Open-set Short Utterance Forensic Speaker Verification using Teacher-Student Network with Explicit Inductive Bias

TL;DR

This paper introduces a teacher-student network approach with explicit inductive bias to enhance forensic speaker verification on small, challenging datasets of short utterances recorded in uncontrolled environments.

Contribution

It proposes a novel knowledge distillation objective function and a fine-tuning strategy to improve speaker verification accuracy in forensic scenarios with limited data.

Findings

The proposed objective function improves teacher-student learning on short utterances.

Fine-tuning with the new strategy outperforms weight decay in domain adaptation.

The approach achieves better verification performance on the forensic corpus.

Abstract

In forensic applications, it is very common that only small naturalistic datasets consisting of short utterances in complex or unknown acoustic environments are available. In this study, we propose a pipeline solution to improve speaker verification on a small actual forensic field dataset. By leveraging large-scale out-of-domain datasets, a knowledge distillation based objective function is proposed for teacher-student learning, which is applied for short utterance forensic speaker verification. The objective function collectively considers speaker classification loss, Kullback-Leibler divergence, and similarity of embeddings. In order to advance the trained deep speaker embedding network to be robust for a small target dataset, we introduce a novel strategy to fine-tune the pre-trained student model towards a forensic target domain by utilizing the model as a finetuning start point and a reference in regularization. The proposed approaches are evaluated on the 1st48-UTD forensic corpus, a newly established naturalistic dataset of actual homicide investigations consisting of short utterances recorded in uncontrolled conditions. We show that the proposed objective function can efficiently improve the performance of teacher-student learning on short utterances and that our fine-tuning strategy outperforms the commonly used weight decay method by providing an explicit inductive bias towards the pre-trained model.