Novel speech processing techniques for robust automatic speech recognition