A VTS-based feature compensation approach to noisy speech recognition using mixture models of distortion | IEEE Conference Publication | IEEE Xplore