Leveraging phonetic context dependent invariant structure for continuous speech recognition | IEEE Conference Publication | IEEE Xplore