Chronological Self-Training for Real-Time Speaker Diarization

TL;DR

This paper introduces a chronological self-training method for real-time speaker diarization that significantly improves accuracy with minimal training data, achieving over 95% accuracy in just one second of enrollment.

Contribution

It proposes a novel self-training approach tailored for real-time diarization that reduces user interaction time while maintaining high accuracy.

Findings

Over 95% accuracy achieved with 1 second of training data

Average diarization error rate as low as 10% on diverse datasets

Effective across 6 different languages

Abstract

Diarization partitions an audio stream into segments based on the voices of the speakers. Real-time diarization systems that include an enrollment step should limit enrollment training samples to reduce user interaction time. Although training on a small number of samples yields poor performance, we show that the accuracy can be improved dramatically using a chronological self-training approach. We studied the tradeoff between training time and classification performance and found that 1 second is sufficient to reach over 95% accuracy. We evaluated on 700 audio conversation files of about 10 minutes each from 6 different languages and demonstrated average diarization error rates as low as 10%.