Learning Deep Embedding with Acoustic and Phoneme Features for Speaker Recognition in FM Broadcasting

Figure 2

Results for the hybrid DNNs with different loss functions and structures.