Universal speaker recognition encoders for different speech segments duration

TL;DR

This paper proposes a universal speaker encoder trained to perform reliably across varying speech segment durations, improving speaker verification accuracy without increasing inference time.

Contribution

The paper introduces a simple training recipe for creating universal speaker encoders applicable to any neural network architecture, enhancing robustness across speech durations.

Findings

Universal encoder improves verification for different speech durations

Evaluation on NIST SRE and VoxCeleb1 benchmarks shows performance gains

Encoder maintains same inference time as the underlying neural network

Abstract

Creating universal speaker encoders which are robust for different acoustic and speech duration conditions is a big challenge today. According to our observations systems trained on short speech segments are optimal for short phrase speaker verification and systems trained on long segments are superior for long segments verification. A system trained simultaneously on pooled short and long speech segments does not give optimal verification results and usually degrades both for short and long segments. This paper addresses the problem of creating universal speaker encoders for different speech segments duration. We describe our simple recipe for training universal speaker encoder for any type of selected neural network architecture. According to our evaluation results of wav2vec-TDNN based systems obtained for NIST SRE and VoxCeleb1 benchmarks the proposed universal encoder provides speaker verification improvements in case of different enrollment and test speech segment duration. The key feature of the proposed encoder is that it has the same inference time as the selected neural network architecture.