In sound perception, humans adapt by tilting the head or relying on other location cues such as visual input when the source of the sound lies on the cone of confusion. On this cone, no time and level differences can be perceived by the ears. If that is the case, how do sensors or computers go about this problem when localizing the source of the sound? I understood that adding microphones to form an array to aid localizing might be a solution but are there any other approaches feasible?
The cone of confusion is based on a very simple approximation: a sphere with two microphones.
Real humans do have a fair bit of vertical localization ability using pinna cues, shoulder reflection, binaural asymmetries and dynamic cues: i.e. change in interaural differences for very small head rotations.
Computer models can operate the same way, depending on what signals are available. If you want to have vertical or front back resolution: don't use a sphere with two microphones. A real human head would be better, but much simpler would be a small microphone array that has spatial extension over the dimension you want to resolve.