Perceptual Filter Optimization for Active Acoustic Equalization

The design of feed-forward filters for hearables with the goal to match a target transfer function from the outside to the inside of the hearable is a concern of applications such as occlusion cancellation or transparent hear-through. A common filter design method is based on the Wiener-Hopf equation, which gives an analytical solution to a least squares cost function. However, this solution strongly depends on the phase response of the target transfer function. We present a cost function and suggest a numerical optimization scheme for the design of a feed-forward filter that does not require a heuristic choice of the phase spectrum for a given target transfer function. This cost function considers the non-linear perception of amplitude and frequency of the human auditory system. We verify the effectiveness of the filter design method based on a set of real acoustic transfer functions of a noise cancelling headphone.