Targeted beamforming active noise control based on disturbance metrics

This work proposes a headphone Active Noise Control (ANC) scheme based on the targeted attenuation of sources that are deemed disturbing for listeners. Initially, a listening test determines the disturbance of distinct noises that coexist within background noise based on their class and their direction of arrival. The listening test is based on such predetermined spatial scenes, binaurally auralized and presented via headphones. The derived disturbance metric, guides the operation of the proposed headphone ANC: any complex auditory scene is subsequently analyzed and via a Sound Event Localization and Detection Neural Network, a beamformer is steered to the source deemed to be the most disturbing. A Time-Domain Beamformer, driven by a phased array formed by the two already existing reference microphones commonly found in the outer shell of ANC-enabled headphones and guides the multi-reference ANC controller in order to provide an improved attenuation of the primary disturbing source, while also significantly attenuating the background noise field to acceptable levels.