Jahangir Alam

Robust Speech Representation Learning via Flow-based Embedding Regularization

2021 · arXiv (Cornell University)

Over the recent years, various deep learning-based methods were proposed for extracting a fixed-dimensional embedding vector from speech signals. Although the deep learning-based embedding extraction methods have shown good performance in numerous tasks including speaker …

An end-to-end approach for the verification problem: learning the right\n distance

2020 · arXiv (Cornell University)

In this contribution, we augment the metric learning setting by introducing a\nparametric pseudo-distance, trained jointly with the encoder. Several\ninterpretations are thus drawn for the learned distance-like model's output. We\nfirst show it approximates a likelihood ratio …