Statistics for Feature extraction and selection for speech recognition