A new method of text-dependent speaker identification using discriminative centroids weighting is proposed in this paper. The characteristics of the proposed method are as follows: feature parameters extraction, vector quantization with the growing cell structures (GCS) algorithm, stochastic fine-tuning of codebooks and discriminative centroids weighting (DCW) according to the uniqueness of personal features. The algorithm is evaluated on a database that includes 25 speakers each of them recorded in 24 different sessions. All 25 speakers spoke the same phrase for 240 times. The overall performance of the system was 99.5 %.