发送短信: Investigation on model selection criteria for speaker identification /