I have a school project that requires me to build a sensor module attached to user's ankle which could sense and recognize basic human motion such as walking, running jumping, sitting and some pedometer applications. I already have the basic idea of how to build the system which consist of using a 6DOF Inertial Measurement Unit (IMU) connecting to a Arduino Uno and a bluetooth module. A Kalman filter will be used to fuse the sensor data.
However, I would like to know how to develop the algorithms used to recognize the human motion and pedometer applications. Any inputs and advises are welcomed.
Phew, I thought this was going to be an electronic nose project.
You are not planning on strapping the Arduino to an ankle are you?
There may be algorithms published already for this, you should search for them. Otherwise do it empirically, gather data when walking, jogging, sprinting, climbing stairs etc. look for a pattern in the data that characterises each motion then write software that looks for the pattern.
You are going to get differences in acceleration and frequency of motion for one thing.
radman:
Phew, I thought this was going to be an electronic nose project.
You are not planning on strapping the Arduino to an ankle are you?
There may be algorithms published already for this, you should search for them. Otherwise do it empirically, gather data when walking, jogging, sprinting, climbing stairs etc. look for a pattern in the data that characterises each motion then write software that looks for the pattern.
You are going to get differences in acceleration and frequency of motion for one thing.
Hi,
Thanks a lot for the reply. Yes i plan to do it by strapping the module to my ankle. But by collecting data empirically, i would have to run the experiment on different people with different weights etc so that i can get a correlation of how weights (or maybe legs length ) can vary with the motion pattern?
