Sign language recognition using the Extreme Learning Machine
Mosiuoa M. Sole, Mohohlo Samuel Tšoeu · 2011
The Extreme Learning Machine (ELM) is a simplified neural network. It non-linearly embeds input data in a higher dimensional space using randomly generated sigmoidal basis functions. The training target vector is then approximated by a linear weighted sum of these basis functions. In this paper the ELM algorithm has been applied to classify static hand gestures that represent different letters of the Auslan(Australian Sign Language) dictionary. ELM offers very fast learning with very consistent performance. It has only one tuning parameter. It can be adapted to multiclass classification and multi output regression without any increase in training time. Its low computational intensity and short training time makes it superior to traditional algorithms such as Hidden Markov Models (HMMs), Single and Recurrent Feed Forward Neural Networks for real time translations. Preliminary experimental results have shown that ELM can produce good generalization performance as a classifier. Increasing the dimensionality of the data results in better separation. Consequently the gestures become more distinguishable improving the probability of correct classification. An investigation into its classification performance for the entire alphabet is currently under way.