全國中小學科展

2025年

Developing a Heart-Rate Monitoring App to Help Families in Identifying Mood Changes for Children Diagnosed With Neurodivergence

This study dives into the benefits of a heart rate (HR) monitoring application to aid families in identifying mood changes in children diagnosed with neurodivergence. Children with neurodivergence often struggle with communicating their emotions, which often results in tantrums or emotional outbursts, and this study plans to address this by creating an app that detects users’ HR to calculate heart rate variability (HRV) and detect when the user’s heart rate variability (HRV) levels become higher than usual. Heart-rate variability is defined as a small variation of the interval between every heartbeat, it’s calculated with the formula of . By looking at the developments of these small variations, it will be 60/𝐵𝑃𝑀 × 1000 easier to notice then the interval for heartbeats are shorter, meaning the body is in need of more blood pumped quickly for support. The app is connected to a heart rate sensor that is worn by the user. The heart-rate sensor frequently uploads data to the app which the app processes and carefully observes while looking for any sudden, dramatic change. The sensor and app was tested and proved to meet the expected requirements of functionality. Four participants with neurodivergence were asked to equip the heart-rate sensor and results showed that different developments of heart-rate variability were able to be detected by the app, these participants varied in their type of neurodivergence as well as their age. As an example, the third participant showed the purpose of the app most visibly, having a resting heart-rate of 86 BPM (697.67 ms) turning into a high 107 BPM (561.68ms) after changing activities. When the user’s sensor detects a sudden spike in heart-rate variability, the app notified the parent account about this change in emotion. This study has supported the relevance of using heart-rate variability to observe changes in mood.

基於高效可更新神經網絡的西洋棋人工智慧應用於嵌入式對弈棋盤

本研究以西洋棋為切入點,採用磁簧開關陣列來偵測棋子位置,並在設計中加入二極體以防止Ghosting效應,進而開發出一款以Arduino Uno開發板為基礎的智慧對弈棋盤。棋盤底部配備RGB LED燈,以便為使用者提供落子提示,並根據不同的落子類型呈現不同的燈光效果。 我們成功地透過簡潔的設計與高效的運算性能,實現了一個能夠識別棋手落子的智慧對弈棋盤,並能根據國際西洋棋規則提供正確的移動提示,讓完全沒有基礎的初學者也能在遊戲中學習並掌握西洋棋的所有規則。此外,我們還引入了基於Minimax演算法的輕量化AI和基於高效可更新神經網絡(NNUE)AI,並探討兩者之間的性能差異,從而使該智慧棋盤在節省運算資源的同時,可以在不連接電腦的前提下,具備一定的棋力,以支持棋手的技能提升與訓練。

以底棲魚生物放大效應探討邊緣海區域性汞汙染Marginal Sea Regional Mercury Pollution Revealed by Biomagnification Effects in Demersal Fish

海洋中有許多重金屬汙染,其中汞元素因為濃度低很容易被忽略,但卻容易經由飲食進入人體,造成嚴重傷害。也因其濃度低不易被測量,現今也少有海洋中汞汙染的完整資料。然而,在生物體中汞濃度會因生物累積及放大作用而較海水的濃度來得高,故本研究利用魚體汞累積速率(MAR)當作追蹤海洋污染之生物指摽,此方法將魚體總汞濃度除以年齡得到的汞累積速率,以去除生長時間的影響因素。 本研究利用習性不常移動之底棲魚種之MAR,分析與生物放大作用相關的掠食階級(Trophic Levels)之相關趨勢,經過篩選,研究分析了31篇過去於大西洋、太平洋及地中海採樣的文獻,將其中資料整理成趨勢圖,比較各區域汙染程度,發現各區的汙染程度呈現差異,同時提供觀測區域海汞汙染的新方法。

