Target Sound

Target sound extraction (TSE) focuses on isolating a desired sound from a mixture of audio sources, preserving its spatial characteristics. Current research emphasizes developing robust algorithms, often employing deep learning architectures like transformers and diffusion probabilistic models, to achieve high-fidelity separation even in noisy environments and with limited or variable information about the target sound. These advancements are crucial for improving applications ranging from hearing aids and noise cancellation to enhancing the accuracy of audio-based surveillance and gravitational wave detection. The field is also exploring active methods, where an agent's movement is optimized to improve sound isolation.