Augmented reality deep gesture network

A computer implemented method for recognizing a hand gesture using a random forest model, comprising:

• performing a training on a random forest model that comprises a plurality of classification decision trees;
• obtaining image data;
• clustering a plurality of pixels from the image data into a plurality of clusters;
• analyzing the plurality of clusters using a rejection cascade with multiple criteria pertaining to depth and size information to filter the plurality of clusters into a plurality of selected candidates;
• analyzing the plurality of selected candidates using the plurality of classification decision trees in the random forest model that has been trained to recognize a hand in a selected candidate among the plurality of selected candidates;
• representing the hand in the selected candidate as a skeletonized hand model comprising a one-dimensional line segment and multiple branches at least by applying a distance or medial axis transform to the selected candidate based at least in part upon a contour of the hand, wherein applying the distance or medial axis transform comprises:
• assigning a first value for points located on the contour of the hand in the selected candidate;
• and
• assigning a second value to additional points located within the contour of the hand based at least in part upon a distance between a respective additional point of the additional points and a surface of the hand;
• performing a lower-dimensional analysis on the skeletonized hand model using regression decision trees in the random forest model to identify a first location of a center of the hand and a second location of a joint of the hand, wherein the plurality of classification decision trees comprises the regression decision trees;
• and
• recognizing a hand gesture using the plurality of classification decision trees based at least in part on the first location of the center of the hand and the second location of the joint of the hand.