Can envelope recovery account for speech recognition based on temporal fine structure