Hao Wu, Hongru Sun, Wanqing Li, Xinguo Yu, Hao Ming, Xiao Luo, Wenbin Zhang, Jiahong Zhao, Yi Guo, Jie Yang. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

In virtual reality (VR), sound sources are convolved with room impulse responses (RIRs) to create immersive and dynamic audio experiences. Assessing the quality of spatial audio synthesis in VR is challenging. Subjective listening tests are accurate, but they are time-consuming and costly. This paper introduces a novel objective quality metric to predict the listening quality (LQ) and localization accuracy (LA) of Ambisonic spatial audio at new positions using spatially interpolated or extrapolated first-order Ambisonic (FOA) RIRs based on known nearby FOA RIRs. The LQ and LA scores are computed based on the Kolmogorov-Smirnov test, to measure the similarity between segments of direct sound and reflections of reference and synthesized FOA RIRs. Results show that these scores strongly correlate with subjective test results, proving the reliability of the proposed method. A major advantage is that it predicts spatial audio quality directly from synthesized FOA RIRs, without requiring convolved signals or being affected by specific sound sources, providing a more practical and efficient solution for development.

Audio adversarial examples impose acoustically imperceptible perturbations to clean audio examples, fooling classification models into producing incorrect results. Transferability is a critical property of audio adversarial examples, making black-box attacks applicable in practice and attracting increasing interest. Despite recent studies achieving transferability across models within the same domain, they consistently fail to achieve transferability across different domains. Given that time-domain and frequency-domain models are the two predominant approaches in audio classification, we observe that adversarial examples generated for one domain demonstrate significantly constrained transferability to the other. To address this limitation, we propose an Adaptive Inter-domain Ensemble (AIE) attack, which integrates transferable adversarial information from both domains and dynamically optimizes their contributions through adaptive weighting, improving the cross-domain transferability of audio adversarial examples. Extensive evaluations on diverse datasets consistently demonstrate that AIE outperforms existing methods, establishing its effectiveness in enhancing adversarial transferability across domains.

The processing and transportation of medical waste pose uncertain threats to the surrounding people and the environment in urban road networks. This paper aims to mitigate such risks under an emergency system with uncertain response times. In more detail, we first formulate an integrated pollution-population risk assessment that estimates the dynamic impact on the exposed population by embedding the emergency response time into the risk measure. Given the variability in traffic conditions, the response time is uncertain, which also affects the associated risks. Taking this randomness into consideration, a bi-objective chance-constrained model is developed to seek optimal facility locations, vehicle acquisitions, as well as route and tour plans, such that both the risk and cost are simultaneously minimized. To meet practical restrictions on medical waste collection, continuously accumulative vehicle load and volume constraints are added to the two-commodity flow formulation. Then, we propose a comprehensive solution procedure that integrates a Back Propagation Neural Network approach within the fuzzy chance constraint framework to address uncertainties. Two multi-objective methods, an augmented ɛ -constraint solution technique and a nearest-neighbor Non-Dominated Sorting Genetic Algorithm II (NSGA-II) algorithm, are implemented respectively for small- and large-scale problem instances. A series of numerical experiments are conducted on a real-life situation in Shanghai city of China to demonstrate the workability of the proposed model and approach. The numerical results show that our recommended system can effectively prevent the overall capacity shortage, reduce the total cost and risk respectively by more than 8% and 11%, as well as lower the transportation risk and distance respectively by nearly 15% and 23%. • Urban medical waste management is studied under uncertain emergency response time. • An integrated pollution-population measure is proposed to estimate dynamic risks. • Vehicle load and volume constraints are improved in two-commodity flow formulation. • BPNN is integrated with fuzzy chance constraints to model uncertainties. • A nearest-neighbor NSGA-II algorithm is proposed for solving large instances.

This paper introduces a novel objective quality metric for predicting the impact on Listening Quality (LQ) and Localisation Accuracy (LA) of interpolated Directional Room Impulse Responses (DRIRs). The DRIRs in this paper are First Order Ambisonic (FOA) representations of the RIR. Interpolation is used to synthesise a DRIR at a new listening position using known DRIRs at neighbouring listening positions. The target application is spatial audio synthesis for Virtual Reality (VR) that adapts to the position of the listener. The LQ and LA scores are based on the Pearson’s Correlation Coefficient (PCC), comparing segments of reference and synthesised DRIRs. These scores compare well with results from other representative objective measures of Ambisonic audio quality using DRIRs convolved with sound sources. A benefit of the proposed approach is predicting the impact on spatial audio quality when using interpolated DRIRs.

This paper proposes the co-prime circular microphone array (CPCMA), which applies co-prime array theory to circular microphone arrays. Compared with a conventional uniform circular array (UCA) with the same number of microphones and radius, the CPCMA achieves beampatterns with a narrower main lobe and fewer side lobes, whilst also avoiding spatial aliasing, thus having no grating lobes above the Nyquist frequency of the UCA. Array gain (AG) results indicate that the CPCMA is better than the UCA at amplifying the desired target signal while suppressing noise from other directions for typical speech signal frequencies. Compared with a UCA with similar performance and the same spatial Nyquist frequency, the CPCMA significantly reduces the required number of microphones. Simulations also indicate advantages of the CPCMA in speech source DOA estimation under high noise and low reverberation, especially when there are simultaneous multiple sources.