Minimum Bayes error feature selection

We consider the problem of designing a linear transformation 2 IR , of rank p n, which projects the features of a classi er x 2 IR onto y = x 2 IR such as to achieve minimum Bayes error (or probability of misclassi cation). Two avenues will be explored: the rst is to maximize the -average divergence between the class densities and the second is to minimize the union Bhattacharyya bound in the range of . While both approaches yield similar performance in practice, they outperform standard LDA features and show a 10% relative improvement in the word error rate over state-of-the-art cepstral features on a large vocabulary telephony speech recognition task.