Identification of Early Reflections using Sparse Recovery

• Applicant: Professor Dr.-Ing. Sascha Spors
• Subject Area: Acoustics
• Term: from 2016 to 2022
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 309256691

The analysis and design of room acoustics is an important research topic for concert halls, theatres, lecture rooms and public spaces. The acoustic response of a typical room can be split into three major contributions: the direct sound, early reflections and diffuse sound. Early reflections play a prominent role for the auditory impression of a room. It is therefore desirable to precisely determine the location and frequency response of early reflections. The spatio-temporal analysis of a sound field by spatially distributed microphones is state of the art in audio signal processing. Microphone arrays together with beamforming have been used to localize reflecting surfaces is rooms. The wavelength of audible sound ranges from 15 meters to 1.5 centimeters. Microphone arrays that could capture the spatial character of sound over the entire audible range require an infeasible number of microphones. This is a consequence of the Shannon Nyquist theorem. Current methods are therefore clearly limited in resolution. The sound field generated by early reflections can be approximated by a superposition of simple acoustic sources. This constitutes a sparse representation of the sound field, which can be exploited. Sparse recovery allows to sample a signal far below the Nyquist rate but still be fully reconstructed. The project aims at explicitly exploiting the sparsity of early reflections in order to precisely determine their location and frequency response. The research is accompanied by a structured evaluation of the algorithms on the basis of a scene database collected in the project.

• DFG Programme: Research Grants