We address the question of whether and how boosting and bagging can be used for speech recognition. In order to do this, we\r\ncompare two different boosting schemes, one at the phoneme level and one at the utterance level, with a phoneme-level bagging\r\nscheme. We control for many parameters and other choices, such as the state inference scheme used. In an unbiased experiment,\r\nwe clearly show that the gain of boosting methods compared to a single hidden Markov model is in all cases only marginal, while\r\nbagging significantly outperforms all other methods. We thus conclude that bagging methods, which have so far been overlooked\r\nin favour of boosting, should be examined more closely as a potentially useful ensemble learning technique for speech recognition.
Loading....