技術簡介
本技術針對 ADHD 孩童在特定情境下的專注力辨識,透過穿戴式裝置(如 Apple Watch 或其他智慧型手環)內建的 加速度計與陀螺儀,即時記錄孩童的動作數據,並利用 AI 模型進行動作辨識。
1. 技術突破與特殊功能
(1) 即時動作辨識
(2) 連續時間分析
(3) 適應不同學習情境
(4) 即時回饋與提醒
2. 技術特點
(1) AI 動作辨識:每秒即時判讀孩童行為,如包含 操作搖桿、寫生字、擦橡皮擦、操作腕錶、其他動作。
(2) 長時間分心分析:計算分心占比,提供專心度報告。
(3) 適應多種學習場景:寫作業、閱讀、考試、上課等皆適用。
(4) 即時提醒機制:透過震動或螢幕通知,提醒孩童恢復專注狀態。
這項技術可應用於 ADHD 兒童的學習輔助,幫助家長與教師更了解孩童的專注模式,並透過即時回饋機制提升孩童的學習專注力。
Abstract
For ADHD children’s attention recognition in specific contexts, we use the built-in accelerometer and gyroscope of wearable devices (e.g., Apple Watch or other smart bracelets) to record children’s movement data in real time, and then use the AI model to recognize their movements.
1. Technological breakthroughs and special features
(1) Real-time motion recognition: Using the Accelerometer and Gyroscope to capture hand movement characteristics, the AI model can recognize the child’s current actions at the per second level, including operating the joystick, writing vocabulary, erasing erasers, operating the watch, and other motor behaviors.
(2) Continuous time analysis: AI algorithms were used to calculate the percentage of distraction to further estimate the child’s Focus Score under long-term monitoring. If prolonged non-learning actions are detected, such as operating a watch or other actions, it may indicate that the child has entered a distracted state.
(3) Adapt to different learning situations: Support different learning situations such as homework, lesson plan training, essay writing, etc. Analyze the focus score according to the changes in the scene, and provide teachers and parents with more accurate reports on concentration.
(4) Instant feedback and reminder: The system can remind children to return to the state of concentration through Haptic Feedback or on-screen prompts according to the trend of children’s concentration decreasing, forming a real-time concentration training mechanism.
2. Technical features
(1) AI Action Recognition: Read the child’s behavior in real time every second, such as operating joystick, writing, erasing eraser, operating wristwatch, and other actions.
(2) Long time distraction analysis: Calculate the percentage of distraction and provide concentration report.
(3) Adapt to a variety of learning scenarios: homework, reading, exams, classes, and so on.
(4) Instant reminder mechanism: Vibration or on-screen notification to remind children to return to the state of concentration.
This technology can be applied to ADHD children’s learning assistance, helping parents and teachers to better understand the child’s concentration pattern and improve the child’s learning concentration through real-time feedback mechanism.
技術規格
1. 硬體規格
感測器
• 加速度計 (Accelerometer):三軸 (X, Y, Z) 資料輸出。
• 陀螺儀 (Gyroscope):三軸 (X, Y, Z) 角速度測量,精確監測手部旋轉與晃動。
裝置需求
• 支援裝置:Apple Watch / Garmin / Fitbit / 其他智慧手環
• 最低運行規格:
o 處理器 (CPU):≥ Dual-Core 1.2GHz
o 記憶體 (RAM):≥ 512MB
o 儲存空間:≥ 50MB
o 無線連接:BLE 4.0 or Wi-Fi
2. 軟體架構
資料收集
AI 動作辨識模型
3. 即時回饋機制
警示與通知
• 震動提醒 (Haptic Feedback):孩童長時間未回到學習狀態時,裝置發送震動提醒。
• 螢幕通知 (On-Screen Alert):如連續 3 分鐘分心,顯示提示訊息提醒孩童回到學習狀態。
Technical Specification
1. Hardware Specification
Sensors
- Accelerometer: 3-axis (X, Y, Z) data output.
- Gyroscope: 3-axis (X, Y, Z) angular velocity measurement to accurately monitor hand rotation and wobble.
