Gesture Recognition Arduino nano 33 ble sense (implementation notes)

Board info: https://store.arduino.cc/usa/nano-33-ble-sense
Arduino On-chip AI: https://www.arduino.cc/en/AI/HomePage

Getting started: installation

• On Arduino app:
  • 下載程式褲：Arduino ➡️ 工具 ➡️ 管理程式庫 ➡️ 搜尋Arduino_TensorFlowLite & Arduino_LSM9DS1 & ArduinoBLE,etc ➡️ Install
  • 下載開發版：Arduino ➡️ 開發版 ➡️ 開發版管理員 ➡️ 搜尋Arduino nRF528x Boards (Mbed OS)或Arduino Nano 33 bl > install
  • 燒錄器(programmer)設定為AVRISP mkll
• On Arduino web editor:
  • Sign in: https://create.arduino.cc
  • Install Arduino Create Agent: https://create.arduino.cc/getting-started/plugin/install (So that web editor could detect your board which is connected to PC via USB port!)
  • Search for libraries you need!

TensorFlow Lite for Microcontrollers examples

Imfortant Ref: https://blog.arduino.cc/2019/10/15/get-started-with-machine-learning-on-arduino/

• Examples in Arduino_TensorFlowLite :micro_speech, person_detection, etc.
• Examples in Arduino_LSM9DS1 helps to read accelerometer and gyroscope(陀螺儀) values
• Examples in ArduinoBLE deal with BLE connection tasks
• 視所需下載程式庫，尚有溫濕度、壓力感測等等範例可使用...

Gesture Recognition Tutorial

• 在Arduino執行IMU_Capture.ino來搜集姿勢資料(手握Arduino進行flex與punch兩種動作，或其他欲進行辨識的動作) ➡️ 複製序列埠data ➡️ 將一種姿勢資料做成.csv檔案。記得將空白列刪除(可透過excel)，否則訓練時會出現nan錯誤。
• 開啟arduino_gesture_recog.ipynb ➡️ 匯入flex.csv與punch.csv檔案 ➡️ 進行訓練 ➡️ 匯出model.h
• 回到Arduino開啟IMU_Classifier.ino
• 在Arduino新增Tab ➡️ 命名為model.h ➡️ 將由colab(ipynb file)匯出之model.h內容複製貼上。(或是直接將model.h放置於IMU_Classifier資料夾內)
• 最後編譯上傳IMU_Classifier.ino至開發版，開啟序列埠
  • 執行flex動作時，序列埠呈現：
    
  • 執行punch動作時，序列埠呈現：
    
• Future Attempt: 利用Emoji_Button.ino來輸出Gesture圖案
• Other pojects: https://create.arduino.cc/projecthub/dgiancono/nano33blesensor-getting-started-with-the-nano-33-ble-sense-8a7eba