Multi-Label Training for Text-Independent Speaker Identification

TL;DR

This paper introduces Multi-Label Training (MLT) for text-independent speaker identification, which improves accuracy by assigning multiple labels to speech segments, leveraging ensemble-like benefits without high computational costs.

Contribution

The paper proposes a novel Multi-Label Training strategy that enhances speaker identification accuracy and robustness, applicable to existing models without significant computational overhead.

Findings

MLT improves identification accuracy in both clean and noisy conditions.

MLT leverages ensemble effects to enhance model performance.

The strategy is easily adaptable to current speaker identification systems.

Abstract

In this paper, we propose a novel strategy for text-independent speaker identification system: Multi-Label Training (MLT). Instead of the commonly used one-to-one correspondence between the speech and the speaker label, we divide all the speeches of each speaker into several subgroups, with each subgroup assigned a different set of labels. During the identification process, a specific speaker is identified as long as the predicted label is the same as one of his/her corresponding labels. We found that this method can force the model to distinguish the data more accurately, and somehow takes advantages of ensemble learning, while avoiding the significant increase of computation and storage burden. In the experiments, we found that not only in clean conditions, but also in noisy conditions with speech enhancement, Multi-Label Training can still achieve better identification performance than commom methods. It should be noted that the proposed strategy can be easily applied to almost all current text-independent speaker identification models to achieve further improvements.