In this study, voice based user identity recognition is introduced for visually impaired persons in learning management systems. The system designed for voice recognition process was performed both spoken and text dependent. For evaluating the performance of system, a total of 22 persons' voices (which 9 persons are visually impaired) recorded in a noiseless environment for training and testing process. Voice recognition mainly consists of preprocessing, feature extraction and classification (decision) steps. Mel Frequency Cepstral Coefficients (MFCC) method was used in the extraction step. In the speaker classification, that is, in the decision phase, Dynamic Time Wrapping (DTW) method, which is known to eliminate errors caused by delay in sound processing applications and give more accurate results, has been preferred. According to the simulation results, the performance of the presented system is 95%.