I also need this feature!
It’s called beam-forming, it’s really great for noisy environments like offices.
An possible workflow would be: [waiting for voice trigger on without beamforming] -> [detects a Voice Trigger (Hotword)] -> [set the beaforming direction to voice]
Or to be simpler, when It starts do detect voice it change the beaforming direction to the speaker, people talking from other directions will get ignored.