Beamforming Using Virtual Microphones for Hearing Aid Applications

This paper presents a novel, low-complexity method for synthesizing virtual microphone (VM) signals to enhance beamforming in hearing aid (HA) applications. Leveraging the W-disjoint orthogonality (WDO) assumption, the proposed approach generates additional VM channels from a typical two-microphone setup, potentially eliminating the need for extra hardware for physical microphone channels. The VM signals are estimated by modeling their relative transfer functions (RTFs) as a power function of the RTF between the physical microphones, enabling both interpolation and extrapolation of VM positions. We demonstrate the integration of the synthesized VM signals into a minimum variance distortionless response (MVDR) beamformer using realistic HA recordings across diverse acoustic scenarios. Experimental results show substantial improvements in signal-to-noise ratio (SNR) and objective speech intelligibility, with further gains achieved by optimizing VM position(s). Compared to existing VM methods, the proposed technique offers superior performance and scalability while maintaining low computational complexity, making it highly suitable for resource-constrained devices such as HAs.