In this paper we discuss the relevance of bandwidth extension for speaker identification tasks. Mainly we want to study if it is possible to recognize voices that have been bandwith extended. For this purpose, we created two different databases (microphonic and ISDN) of speech signals that were bandwidth extended from telephone bandwidth ([300, 3400] Hz) to full bandwidth ([100, 8000] Hz). We have evaluated different parameterizations, and we have found that the MELCEPST parameterization can take advantage of the bandwidth extension algorithms in several situations.