Hello, recently I am working on the design of microphone array beamforming using convex optimization.I designed the circular array beamforming with the sidelobe suppresion of 20db, 30db, 50db and 100db in the interference direction of 180 degre and 0degre. The result is evident in the beam pattern figure(the sidelobe suppression has reached all the expected db). However when I tested the beamforming on audio data, the maximum db suppression is 30db. Meaning when I set the db to 50 or 100 db, the suppression result is the same as for 30db. I need help to clarify this, maybe I missed some theories and concepts of microphone array beamforming design. Thank you for your reply
Hi,
There are many ways this can happen. Since I don't know the details, here are some possible reasons.
1) Array errors: My most general guess is that the audio data is recorded on an array that is not well calibrated. You can compute something called "the expected beampattern" given some errors in the arrays elements/microphones (see for instance van Trees' Optimum array processing), and it turns out it takes exceptionally low errors to reach 50 or 100 dB suppression on average in the beampattern nulls.
2) Model errors: Depending on where you are recording, you may have multipath (reflections from walls, the ground, etc.). If you are closer to it, maybe you have an extended source? In other words, the source is not covered by the sharp null you have placed.
Best,
Håvard
- Badges
- Science topic