Depth-First Neural Architecture With Attentive Feature Fusion for Efficient Speaker Verification | IEEE Journals & Magazine | IEEE Xplore