Parallel implementation of Artificial Neural Network training for speech recognition