Zero-Delay Spatial Audio Rendering for Immersive Networked Music Performances

Low-latency spatial audio processing is essential to reduce the overall latency in immersive networked music performances (NMP). In spatial audio rendering chains, most components, including source filtering for directivity, occlusion, and environmental effects, can be implemented with negligible latency using minimum-phase IIR filters. However, Head-Related Impulse Responses (HRIRs), which are fundamental to binaural rendering, often introduce non-negligible latency when measured HRIRs are used. In this paper, we discuss the sources of latency and present a zero-delay implementation based on minimum-phase FIR filters and delay lines. Through a controlled listening experiment involving 14 expert musicians, we evaluated the perceptual transparency of the zero-delay rendering with respect to a reference implementation without latency optimization. The experiment comprised two zero-delay versions with different filter lengths and involved samples for five different source directions and four different signal types. The results showed that participants could not distinguish the reference from a zero-delay version with 64 filter taps at 48 kHz sampling rate, indicating that this configuration is suitable for spatial audio rendering in NMP systems.