Joint Training of Speaker Embedding Extractor, Speech and Overlap Detection for Diarization

TL;DR

This paper introduces a joint training approach for speaker embedding, speech activity, and overlap detection in diarization, reducing inference time and simplifying the pipeline compared to traditional modular systems.

Contribution

The work presents a unified model trained end-to-end for multiple diarization tasks, improving efficiency and performance over separate module training.

Findings

Achieves competitive diarization performance with reduced inference time.

Simplifies the diarization pipeline by integrating multiple tasks into a single model.

Demonstrates the feasibility of end-to-end training for speaker embedding, VAD, and OSD.

Abstract

In spite of the popularity of end-to-end diarization systems nowadays, modular systems comprised of voice activity detection (VAD), speaker embedding extraction plus clustering, and overlapped speech detection (OSD) plus handling still attain competitive performance in many conditions. However, one of the main drawbacks of modular systems is the need to run (and train) different modules independently. In this work, we propose an approach to jointly train a model to produce speaker embeddings, VAD and OSD simultaneously and reach competitive performance at a fraction of the inference time of a standard approach. Furthermore, the joint inference leads to a simplified overall pipeline which brings us one step closer to a unified clustering-based method that can be trained end-to-end towards a diarization-specific objective.