Unsupervised speaker adaptation of deep neural network based on the combination of speaker codes and singular value decomposition for speech recognition | IEEE Conference Publication | IEEE Xplore