Robust bi-modal speech recognition based on state synchronous modeling and stream weight optimization | IEEE Conference Publication | IEEE Xplore