We propose a deep learning-based method for speaker verification. Our team has worked on this project for more than 1 year and the accuracy has passed over benchmarks such as the accuracy of the paper by Andrew Zisserman Group at Oxford University. Contrary to other methods that are text-dependent, our system is text and language independent.

The processing speed of our model is less than 1 sec. The system identifies a person by just checking two voices with a length of 4 secs. Our system works perfectly in English, French, Spanish, German, Persian and Arabic, even in a noisy environment.