The field of speech separation and spatial audio is moving towards more realistic and diverse acoustic scenarios, with a focus on improving speech accessibility for children in noisy classrooms and advancing robust and efficient direction-of-arrival estimation. Researchers are exploring new architectures and training strategies to improve speech separation quality, such as spatially aware architectures and targeted adaptation. Additionally, there is a growing interest in using large language models to synthesize more realistic spatial audio scenes. Noteworthy papers include:

• A study on speech separation for hearing-impaired children in the classroom, which demonstrated that spatially aware architectures combined with targeted adaptation can improve speech accessibility.
• A paper on DOA estimation with a lightweight network on LLM-aided simulated acoustic scenes, which proposed a lightweight DOA estimation model that achieves satisfactory accuracy and robustness while maintaining low computational complexity.