English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

This paper studies modulation spectrum features ($\Phi$) and mel-frequencycepstral coefficients ($\Psi$) in joint speaker diarization and identification(JSID). JSID is important as speaker diarization on its own to distinguishspeakers is insufficient for many applications, it is often necessary toidentify speakers as well. Machine learning models are set up usingconvolutional neural networks (CNNs) on $\Phi$ and recurrent neural networks$\unicode{x2013}$ long short-term memory (LSTMs) on $\Psi$, then concatenatinginto fully connected layers.  Experiment 1 shows models on both $\Phi$ and $\Psi$ have better diarizationerror rates (DERs) than models on either alone; a CNN on $\Phi$ has DER29.09\%, compared to 27.78\% for a LSTM on $\Psi$ and 19.44\% for a model onboth. Experiment 1 also investigates aleatoric uncertainties and shows themodel on both $\Phi$ and $\Psi$ has mean entropy 0.927~bits (out of 4~bits) forcorrect predictions compared to 1.896~bits for incorrect predictions which,along with entropy histogram shapes, shows the model helpfully indicates whereit is uncertain.  Experiment 2 investigates epistemic uncertainties as well as aleatoric usingMonte Carlo dropout (MCD). It compares models on both $\Phi$ and $\Psi$ withmodels trained on x-vectors ($X$), before applying Kalman filter smoothing onepistemic uncertainties for resegmentation and model ensembles. While the twomodels on $X$ (DERs 10.23\% and 9.74\%) outperform those on $\Phi$ and $\Psi$(DER 17.85\%) after their individual Kalman filter smoothing, combining themusing a Kalman filter smoothing method improves the DER to 9.29\%. Aleatoricuncertainties are higher for incorrect predictions.  Both Experiments show models on $\Phi$ do not distinguish overlappingspeakers as well as anticipated. However, Experiment 2 shows model ensembles dobetter with overlapping speakers than individual models do.

Uncertainty Quantification in Machine Learning for Joint Speaker Diarization and Identification