Efficient text-independent speaker verification with structural Gaussian mixture models and neural network | IEEE Journals & Magazine | IEEE Xplore