Abstract: In the last decade, the demand for learning through mobile devices has been increased, however, the security and authentication of these systems have less attention. This is because of researchers desirability to be more famous by unauthenticated publishing of their articles. Therefore, this study presents a proposed voice authentication for mobile learning (m-learning) system as a secure solution. In the proposed system, each of the server and clients (learners) in the designed learning system is provided with voice features extraction algorithm. HPSO algorithm is used for extraction the wavelet frequency domain features. These extracted features are then matched with stored database in order to give the permission of learning system accessibility. For (LL subband) FAR is 0.0, FRR is 0.01 and CVR is 99%. For (LL, LH, HL and HH) FAR is 0.0, FRR is 0.0 and CVR is 100%. The voice recognition time is about 1.04 sec.