Binaural Beamforming for Hearing Aids

People with sensorineural hearing loss often rely on hearing aids, yet many still struggle to understand speech in everyday noisy environments. A major challenge is that advanced noise-reduction algorithms, although powerful in theory, tend to lose effectiveness when implemented on modern wearable devices—such as glasses equipped with microphones—especially when the user or surrounding sound sources are moving. In this talk, I will introduce a new approach designed to overcome these limitations. The method enhances traditional binaural processing by focusing on a “region of interest,” allowing the system to better preserve important spatial cues while suppressing multiple interfering sounds. Simulation results in challenging reverberant settings demonstrate significant gains, including more stable performance when sound directions are uncertain and better control over noise amplification. In particular, the new approach proves the robustness to be 2.5 times more stable than conventional methods. Ultimately, this work aims to bring hearing-aid technology closer to natural listening, enabling smoother integration between hardware and human perception and offering users clearer, more comfortable hearing in real-world situations.

M.Sc. student under the supervision of Prof. Israel Cohen.