BINAQUAL: A Full-Reference Objective Localization Similarity Metric for Binaural Audio

TL;DR

BINAQUAL introduces a full-reference objective metric for binaural localization similarity by adapting the AMBIQUAL framework to two-channel playback. It computes NSIM-based phaseogram similarities per channel and combines them via a product, operating on a 2048-point STFT with a 32-band gammatone filterbank to yield a score in $[0,1]$. The authors validate BINAQUAL against five research questions using four synthetic datasets and demonstrate strong correlation with subjective MUSHRA scores, while revealing perceptual limitations such as front-back confusion and pure-tone localization challenges. The work provides a practical benchmark for evaluating spatial fidelity in binaural audio and releases the SynBAD dataset and code to support reproducibility and further research.

Abstract

Spatial audio enhances immersion in applications such as virtual reality, augmented reality, gaming, and cinema by creating a three-dimensional auditory experience. Ensuring the spatial fidelity of binaural audio is crucial, given that processes such as compression, encoding, or transmission can alter localization cues. While subjective listening tests like MUSHRA remain the gold standard for evaluating spatial localization quality, they are costly and time-consuming. This paper introduces BINAQUAL, a full-reference objective metric designed to assess localization similarity in binaural audio recordings. BINAQUAL adapts the AMBIQUAL metric, originally developed for localization quality assessment in ambisonics audio format to the binaural domain. We evaluate BINAQUAL across five key research questions, examining its sensitivity to variations in sound source locations, angle interpolations, surround speaker layouts, audio degradations, and content diversity. Results demonstrate that BINAQUAL effectively differentiates between subtle spatial variations and correlates strongly with subjective listening tests, making it a reliable metric for binaural localization quality assessment. The proposed metric provides a robust benchmark for ensuring spatial accuracy in binaural audio processing, paving the way for improved objective evaluations in immersive audio applications.

Figures (6)

• Figure 1: A high-level representation of the BINAQUAL model. (A) The process to calculate NSIM scores for each channel of reference and test signals. (B) The process to calculate localization similarity from NSIM scores.
• Figure 2: Localization similarity as a function of azimuth and elevation with fixed reference audio source, localized at azimuth=0°, elevation=0° for (a) Castanets and (b) Female speech w. reverb test signals.
• Figure 3: Localization similarity scores distributed on a half sphere for various contents. The corresponding reference audio sources were localized at azimuth = 0°, elevation = 0°. Box-Cox transform was applied to LS scores to spread out the distribution and make variations more noticeable.
• Figure 4: Localization similarity between single-source interpolated test signals in the angle interpolations dataset and their corresponding references. (a) Reference is localized at azimuth = 50°, elevation = 0°. (b) Reference is localized at azimuth = 90°, elevation = 0°.
• Figure 5: The average localization similarity between multi-point source samples of the surround layouts dataset, rendered for different loudspeaker layouts, and corresponding reference signals without any layouts. (a) Point source angles without elevation, (b) Point source angles with elevation.