SELD U-Net: Joint Optimization of Sound Event Localization and Detection with Noise Reduction

Sound event localization and detection (SELD) is a combined task that classifies acoustic events from audio signals, estimates temporal boundaries, identifies locations. With the advancement of industries utilizing SELD has been applied in various fields, deep-learning-based research being conducted for its effective application. However, current focuses mainly on performance improvement noise-free environments, which leads to degradation issues noisy environments. To address this problem, study proposes robust U-Net model performs The proposed combines remove noise SELDnet perform SELD. was trained evaluated using environmental data with sizes. Consequently, it confirmed superior compared existing deep learning-based models environments high levels noise.