To test the performance of vowel cepstrum in forensic speaker recognition, a forensic speaker identification method based on likelihood ratio and MFCC features is presented in this paper. This method is tested in vowel /a/ of 45 people’s telephone dialog recordings and shows high identification ratio. Experiment results show that not only can the method identify the speaker, but also quantify the evidence strength according the acoustic difference between the questioned recording and the suspect’s recording, and provide the scientific and reasonable evaluation results to court. Compare to the manual method in formants extraction and pitch extraction, the usage of auto extraction of features increases efficiency and performance of forensic speaker identification system.