TCG CREST System Description for the DISPLACE-M Challenge

2026-03-02

AI Summary

This work addresses the degradation of speaker diarization performance in rural healthcare settings due to acoustic noise, within the context of Track 1 of the DISPLACE-M Challenge. It systematically evaluates the impact of voice activity detection (VAD) and clustering strategies, proposing an agglomerative hierarchical clustering (AHC) approach enhanced with large-window median filtering and several improved spectral clustering variants—SC-adapt, SC-PNA, and SC-MK. The framework integrates ECAPA-TDNN embeddings, WavLM pretrained features, and an end-to-end Diarizen architecture. Compared to the SpeechBrain baseline, the Diarizen system achieves a 39% relative reduction in diarization error rate (DER). The best-performing system attains DERs of 10.37% and 9.21% on the development and evaluation sets, respectively, securing sixth place in Phase I of the challenge.

Technology Category

• Knowledge Representation and Reasoning: Diagnosis and Abductive Reasoning
• Natural Language Processing: Speech
• Machine Learning: Clustering

Application Category

• Search and Retrieval-Augmented AI: Vertical and domain-specific search
• Systems and Infrastructure for Web, Mobile and WoT: Experiences and lessons learnt from Web-based algorithms and system deployments
• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation

Abstract

This report presents the TCG CREST system description for Track 1 (Speaker Diarization) of the DISPLACE-M challenge, focusing on naturalistic medical conversations in noisy rural-healthcare scenarios. Our study evaluates the impact of various voice activity detection (VAD) methods and advanced clustering algorithms on overall speaker diarization (SD) performance. We compare and analyze two SD frameworks: a modular pipeline utilizing SpeechBrain with ECAPA-TDNN embeddings, and a state-of-the-art (SOTA) hybrid end-to-end neural diarization system, Diarizen, built on top of a pre-trained WavLM. With these frameworks, we explore diverse clustering techniques, including agglomerative hierarchical clustering (AHC), and multiple novel variants of spectral clustering, such as SC-adapt, SC-PNA, and SC-MK. Experimental results demonstrate that the Diarizen system provides an approximate $39\%$ relative improvement in the diarization error rate (DER) on the post-evaluation analysis of Phase~I compared to the SpeechBrain baseline. Our best-performing submitted system employing the Diarizen baseline with AHC employing a median filtering with a larger context window of $29$ achieved a DER of 10.37\% on the development and 9.21\% on the evaluation sets, respectively. Our team ranked sixth out of the 11 participating teams after the Phase~I evaluation.

Problem

• speaker diarization
• noisy rural-healthcare scenarios
• naturalistic medical conversations
• diarization error rate
• voice activity detection

Innovation

• speaker diarization
• Diarizen
• spectral clustering
• WavLM
• agglomerative hierarchical clustering