the development of smart
的深度學習後訓練算法，顯著提升 healthcare.
胸片疾病診斷的準確率，為醫療診
Wu Zewei specialised in visual multi-object tracking, proposing
斷提供更精準的 AI 輔助工具，推動
an innovative approach that treats detection algorithms as
智慧醫療發展。
generalised “online sensors”. By analysing dynamic patterns
吳澤偉的研究專注於視覺多目標跟 through a time-series perspective, he designed an adaptive
蹤。他提出將目標檢測算法視作一 short-trajectory tracker capable of maintaining stability even in
種廣義的“在線傳感器”，以時間 low-confidence detection scenarios.
序列的觀點分析目標分佈的動態特
Zhu Han turned his attention to communication security in
性，設計出一種自適應短軌跡跟蹤
the IoT era, particularly in ambient backscatter systems. He
器。這項創新使系統即使在低置信
introduced a novel system architecture that enhances both
度的檢測場景中，也能穩定運作。
stability and confidentiality in data transmission, marking a
朱含守護著物聯網時代的通信安全， critical step toward secure and reliable IoT communications.
特別關注環境反向散射通信中的安
Shen Lu addressed a longstanding challenge in Chinese
全性與可靠性提升。他提出一種新
information processing: handwritten text recognition and
穎的系統架構，同時加強資訊傳輸
glyph restoration in noisy environments. Her model effectively
的穩定與保密性，成為未來物聯網
reconstructs and identifies distorted characters in Chinese
通信的關鍵一步。
contexts, laying the foundation for applications such as
沈璐將研究視角轉向中文資訊處理 digitising historical documents.
中的難題——噪聲環境中的手寫文
字辨識與字形修復。她開發的模型，
能夠針對包含干擾的手寫字進行有
效識別與重建，尤其在中文場景中
取得不錯效果，為未來歷史文獻數
位化等應用奠定基礎。

22