Real-time Bengali and Chinese numeral signs recognition using contour matching

This paper presents a real-time Bengali and Chinese numeral signs recognition system using contour matching. The system converts the captured image into gray scale image. After histogram equalizing and smoothing the gray scale image, the system detects edges using Canny edge detection algorithm. Edge linking is done using morphological operation. The system extracts contours and calculates contour length and area from the detected edges. The system filters unwanted contours based on minimum contour length and area. Then the system encodes the remaining contours into Vector Contours (VC). The system resizes the encoded VC into predefined size. The system generates feature vector based on equalized VC, Auto-Correlation Coefficient (ACC), Normalized ACC and ACC descriptors of equalized VC, which will be used for training and/or testing process. The system recognizes the hand signs based on maximum similarity between test contour and predefined training contours of hand signs using Inter-Correlation Function (ICF). The system is trained using 1000 (10×10×10) contour templates separately for both ten (*** to ***) Bengali and ten (0 to 9) Chinese numeral signs from 10 signers. The system is tested using another 1000 (10×10×10) contour templates separately for both ten Bengali and Chinese numeral signs achieving recognition accuracy of 95.80% for Bengali numeral signs and 95.90% for Chinese numeral signs with computational cost of 8.023 milliseconds per frame.