As the technology is expanding its roots in multimedia, the peoples are more concerning about the security and speed of their work. By viewing this, the need of voice recognition system is increasing. So, this has become a dominant research topic. It includes training and testing of voices. Sampling, feature extraction, database creation, classifying all these steps are used to recognize a voice. Voice recognition are of two types. First is speaker dependent which require training of software, by reading some pages of text. Second one is speaker independent in which there is no requirement of software training. This can be done in time domain by correlation but, it will lack accuracy and speed, so fast Fourier transform is used to find its features. Although, there are many limitations of it for e.g. variation in person’s voice with time, crowded rooms etc.