影像辨識 在 智慧運輸系統 的應用 自動跟隨

本研究旨在開發一款能自動跟隨人的AI智慧運輸車,應用在 醫院、商場等環境,能減輕工作人員的負擔。本系統融合YOLOv8影像辨識技術和物聯網,使運輸車具備跟隨人體的能力。該系統通過物聯網,將車上攝影機接收的影像傳送至電腦,進行即時運算,不必在車上配備高階的微型電腦或GPU以降低成本。使用YOLOv8模型辨識人體與手勢,以arduino ESP32開發板作為主控單元控制減速馬達,使運輸車能自動跟隨人員。本研究不像傳統自動導引車(AGV),因為依賴固定路徑,而無法用在多變的場景,且不適合與人協同工作。將本系統應用於醫療院所或零售業商場等 ,經常出現 變障礙物的場所時,能與人員配合執行任務,例如當作病人的點滴架、輔助護理師工作的醫務車,以及賣場中協助工作人員上貨的籠車,能有效降低勞力負擔。

開發影像辨識系統應用於離岸流偵測預警

離岸流是海灘安全的重大隱患,台灣有30處海岸經衛星拍攝到離岸流。其發生時間和地點不定,且會迅速將人帶離岸邊,最遠約100公尺。民眾常因掙扎而體力不支,導致溺斃。為減少此類事故,本研究開發了一套離岸流偵測預警系統,構建出準確度超過95%的辨識模型。 因台灣無離岸流圖庫,本研究除國外圖庫,也加入台灣GoOcean等平台的即時影像,以貼合台灣實際狀況。經影像前處理後,共約四萬張圖片和五十部影片用於訓練YOLO v8模型。當偵測到離岸流,即時透過Discord發送警報至救難中心。我們也加入人形辨識功能,若系統同時偵測到人與離岸流,則加強警戒等級,從而提升救援效率。 完成離岸流辨識系統後,我們著手開發預警功能。將辨識模型應用於各地監視器,蒐集更多離岸流資料,再先後使用YOLO v8和Transformer提取生成特徵,預測其發生,以達成預警的功能。

降脂轉肌–將脂肪轉變成肌肉的可能性探討

先前研究發現一種多元不飽和脂肪酸 15-keto-PGE2 能減少肌肉慢性發炎以及增加肌肉生長因子胰島素的敏感性,具有增加肌肉的潛力。我們探討此脂肪酸在小鼠中將脂肪組織轉成肌肉組織的可能性。 我們先以 15-keto-PGE2 治療肥胖型肌少症小鼠後,再以全基因組mRNA定序,發現在小鼠皮下脂肪中,肌肉特異基因表現量高度上升。透過基因表現路徑分析軟體 GSEA證實此脂肪酸會高度引發皮下脂肪中,與肌肉分化、肌肉收縮與肌肉結構相關的生物路徑。另外, 15-keto-PGE2 也可以誘導脂肪前驅細胞3T3-L1中與肌肉生成相關的基因表現。 最後,我們發現以 15-keto-PGE2 治療的老年肥胖型肌少症小鼠,體重、血糖、脂肪比例下降,肌肉質量及力量上升,證實了它在體內同時減少脂肪並增加肌肉量的效果。

以深度學習進行籃球慣用動作分析

本研究聚焦於籃球員的慣用動作分析,透過深度學習技術開發了一套籃球動作分析系統,旨在準確分析籃球員在籃球運動中的個人動作特徵來進行動作辨識。我們透過自行蒐集籃球動作的影片,並使用MMAction2這個資源庫來進行動作辨識模型的訓練,將訓練好的動作辨識模型用開發慣用動作分析系統。系統流程首先使用滑動視窗(Sliding Window)的機制將即時拍攝的影像變成有序列的連續影像片段,再即時傳送至進攻動作辨識的深度學習模型中,來辨識出連續影像片段中的動作序列屬於何種特定動作,藉此將多個連續影像片段中的動作序列各自轉換為單一動作單元並依次輸出。最終,系統基於前述單一動作資料進行綜合分析,以統計使用者的籃球慣用動作。此分析系統能為籃球愛好者提供清晰的動作偏好資料,具有提升訓練成效的潛力,同時為籃球技術分析與訓練提供了一個精確的數據分析工具。

