Design of Low-Order IIR Filters Based on Hankel Nuclear Norm Regularization for Achieving Acoustic Transparency

Inhalt:
or the well-known occlusion effect, can be mitigated by altering the system response through a digitally filtered sound playback. In resource-constrained hearing devices such as hearables and hearing aids, an implementation using infinite impulse response (IIR) filters is often preferred over finite impulse response (FIR) filters. This work focuses on indirect design methods for IIR filters to achieve acoustic transparency, i.e., an FIR filter is designed and in a second step approximated by a low order IIR filter. Both steps are unaware of each other in conventional methods, in which case the approximation can greatly reduce the performance. To mitigate this effect, we combine both steps in a novel approximation-aware filter optimization algorithm based on the nuclear norm heuristic. Using real-world data, we conduct case studies for an acoustically transparent hear-through and a vibration sensor-based occlusion effect reduction system. The conducted simulations indicate a performance improvement in resource-constrained applications.