Direct and Residual Subspace Decomposition of Spatial Room Impulse Responses

Psychoacoustic experiments have shown that directional properties of the direct sound, salient reflections, and late reverberation an acoustic room response can a distinct influence on auditory perception given room. Spatial impulse responses (SRIRs) capture those thus are used for direction-dependent analysis virtual rendering. This work proposes subspace method decomposes SRIRs into part, which comprises sound residual, to facilitate enhanced rendering methods by providing individual access these components. The proposed is based generalized singular value decomposition interprets residual as noise be separated from other components reverberation. Large values attributed then obtained low-rank approximation SRIR. By advancing end SRIR toward beginning while iteratively updating estimate, adapts spatio-temporal variations residual. evaluated using spatio-spectral error measure simulated different rooms, microphone arrays, ratios energy. creates lower errors than existing approaches in all tested scenarios, including scenario with two simultaneous reflections. A case study measured shows applicability under real-world conditions. reference implementation provided.