The proposed embedded system supports real time recognition of alphabetic symbols from images of static hands gesture.
Each image is captured and converted into a binary image, which is divided in grids for calculating the percentage of white pixels. This image feature is input for a Extreme Learning Machine (ELM) neural network that recognizes the symbol.
For improving the recognition process, a checker unit confirms the recognized symbol for a set of images captured.
Through a hardware implementation that explores parallelism of operations and pipelining, it was possible to ensure that the system is able to identify symbols without loss of video frames during recognition.