We implement a biometric authentication system on the Android platform, which is based on text-dependent speaker recognition. The Android version used in the application is Android 4.0. The application makes use of the Modular Audio Recognition Framework, from which many of the algorithms are adapted in the processes of preprocessing and feature extraction. In addition, we employ the Dynamic Time Warping (DTW) algorithm for the comparison of different voice features. A training procedure is implemented, using the DTW algorithm to align features. Furthermore, we introduce personal thresholds, based on which the performance for each individual user can be further optimized.We have carried out several tests in order to evaluate the performance of the developed system. The tests are performed on 16 persons, with in total 240 voice samples, of which 15 samples are from each person. As a result, for authentication, one of the optimal trade-offs of the False Acceptance Rate (FAR) and False Rejection Rate (FRR) achieved by the system is shown to be 13% and 12%, respectively. For identification, the system could identify the user correctly with a rate of 81%. Our results show that one can actually improve the system performance in terms of FAR and FRR significantly, through using the training procedure and the personal thresholds.