by Kinect Fitness Trainer

**The following are the slides with the slide numbers as presented in class.
 
1. Kinect: Overview of Project

  • Exercise Tutor: A System for In-home Therapeutic Exercise Guidance
  • Enable gesture recognition and use it as an Exercise Tutor.

2. KINECT GESTURE RECOGNITION

Software

  • Investigate the Kinect SDK 1.0.
  • The release of SDK 1.0
  • Improved quality of code and ease of use.
  • Enables the sensor to see objects as close as 40 centimeters.
  • Investigate OpenNI another helpful tool used for gesture recognition.

3. KINECT GESTURE RECOGNITION

Software

  • Investigate the Kinect SDK 1.0.
  • The release of SDK 1.0
  • Improved quality of code and ease of use.
  • Enables the sensor to see objects as close as 40 centimeters.
  • Investigate OpenNI another helpful tool used for gesture recognition.

4. Problems Encountered:

  • Non supportive new release of Kinect SDK 1.0
  • Old examples - time consuming process                                                                                                                   -considering that many functions were removed.
  • OpenNI and Microsoft Kinect SDK are non-supportive of each other.
  • Stick to Microsoft Kinect SDK Beta but future trouble with migration of code to 1.0.
  • OpenNI and Microsoft Kinect SDK are non-supportive of each other.
  • Help in the form of Microsoft.Kinect.Migration.dll.
  • dll will highlight and explain what needs to be changed and why.

5. So far achieved gestures:

                 Exercises
  • Upper right punch
  • Upper left punch
  • Left kick
  • Right kick
  • Left arm curl
  • Right arm curl
  • Squats                                                                                                                                         .   

          Other achievements
  • Voice Recognition
  • Ease of Access to the program
  • As simple as- Start && Stop

6. Limitations

  • Two known issues:
1. Blocked by an object
2. Overlapping joints.

  • Complex for Kinect to detect exercises from proposal.

  • Joints not recognized at some point of exercise. E.g. Crossed arms...

  • Detection of  the motion in real time.

7.             Exercise                                  Multiplanar joints assessed

  • Hamstring stretch-             Spine (cerical-thoracic; thoraco-lumbosacral); hip
  • Can turn-                          Shoulder, elbow, forearm
  • Sit to stand-                      Ankle, knee, hip, pelvis, spine, head
  • Lateral weight shift-           Ankle, knee, hip, pelvis, spine,   head
  • Video game bowling-        Trunk, shoulder, elbow, forearm
  • Hip abduction-                  Trunk, pelvis, hip, knee

                ----------Joints assessed in validation study


8.  Health Science requirements For Exercises

  • Critical parameters for exercise:
1. Velocity of movement, faster OR slower
2. Acceleration of the movement
3. Excursion of motion/segment
4. Alignment of segments
    (a) relative to other segments
    (b) in relation to gravity
    (c) in rotation – a critical component

9. Fix: Use angles/detect motion
  • Calculate the angles between joints we should be able to start recognizing different exercises from proposal.
  • Use X,Y,Z values to figure out angles.
  • Be able to record and train exercises.

19. Fuzzy Logic

  • Binary calculations lead to interesting errors
  • Basic movements cause unforeseen errors
  • Defines the gray area humans live by
  • Tall man example


20.  Fuzzy Logic

Would be applied to skeleton data before rendering.
Enable the Kinect to make assumptions
Squatting


22.
Demonstration!!!!


23.
End of Slides