cw471802 / hand-gesture-recognition

Detect number of fingers from image or video using OpenCV

Hand Gesture Recognition

• Extract and segment hand region from the video sequence.

• Recognize the number of fingers from the segmented hand region by using Convex Hull.

Getting Started

How to use

git clone https://github.com/aakashjhawar/Hand-Gesture-Recognition.git
cd Hand-Gesture-Recognition

Run the Finger_count.ipynb Jupyter Notebook

Prerequisites

• Python 3.5
• OpenCV

sudo apt-get install python-opencv

Procedure

• Strategy for counting fingers
  • Garb an ROI (Region of interest)
    • Calculate a running average background value for 60 frames of video
    • Once average value is found, then the hand can enter the ROI
• Set a ROI and calculate the average running value for some amount of frames
• Then once a hand enters, we can detect change and apply thresholding
• Strategy for counting fingers * Once the hand enters the ROI, we will use a Convel Hull to draw a polygon around the hand * Using some maths, we'll calculate the center of the hand against the angle of outer points to infer finger count
• The next step is to use thresholding to grab the hand segment from the ROI
• Now that we have the hand segment, the next step is to actually count the fingers behind held up
• We can do this by utilizing a Convex Hull
• A convex hull draws a polygon by connecting points around the most external points in a frame
• In our case, this set of points is actually just our threshold image of a hand.
• We can expect a general shape of our polygon to be something like
• Then using a ratio of that distance we create a circle
• Any points outside of this circle far away enough from the bottom, should be extended fingers

Working

Image-

Result

One

Two

Three

Four

Five

Segmentation of some image are improper as the lighting in the room was uneven.