Combination of deep speaker embeddings for diarisation. (September 2021)

• Record Type: Journal Article
• Authors: Sun, Guangzhi; Zhang, Chao; Woodland, Philip C.
• Is Part Of: Neural networks. Volume 141(2021)
• Page Start: 372
• Page End: 384
• Publication Date: 2021-09
• Subjects: Speaker diarisation -- System combination -- Speaker embedding -- Attention mechanism -- Gating mechanism -- Bilinear pooling; Neural computers -- Periodicals; Neural networks (Computer science) -- Periodicals; Neural networks (Neurobiology) -- Periodicals; Nervous System -- Periodicals; Ordinateurs neuronaux -- Périodiques; Réseaux neuronaux (Informatique) -- Périodiques; Réseaux neuronaux (Neurobiologie) -- Périodiques; Neural computers; Neural networks (Computer science); Neural networks (Neurobiology); Periodicals; 006.32
• DOI: 10.1016/j.neunet.2021.04.020
• Languages: English
• ISSNs: 0893-6080

Abstract

Abstract: Significant progress has recently been made in speaker diarisation after the introduction of d-vectors as speaker embeddings extracted from neural network (NN) speaker classifiers for clustering speech segments. To extract better-performing and more robust speaker embeddings, this paper proposes a c-vector method by combining multiple sets of complementary d-vectors derived from systems with different NN components. Three structures are used to implement the c-vectors, namely 2D self-attentive, gated additive, and bilinear pooling structures, relying on attention mechanisms, a gating mechanism, and a low-rank bilinear pooling mechanism respectively. Furthermore, a neural-based single-pass speaker diarisation pipeline is also proposed in this paper, which uses NNs to achieve voice activity detection, speaker change point detection, and speaker embedding extraction. Experiments and detailed analyses are conducted on the challenging AMI and NIST RT05 datasets which consist of real meetings with 4–10 speakers and a wide range of acoustic conditions. For systems trained on the AMI training set, relative speaker error rate (SER) reductions of 13% and 29% are obtained by using c-vectors instead of d-vectors on the AMI dev and eval sets respectively, and a relative SER reduction of 15% in SER is observed on RT05, which shows the robustness of the proposed methods. By incorporating VoxCeleb data into the training set, the best c-vector system achieved 7%, 17% and 16% relative SER reduction compared to the d-vector on the AMI dev, eval and RT05 sets respectively. Highlights: A complete single pass neural network-based diarisation pipeline 2D self-attentive, gated additive and bilinear pooling combination methods Stacked combination using 2D self-attentive and a bilinear pooling structure 29% relative SER reductions on the AMI eval set using AMI training data 17% relative SER reductions on AMI eval set using additional VoxCeleb training data