Spatial-aware Speaker Diarization for Multi-channel Multi-party Meeting

TL;DR

This paper introduces a spatial-aware speaker diarization system utilizing multi-channel audio, a novel neural network architecture, and spatial features to improve accuracy and overlapped speech detection in multi-party meetings.

Contribution

It proposes DMSNet, a novel multi-stream neural network with attention superdirective beamforming for robust speaker diarization in multi-channel recordings.

Findings

Achieved 93.53% accuracy in overlapped speech detection.

Reduced diarization error rate from 13.45% to 7.64%.

Enhanced robustness of speaker embeddings using spatial information.

Abstract

This paper describes a spatial-aware speaker diarization system for the multi-channel multi-party meeting. The diarization system obtains direction information of speaker by microphone array. Speaker spatial embedding is generated by xvector and s-vector derived from superdirective beamforming (SDB) which makes the embedding more robust. Specifically, we propose a novel multi-channel sequence-to-sequence neural network architecture named discriminative multi-stream neural network (DMSNet) which consists of attention superdirective beamforming (ASDB) block and Conformer encoder. The proposed ASDB is a self-adapted channel-wise block that extracts the latent spatial features of array audios by modeling interdependencies between channels. We explore DMSNet to address overlapped speech problem on multi-channel audio and achieve 93.53% accuracy on evaluation set. By performing DMSNet based overlapped speech detection (OSD) module, the diarization error rate (DER) of cluster-based diarization system decrease significantly from 13.45% to 7.64%.