What If a Spatial Filter Had Depth?

Beamforming is one of our fundamental tools for spatial filtering. Electronically steer an array in a direction, and signals arriving from that direction pass while signals arriving from other directions are attenuated. But in photographic terms, the resulting spatial filter has essentially “infinite depth of field”: signals along the same “look direction” are not distinguished by range.

The “flower” and the “mountain” behind it both remain in focus.

What if we could focus sound more like a camera, adding depth to direction so that we can select a particular location in 3D space?

This talk introduces acoustic focusing, a physics-based spatial-filtering approach. Rather than treating multipath propagation simply as an impairment, the propagation environment becomes part of the spatial model - and a source of additional spatial information.

The result is an acoustic analogue of photographic focus and depth of field. We can focus on a subject at a particular location, blur the acoustic foreground and background, change the depth of field, and even refocus the recording on a different subject after it has been made.

We’ll connect this photographic model to familiar audio DSP concepts including beamforming, free-space propagation, the acoustic wave equation, and information theory.

Real-world examples will demonstrate some initially counterintuitive behavior: focusing beyond a much louder intervening source, rejecting sounds beyond the subject, and changing which acoustic subject is in focus after the recording has been made.