Improving speech recognition performance via phone-dependent VQ codebooks and adaptive language models in SPHINX-II | IEEE Conference Publication | IEEE Xplore