Improving monaural speaker identification by double-talk detection