Automatic speech recognition and intrinsic speech variation