Speaker: Dr. hab. Szymon DrgasPoznan University of Technology, Poznań, Poland
Will speak about: "Machine learning-based methods for hearing loss compensation and speech intelligibility prediction"
Date, time: March 4 at 13:00 in the Prof. Arkadiusz Piekara Auditorium, Collegium Physicum.
Abstract: Recent advances in machine learning enable the estimation of a clean, intelligible target signal from a noisy input. For hearing-impaired listeners, however, this task is more challenging because the definition of the target signal is not well established; that is, it remains unclear which signal would provide optimal intelligibility and quality for an individual with a specific hearing loss profile.
In this talk, I will discuss various auditory models, neural network architectures for speech enhancement, and loss functions that incorporate characteristics of a listener’s hearing condition. I will also address the combination of neural networks and traditional signal processing methods for speech intelligibility prediction.
Chairman: Prof. A. Józefczak
Additional Information: All are invited to Coffee with the Speaker, starting at 12:30 in front of the auditorium. Please follow the news on the MTPR website (http://mtpr.amu.edu.pl), where information about upcoming events will be available soon.
Those who cannot participate in person can follow the talk using Microsoft Teams: