Cross-Task Adaptation and Speaker Adaptation in Air Traffic Control Tasks

When we want to develop a recognition system for a new environment, we have to decide which is the best option in what respect to the acoustic modeling: developing acoustic models from scratch using the data available for the new environment or to do cross-task adaptation starting from reliable HMM models. In this paper, we show the performance of several alternatives, comparing cross-task MAP and MLLR adaptation, in two speech recognizers related to air traffic control tasks, one for spontaneous speech and the other one for a command interface. For the spontaneous speech recognizer, we also include the comparison between MAP and MLLR for speaker adaptation using a variable amount of adaptation data. We show how MAP outperforms MLLR when there is enough data available and the threshold points where both techniques provide similar performance or MLLR is better than MAP. In all cases, we show the effectiveness of variance adaptation.