Data Set Development for Gesture Recognition and Implementation using MATLAB and Xilinx Vivado

Gesture recognition system is an algorithmic based system which can interpret different hand gestures, which are useful in human machine interface. In this paper, two datasets are used for training and testing. The first data set developed by the authors, has a total of 250 images, across 10 classes. Images of ten different hand gestures of one of the authors were captured and segregated into ten corresponding classes. For each hand gesture, 25 different images are captured in varying positions. To increase number of images in the data set, a second data set which consists of 1000 images, across 10 classes, is also considered from an open source. In the first data set, 250 images are considered for training and the resolution of each image is 432 x 578. In the second data set, 1000 images are used for training. The trained images are converted into gray scale and stored in the binary format. A gesture recognition algorithm is designed and 90 images are used for testing in both datasets. The gesture recognition algorithm is modeled in MATLAB for accuracy of 99.2% for the first data set and 98.6% for the second data set. The next stage of the project is conversion of designed MATLAB code into verilog code using HDL coder. The verilog code is functionally simulated, synthesized, and implemented in Xilinx vivado.