Spatial segmentation of impulse response for room reflection analysis and auralization

Parametric spatial audio reproduction methods enable the analysis and synthesis of room acoustics using measured spatial room impulse responses (SRIRs). However, limitations in representing early reflections can restrict control over these components. This paper introduces the Spatial Segmentation of Impulse Response (SSIR) method to address these challenges and evaluates its perceptual accuracy against existing approaches. SSIR detects and segments early sound events, facilitating a more compact spatial representation and manipulation of individual sound events while maintaining the parametric framework of the Spatial Decomposition Method (SDM). Two perceptual experiments evaluated SSIR: the first assessed overall fidelity, comparing SSIR against an optimized SDM variant (SDM-BPE), a reflection extraction method, and a dummy head reference; the second evaluated spatial and timbral fidelity, testing SSIR against the original SDM formulation and SDM-BPE. Results showed significant improvements in spatial and timbral fidelity over standard SDM. Importantly, no significant perceptual differences were found between SSIR and SDM-BPE. SSIR outperformed an early reflection extraction method. Objective metrics indicated no significant impairments in apparent source width or externalization. SSIR maintains perceptual performance equivalent to state-of-the-art methods while reducing the complexity of spatial data and enabling reflection control. This makes it suitable for psychoacoustic studies utilizing parametric spatial audio reproduction systems.