方格裡的秘密—隨機分布的機率探討

本文研究了一個信息完全公開的組合遊戲,探討當一群人被完全隨機的分配到模型裡時,其初始位置與特定位置所形成的包圍關係,並探討最佳的人力分配。本研究通過座標解析與不等關係的代數運算等方法,成功找出獲勝條件對於遊戲雙方的限制,並進一步解決問題。在研究的過程中,也將結論擴展到不同模型,探討不同模型對於遊戲造成的影響,並比較其結論有何區別。

Eradicating Cystic Fibrosis Biofilms by a Novel Non-Toxic, Multi-Pathway Salicylate Therapy

1.1. Cystic Fibrosis Biofilms Biofilms are bacterial aggregates in a matrix of polysaccharides, proteins and nucleic acids (Donlan, 2002). They account for 80% of all chronic infections and cause over 500,000 deaths annually. Cystic fibrosis (CF) is a genetic disorder characterized by mucus accumulation in the respiratory tracts (Morrison et al., 2020). This impairs mucociliary clearance, allowing chronic colonization by bacterial biofilms, leading to fatal respiratory failure, lung scarring, and necrosis of pulmonary epithelial tissues (Martin et al., 2021). 1.2. Obstacles in Current Treatments Three major therapies are used against CF biofilms: (1) aminoglycoside antibiotics like tobramycin, (2)non-aminoglycoside antibiotics such as ciprofloxacin and vancomycin, and (3) non-antibiotic therapies including flushing, chlorination, and ultraviolet disinfection. These have two major flaws. First, they are cytotoxic; 30% of patients experience acute kidney injury after three days of intravenous aminoglycoside therapy (Joyce et al., 2017). Furthermore, non-aminoglycoside therapies can cause phospholipid buildup in lysosomes of proximal tubule epithelial cells, accounting for 10-20% of acute renal failure cases. Second, antibiotic resistance due to horizontal gene transfer and mutations has significantly reduced treatment effectiveness. Therefore, cystic fibrosis biofilms remain a critical threat with few effective treatments. 1.3. Salicylate Derivatives This project tackled this issue using an innovative non-antibiotic approach with salicylate derivatives. Salicylates, a class of benzoic acids—benzene-based carboxylic acids (Figure 1)—used in painkillers and blood thinners, were investigated for their antibiofilm potential through a 3-step process: 1. Literature review: Identified three key biofilm therapeutic targets: quorum sensing, bacterial adhesion, and cell motility. Disrupting these pathways would result in biofilm eradication. 2. Molecule Identification: Recognized key molecules in each pathway: LasR, adhesins, and flagellin. Inhibiting these molecules would disrupt the pathways. 3. Screening: Found that salicylates could inhibit the identified molecules, though they had never been tested against cystic fibrosis biofilms.

正本「塑」源-探討河川塑膠微粒

基於對鄉土的關懷,我們選定高屏溪流域,來研究河川中塑膠微粒的尺寸大小和含量,透過光譜來了解塑膠微粒的材質。高屏溪流域是跨高雄及屏東地區最大的河流,我們共分為枯水期及豐水期來採水,比較不同季節之塑膠微粒類別,若以形狀區分:枯水期塑膠纖維最多,豐水期塑膠薄膜最多。若以尺寸區分:枯水期100~500μm的佔最多,豐水期25~100μm的佔最多。若以材質區分,枯水期含量最多分別是PE及PP,豐水期含量最多分別是PE及PET。若從顏色區分,大多是白色及透明為主要顏色。以上實驗結果可供主管機關列為重要監測參考,特別是國內少有研究單位或機構,在進行高屏溪流域塑膠微粒類別、含量之調查與研究,因此本研究重大發現為國內創舉深具意義,可供相關單位後續深究參考之價值性。