Attentive Statistics Pooling for Deep Speaker Embedding

TLDR

Conventional speaker embedding averages frame‑level features across all frames of an utterance to produce an utterance‑level representation. The paper proposes attentive statistics pooling for deep speaker embedding in text‑independent speaker verification. The method uses an attention mechanism to weight frames, producing weighted means and weighted standard deviations for speaker embeddings. The approach captures long‑term speaker variations and reduces EERs by 7.5 % on NIST SRE 2012 and 8.1 % on VoxCeleb compared to conventional pooling.

Abstract

This paper proposes attentive statistics pooling for deep speaker embedding in text-independent speaker verification. In conventional speaker embedding, frame-level features are averaged over all the frames of a single utterance to form an utterance-level feature. Our method utilizes an attention mechanism to give different weights to different frames and generates not only weighted means but also weighted standard deviations. In this way, it can capture long-term variations in speaker characteristics more effectively. An evaluation on the NIST SRE 2012 and the VoxCeleb data sets shows that it reduces equal error rates (EERs) from the conventional method by 7.5% and 8.1%, respectively.