The most commonly method use by people to protect their secured data or information is using password or PIN/ID protection. This method require user to authenticate themselves by entering password that they had already created. However, due to lack of security the data is not secured enough. There are cases of fraud and theft when people can easily know the password. But as time goes by, there is a new technology known as Biometric Identification System. It uses biometric characteristics of an individual that is unique and different from everyone else and therefore can be use to authenticate the user authority access. This research is an implementation of speech recognition as medium security access control to restricted services such as phone banking system, voicemail or access to database services. First, speaker signal will go to pre-treatment process, where it will remove the background noise. Then, features from speech signal will be extracted using Me1 Frequency Cepstrum Coefficients (MFCC) method. Then, using Vector Quantization, the features will be matched with the reference speech in database. The real speaker is identified by clustering the speech signal from the tested speaker to codebook of each speaker using K-means algorithm and the speaker with the lowest distortion Euclidean distances is chose as the correct speaker. The main focus of tlns research is speaker identification, which compared speech signal from unknown speaker to a database of known speaker using text-dependent utterances. From the experimental results, this method has showed that it was able to recognize the correct voice perfectly.