On the array configuration and accuracy of remote in-ear level sensing for in-vehicle noise control applications

Measuring in the ear of the operator of a vehicle may be useful in several applications such as noise assessment, adaptive communication, or active noise control. However, wearing in-ear microphones is not very comfortable. This work proposed a design methodology for an array-based-virtual-microphone for remote in-ear level sensing. To extend the sensing capabilities of current virtual microphone methods, a method that besides robust filter design also includes microphone placement optimization is proposed. Motivated by the recent progress in sparse signal processing, an unconstrained optimization is proposed which keeps the fidelity of in-ear level while reducing the number of initial active microphones through the log-sum-sparsity-induced regularization term. It has been compared to several screening methods which are optimizing for system diversity and are often used for optimal sensor placement. To study the importance of microphones being locally placed around the ears, several methods to maximize the system redundancy are also evaluated. The selection approach based on the sparsity coding theory achieves the best remote sensing performance with the least number of microphones. It results in the array configuration which balances well between the local (around the head) and global placement (throughout the vehicle roof). Sensitivity of the design to the natural head movements of 20 and 45 degrees is assessed, showing an improved system robustness by accounting for those variations in the design.