用於說話人識別的度量學習的防禦（CS SD）

• 2020 年 3 月 27 日
• 筆記

本文的目標是對不可見的說話者進行「開集」的說話者識別，其中理想的嵌入應該能夠將資訊壓縮成具有小的類內(同一說話者)和大的類間(不同說話者)距離的緊湊的話語級表示。

人們普遍認為，通過分類目標訓練的網路要優於度量學習方法。在本文中，我們對VoxCeleb數據集上的說話人識別的最新丟失函數進行了廣泛的評估。我們的研究表明：與基於分類的損失相比，即使是最普通的三重損失，也顯示出了良好的競爭性能；而那些使用度量學習目標訓練的三重損失，也比最先進的方法表現得更好。

原文題目：In defence of metric learning for speaker recognition

原文：The objective of this paper is 'open-set' speaker recognition of unseen speakers, where ideal embeddings should be able to condense information into a compact utterance-level representation that has small intra-class (same speaker) and large inter-class (different speakers) distance.

A popular belief in speaker recognition is that networks trained with classification objectives outperform metric learning methods. In this paper, we present an extensive evaluation of most recent loss functions for speaker recognition on the VoxCeleb dataset. We demonstrate that even the vanilla triplet loss shows competitive performance compared to classification-based losses, and those trained with our angular metric learning objective outperform state-of-the-art methods.

原文作者：Joon Son Chung, Jaesung Huh, Seongkyu Mun, Minjae Lee, Hee Soo Heo, Soyeon Choe, Chiheon Ham, Sunghwan Jung, Bong-Jin Lee, Icksang Han

原文地址：https://arxiv.org/abs/2003.11982