Satellite Workshop

on


Speaker and Language Identification, Verification, and Diarization


2nd December 2023



 

 Organized by

 National Institute of Technology Goa , India



INTRODUCTION


Speaker and language identification are two important speech processing tasks. It is well known that the speech has various kinds of information such as what is spoken, who has spoken, in what emotion it is spoken, which language was used etc.  Speaker identification and verification is important for voice based authentication systems. In a multi-speaker conversation which is very common in a meeting room, segmenting the speech based on speaker switching point is essential to make sure who spoke what at what time. The process of splitting a speech signal involving multiple speakers into segments of speech corresponding to individual speakers is speaker diarization. Language identification involves identifying the identity of language spoken in speech utterance. It is desirable for the language identification systems to be speaker independent. A language diarization system involves automatically identifying and segmenting languages in a code-switch conversation. This is very helpful in automatic processing of speech signals in multilingual societies like india. Hence the theme of this satellite event corresponding to SPECOM 2023 is “Speaker and Language Identification, Verification and Diarization”. This workshop is a paper based workshop that involves two invited talks and paper presentations concerned to the theme of the workshop. 

INVITED TALKS

Prof. Hema A. Murthy

IIT Madras, India

Title: Signal Processing guided Machine Learning

Abstract: The buzzword for building applications of practical relevance is “big data” today. This has led to a separate field called “Data Science” being offered by various universities, and Institutes. The field of “data science” has grown to accommodate for the variability in the underlying statistical structure that exists natural signals. Both classical machine learning and deep learning rely on the availability of large amount of a wide variety of data. Deep learning models which are massive neural networks ultimately learn the underlying structure of data.

While Deep Learning has revolutionized machine learning, in this talk we focus on the use of signal processing to preprocess or mine existing data, so that accurate data is presented to machine learning models. Domain specific signal processing has the capability to identify events in a signal. The event itself may have varying statistical characteristics. Presentation of detected events to machine learning enables faster convergence, and a smaller data footprint. We draw examples from Speech and EEG signals.

Dr. Oldřich Plchot 

Brno University of Technology, Czechia

Title: Current and emerging trends in extracting speaker embeddings 

Abstract: This talk will cover state-of-the-art and emerging methods for extracting speaker representations (embeddings) from speech. We will compare unsupervised, self-supervised, weakly supervised, and fully supervised approaches and discuss various applications and use cases that fit the methods. More focus will be given to self-supervised methods with Transformers and their use for extracting speaker representations, as these models have quickly risen in popularity and become an integral part of state-of-the-art speech modeling for automatic speech recognition. Apart from fine-tuning them to extract speaker embeddings, we will discuss strategies for domain adaptation and self-pretraining.

PAPER SUBMISSION DETAILS

Papers are invited from researchers who are working on the topics under the theme of the workshop, “Speaker and Language Identification, Verification and diarization”.  Also the researchers who had participated in Indic Multi-Lingual Speaker Verification (I-MSV 2022) (https://vlsp.org.vn/cocosda2022/i-msv) and DISPLACE challenge (https://displace2023.github.io/) are encouraged to submit their result to this workshop. To encourage research in this direction, anybody who did not participate in the said challenge can still work on the datasets shared in the above link and submit their research findings in the form of a paper. The paper submission details can be found at: https://www.iitdh.ac.in/specom-2023/Submissions.html .



VENUE

The satellite workshop will be organized at Abolim Hall of International Centre Goa, Dona Paula (ICG-Dona Paula). 

PROGRAM

For program schedule for Satellite Workshop, visit out Program Schedule page.

REGISTRATION

For further details on Registration, visit our Registration page.