Ant Learning Algorithm for Gesture Recognition

Fra Robin

Gå til: navigasjon, søk

For his Master Thesis, Sichao Song delveloped an ant learning algorithm for gesture recognition. Details of the algorithm applied to accelerometer-based classification can be found in Song's thesis, and in the paper An Ant Learning Algorithm for Gesture Recognition presented at the IEEE Congress on Evolutionary Computation (CEC), in Cancun, Mexico 2013.


Using the Ant Learning Algorithm with the General Purpose Sensor Platform (GPSP)

We use the ant learning algorithm in combination with the GPSP and FSRs embedded in a sole to recognize gaits and other foot movement.

User Manual

  • By default, the GPSP will broadcast its IP on a network with SSID "EPICSdemo".

Manual Setup

  1. Make sure your computer is at the EPiCSdemo network
  2. Use either java GUI_main or the Max patch GPSP-setup.maxpat to specify the data destination IP and port. SSID may also be changed here.
    • wlan rate: 6Mbit/s
  3. Use either java GUI_data the Max patch GPSP_GUI to specify the desired data streams, change filtering and streaming type, and to start/stop streaming.
    • only the three force sensors are used in the current implementation (default: AD0, AD1, AD2)
    • set moving average filter coefficient: 3
    • remove the time stamp
    • remove the digital pins
  4. After making the settings, choose stream continuously and then start to send OSC Packets.

Automatic Setup

  • Make sure your computer is at the EPiCSdemo network
  • Run the Max patch GPSPreceive.maxpat to set up IP adresses and ports for all GPSPs on the network.
Personlige verktøy