Meeting Topic: The simplicity of hypercomplexity: deep learning for immersive audio
Moderator Name: Aurelio Uncini
Speaker Name: Danilo Comminiello, Associate Professor at Università Sapienza di Roma - Italy
Meeting Location: Zoom virtual Meeting (5pm Central European Time/Italian time zone)
The Intelligent Signal Processing and Multi-Media (ISPAMM) research team of the Department of Information Engineering, Electronics and Telecommunications of Sapienza University of Rome, together with the Italian Section of the Audio Engineering Society, organized a series of meetings on the theme: Artificial Intelligence for Sound Synthesis and Analysis.
Learning sound sources in 3D space is an increasingly central problem in many immersive audio applications. Among the most promising recent methods for studying 3D audio and acoustic signals, deep learning algorithms in the hypercomplex domain show interesting capabilities in exploiting the internal dependencies of multidimensional signals. Spherical harmonic decomposition in the quaternionic hypercomplex domain allows 3D acoustic signals to be analyzed by quaternionic neural networks that learn 3D sound sources with excellent performance and significant computational savings compared to classical deep learning methods. Some application examples, such as 3D acoustic source identification and localization, will also be presented in the seminar.
Written By: Aurelio Uncini