high-hand

Hand tracker and gesture learner based on OpenCV.

Rapid coding storm

Requirement

• A cpp/python demo of hand tracker.

Step.0. Learn OpenCV. (FINISHED)
Step.1. Read image from video device. (FINISHED)
Step.2. Record the background and set stdDEV threshold. (OPTIONAL)
Step.3. Calculate the diff between current image and the recorded background. (OPTIONAL)
Step.4.0 Extract the skin using YUV ellispe-skin-model and HSV model. (FINISHED)
Step.4.1 Mask erosion and dilation to remove the noise. (FINISHED)
Step.5. Draw the hand's contour and display it on the screen (FINISHED)
Step.6. Track the center of hand's contour. (FINISHED)

• Experiment of gesture analyse and recognition based on Deep Learning (OPTIONAL)

0. Study the field of picture recogniton, OCR, SVM, etc.
1. Features extraction
2. Training and testing samples generation
3. Machine learning algorithm experiment
4. Result analyse

• Trace Recoginition

0. Eight-direction Trace State Machine
1. Hidden Markov Model (HMM)
2. Use Edit Distance to fit the trace. (FINISHED)

Issue

• The camera doesn't perform well. So it needs some Filtering (corrosion and expansion) or Noise Reduction algorithms.
• The pace is spike, so it needs to be smooth.

Demo

• The convex hull of hand

• Gesture recognition

Reference

OpenCV API Reference