A Comparative Study of Bottom-Up and Top-Down Approaches to Speaker Diarization

Abstract

Abstract-This paper presents a theoretical framework to analyze the relative merits of the two most general, dominant approaches to speaker diarization involving bottom-up and top-down hierarchical clustering. We present an original qualitative comparison which argues how the two approaches are likely to exhibit different behavior in speaker inventory optimization and model training: bottom-up approaches will capture comparatively purer models and will thus be more sensitive to nuisance variation such as that related to the speech content; top-down approaches, in contrast, will produce less discriminative speaker models but, importantly, models which are potentially better normalized against nuisance variation. We report experiments conducted on two standard, single-channel NIST RT evaluation datasets which validate our hypotheses. Results show that competitive performance can be achieved with both bottom-up and top-down approaches (average DERs of 21% and 22%), and that neither approach is superior. Speaker purification, which aims to improve speaker discrimination, gives more consistent improvements with the top-down system than with the bottom-up system (average DERs of 19% and 25%), thereby confirming that the top-down system is less discriminative and that the bottom-up system is less stable. Finally, we report a new combination strategy that exploits the merits of the two approaches. Combination delivers an average DER of 17% and confirms the intrinsic complementary of the two approaches.

Latent class model with application to speaker diarization

“…The top-down approach is reported to give worse performance on the NIST RT database [25] and has thus received less attention. However, paper [40] makes a thorough comparative study of these two approaches and demonstrates that these two approaches have similar performance.…”

“…Although random initialization works well in most cases, LCM and VB systems tend to assign the segments to each speaker evenly in the case where a single speaker dominates the whole conversation, leading to poor results. According to the comparative study [40], we know that the bottom-up approach will capture comparatively purer models. Therefore, we recommend an informative AHC initialization method, similar to our previous paper [51].…”

Adaptive and online speaker diarization for meeting data

“…Speaker diarization [1][2][3] is an unsupervised statistical pattern recognition task which aims to determine 'who spoke when' in a given audio stream. Speaker diarization has become a key, enabling technology in a wide variety of tasks including document processing, structuring and navigation, information retrieval, meta-data extraction and copyright detection.…”

“…Historically, the state-of-the art in speaker diarization for meetings has evolved around the implementation of offline systems, such as bottom-up and top-down hierarchical clustering approaches [3][4][5]. In both cases, speakers are modelled with Gaussian mixture models (GMMs) which are interconnected to form an ergodic hidden Markov model (HMM) in which the transitions represent speaker turns.…”

Variational Bayes based I-vector for speaker diarization of telephone conversations

“…As for speaker diarization, many research works are based on agglomerative and divisive hierachical manner such as top-down or bottom-up algorithms [2]. The bottom-up approach is by far the most popular system, that is, hierachical agglomerative clustering (HAC).…”