Hello, I try to design with tensorflow a model to recognize speaker with audio wav files. I use VoxCeleb dataset. I read lots of paper on it but I don't achieve to design a proper model. I just get 30% accuracy on my dataset. I just use MFCC algorithm to extract patterns of my sound signal. I put the patterns into a "mnist like 2D model". But the result is not really good. Can someone give me advices?