Discriminative speaker adaptation and environmental robustness in automatic speech recognition