Device Requirements
- Supported Devices: Apple Watch / Garmin / Fitbit / other smart bracelets
- Minimum Running Specifications:
˙ Processor (CPU): ≥ Dual-Core 1.2GHzo
˙ Memory (RAM): ≥ 512MBo
˙ Storage Space: ≥ 50MBo
˙ Wi-Fi Connection: BLE 4.0 or Wi-Fi2
2. Software Architecture
Data Collection
- Sampling Rate: 50Hz or above
- Data Format: CSV / JSON
- Data content:
˙ Acceleration(XYZ)
˙ Gravity(XYZ)
˙ Rotation Rate(XYZ)
˙ Attitude(raw, yaw, pitch) o
˙ Time Stamp
AI Motion Recognition Model
- Recognizable Behavior:
o Scenario Analysis: Writing Assignment, Lesson Plan Training, Composition Writing
o Motion Recognition: Operation of joysticks, writing vocabulary, writing essays
, and other activities. o Motion Recognition: operate joystick, write vocabulary, erase eraser, operate wristwatch, other actions
- Classification Accuracy: ≥ 90%
- Response Time: ≤ 1 second (Instant Computing)
3.Haptic Feedback:
Haptic Feedback is sent to the device to remind the child when he/she has not returned to learning for a long period of time.
Haptic Feedback: If the child does not return to learning for a long time, the device sends a vibration alert.
On-Screen Alert: If the child is distracted for 3 consecutive minutes, the device will display a message to remind the child to return to learning.
技術特色
1. 即時動作辨識與專注力分析
透過 AI 模型每秒辨識孩童當前行為。
可辨識 學習情境 (寫作業、教案訓練) 與 細節動作 (操作搖桿、寫生字、擦橡皮擦、操作腕錶、其他動作等)。
準確率達 90% 以上,可即時提供分心提醒。
2. 長時間學習狀態分析
AI 演算法計算 分心占比,並推算 專心度 (Focus Score),提供數據化指標。
可生成 學習時間統計報告,記錄孩童每日 / 每週專注力變化趨勢。
3. 即時回饋與智能提醒
震動提醒 (Haptic Feedback):當孩童連續分心過久,穿戴裝置會震動提醒。
螢幕提示 (On-Screen Alert):可顯示「請專心學習」等訊息,幫助孩童恢復注意力。
應用範圍
本技術具備即時專注力辨識、長時間行為監測與智能回饋機制,可廣泛應用於 教育科技 (EdTech)、心理健康 (Mental Health)、數位健康照護 (Digital Health)、智慧穿戴設備 (Wearable AI) 及 人因工程 (Human Factors Engineering) 等領域。在 學校與家庭學習監測 方面,本技術可幫助家長與教師了解 ADHD 孩童的學習專注度變化,透過震動與螢幕提醒引導孩童回到學習狀態,並提供長期數據報告,協助制定個人化學習策略。在 心理與醫療輔助應用 上,該技術可整合至 ADHD 行為治療計畫,作為專注力訓練工具,與心理輔導中心、兒童發展診所合作,提供科學化的行為監測與干預參考。此外,本技術亦可運用於 企業員工專注力訓練、考試監測、職業安全監控 等場景,透過分析長時間工作專注度、分心行為,優化生產力與專業培訓成效。在 智慧穿戴與未來應用擴展 方面,技術可整合至 Apple Watch、Fitbit、Garmin 等穿戴裝置,未來更可與 腦波感測 (EEG) 與心率變異性 (HRV) 指標 結合,發展更精準的專注力監測系統,應用於運動訓練、軍事戰術決策與高風險作業監控。本技術不僅適用於 ADHD 兒童,也可進一步推廣至一般族群,作為提升專注力與行為監測的智慧解決方案。
接受技術者具備基礎建議(設備)
■ 穿戴式裝置 :
內建加速度計或陀螺儀,至少支援50HZ取樣頻率
內建低功號藍芽或WIFI
接受技術者具備基礎建議(專業)
■ 使用者 : 會操作智慧穿戴式裝置的任何人
■ 系統維護者
˙ 1. 人工智慧與機器學習
˙ 專業背景:AI、機器學習 (ML)、深度學習 (DL)
˙ 主要職責:
˙ 設計並訓練 動作辨識 AI 模型
˙ 進行 數據預處理與特徵提取(Gravity、Acceleration、Rotation 參數)
˙ 優化 模型效能 以確保低延遲高準確度的即時分析
˙ 整合 AI 模型至行動應用
˙ 2. 穿戴式裝置開發
˙ 專業背景:嵌入式系統、IoT、感測器數據處理
˙ 主要職責:開發腕錶上的感測數據收集應
聯絡資訊
聯絡人:黎和欣 智慧醫療與照護服務組
電話:+886-3-5913411 或 Email:kernoli@itri.org.tw
客服專線:+886-800-45-8899
傳真:+886-3-5837446