Triplet Based Embedding Distance and Similarity Learning for Text-independent Speaker Verification | IEEE Conference Publication | IEEE Xplore