Memory efficient sign language recognition system based on WiSARD weightless neural network technique

A novel single hand Sign Language Recognition (SLR) method utilizing Weightless Neural Network (WNN) known as a RAMnet or n-tuple network is evaluated as a memory efficient technique. In contrast with standard multilayer perceptron neural network (MLPNN), the RAMnet does not require long iteration of presentation of training data in the training phase i.e. long training time, weights adjustments and activation function calculations. This alternative technique is based on WiSARD Neural Network. Two algorithms are proposed and implemented: one for hand tracking through image sequences, and the second for SLR using WiSARD. The pixels containing the hand region are used to train the Sign Language (SL) alphabetical symbol discriminators. Each sign is assigned a discriminator, thus there are as many discriminator as signs to be recognized. The system performance is checked by evaluating for single handed SL English alphabets and produced result of accuracy 75%. The conventional Random Access Memory (RAM) is used to save and process binary image sequences in WiSARD that are addressed by Boolean input and produced Boolean output. The system can implement signs for any other sign language, thus it is language independent.