Explicit Intention Recognition by Using Wrist-type Wearable Device

The project aims to develop a wrist-type wearable sensor system for IoT interaction.

ETC

Intelligence

v
September 22, 2019

Wrist Type Wearable Sensor System

  • Development of Wrist Type Wearable Sensor Prototype
    • Explicit Intention Recognition and Performance Evaluation Available
    • Composed of 1 Sensor FPCB and 2 Board
      • Main Board
        • Composed of MCU, Kinematic Sensor Module, Bluetooth Communication Module, and so on.
      • EMG Sensor Board
        • Composed of pre-AMP, ADC Chip to Acquire the EMG Signal
      • Sensor FPCB
        • Composed of 20 Pair of Electrodes
          • 2 Pair for Ground
          • 16 Pair for EMG Aquisition
    • Can Acquire 16 Channel EMG and 6 Axis Kinematic Data Simultaneously
      • Sampling Rate: 1024Hz
    • Acquired Data: Sent to Computer via Bluetooth Communication
      • Explicit Intention Recognition and Performance Evaluation Available

     

  • Maximum Likelihood Estimation based Explicit Intention Recognition System
    • Process

  • Experiment Set Up
    • Experiment Environment
      • 6 Kind of Explicit Intention(Finger or Wrist Motion) Defined
        • Focused on Availability and Necessity
      • 30 Repetition for Each Explicit Intention
        • Maintained for About 0.2 Second

  • Performance Evaluation
    • Recognition Method
      • Training Set: Randomly Selected 15 Trials for Each Intention
      • Test Set: 15 Trials Which is Not Selected as Training Set
      • Cross-validation is Used
      • Gaussian Model based Maximum Likelihood Estimation is Used
    • Result
      • Shows 96.67% of Average Recognition Accuracy

  • Motion Duration Recogition Based Extrinsic Intention Recognition System
    • Process

  • Experiment Set Up
    • Experiment Environment
      • Motion Start: Holding Grab Motion
      • Motion End: Changing Grab to Extension
      • Recognize Time between Motion Start Point and Motion End Point
      • 2 Different Motion is Used
        • Short Motion: Duration Under 0.5 Second
        • Long Motion: Duration Over 1 Second

  • Performance Evaluation
    • Do Not Need Training
    • Selected Answer Section by using Accelerometer Data from Finger
    • Identified the Time Error by Comparing Answer Section and Recognized Section (△e_1& △e_2 in the figure below)
    • Motion Duration Time Recognition Method
    • Result
      • Short Motion: Occurs Average 0.08 Second of Duration Time Error
      • Long Motion: Occurs Average 0.14 Second of Duration Time Error

 

IoT Interaction

  • IoT Device Control System
    • Transmit the Acquired Data from Wearable Device to Computer
    • User Intention Recognition on Computer
    • Control PC Software
      • Google Map
      • Media Player
    • Control IoT Device
      • Smart Bulb
      • Smart Door Lock
      • Smart Switch
      • Smart Temperature Control System
    • Smart Device Control System based on Wrist Type Wearable Device

Explicit Intention Recognition by Using Wrist-type Wearable Device


Introduction

  • The project aims to develop a wrist-type wearable sensor system for IoT interaction.
  • To achieve the goal of the project, we develop wrist-type wearable sensor prototype which records multi-channel EMG and 3 axis accelerometer and gyroscope.
    • Data from the wrist-type wearable device is transmitted to computer by bluetooth communication module.
  • The various signal processing and data analysis including explicit intention recognition is progressed on computer.
    • Control command correspond to the recognized explicit intention is sent to various smart devices.
  • IoT interaction by using the control command is the final goal of the project.