Is it possible to perform acoustic beamforming with a 2 microphone ULA for real world speech ?

I'm trying to perform beamforming for a source standing at a specific angle, using a 2 microphone ULA (using a smartphone stereo mode recording) for real-world signals (varying frequencies). So far I've tried applying Delay-and-Sum, Frost, LCMV, MVDR, GSC, Subband-Phase-Shift beamforming algorithms available in phased array toolbox. However none has achieved any improvement at all.

So I want to ask this: Is it possible to achieve beamforming for real-world acoustic signal using the 2 microphone ULA as mentioned above ? If so, could you please suggest what algorithm/approach should I be looking for ?

Answer 1

[Honglei Chen]

I'm not an expert on microphone arrays but here is how I look at it.

The performance of a beamforming algorithm depends on the setting of your array. To begin with, you have only two elements so you have limited degrees of freedom to work with. Therefore, for most conventional beamforming algorithms, you would expect a rather wide beamwidth. If you have sources that are close to each other, then it probably won't help much.

On the other hand, there are also something called differential array for microphone arrays. Those arrays are often quite small so the conventional beamforming algorithms are not very effective. Instead of forming a beam toward a certain direction, the differential processing steers the null to a certain direction to cancel out interferences. Maybe that's what you want? For example, if you use a weights of [1;1], then you formed a wide beam toward the boresight. But if you use [1;-1] instead, then you formed a null toward that direction.

I hope this helps.