Machine learning for speaker recognition
Seminar with Sandro Cumani (DAUIN - Politecnico di Torino)
The growth of possible applications, the increase in computational power, and the widespread adoption of portable devices has produced, in the last years, an increasing interest from the academic community, industries and governments in automatic systems able to process speech segments.
As a result, speech technology is becoming more and more pervasive, as it allows for natural user-machine interactions. In the field of Cybersecurity, voice can be effectively employed as alternative to fingerprint or face recognition. Furthermore, speaker identification plays a relevant role in forensics and surveillance.
Due to the importance of these topics, the Speech Recognition Group (SRG) of Politecnico di Torino has been actively involved for several years in the development of novel speaker identification and verification approaches based on Machine Learning and Artificial Intelligence techniques.
This talk presents an overview of the main machine learning approaches to speaker recognition, focusing on the contributions of the SRG group to the state-of-the-art. In particular, the talk will address the following topics: speech segment representation, computational complexity of voiceprint extraction, voiceprint classification, score calibration.
Sarà disponibile un leggero rinfresco per coloro che si saranno registrati entro il 19 febbraio sul sito di Eventbrite.