 Gain vs angle
|
An array of microphones can be used to sensitize pickup in one direction by suppressing pickup from a different direction. The sound coming from a given direction arrives in equal phase, and their sum produces a stronger signal than when the sound arrives off to the side. This produces a strong main lobe of gain, as shown in the gain / polar plot to the left. This technique is called beamforming, and is very useful when the direction of sound is likely to come from a fixed direction. If the direction is not fixed, and there is some way of determining the direction, then the main lobe can be steered. Microphone placement and phase equalization is critical, so its good to involve a DSP consulting service early in the design.
Sound that arrives significantly at an angle to the main lobe, arrives at the mics at different times, and therefore each signal is phase delayed. In this case, summing these signals will add to be a reduced magnitude, compared to adding two signals in phase. How much suppression depends on the frequency of the sound and the spacing of the mics. Applications include voice command systems in cars, where the driver is at a known position, and road noise comes from the side. Another application is a set top box where it is relatively true that the person giving commands is positioned in front of the box. |
 Wavefronts arriving at different angles have
different relative delays
 adding artificial delay to steer main lobe
|
Delay can be artificially added to the mic signals, by re-sampling at a higher rate and interpolation. This results in steering the main lobe from side to the other. It can also alter the width of the lobes. Because this method depends on frequency, then adding any delay will be frequency dependent, as will the spacing of the mics. Ideally mic spacing between two mics should be 1/2 wavelength apart. At 1000 Hz, this is 17.5 cm, or about 6.7 inches. The method of adding delay is called delay-sum beamforming.
Because of this sensitivity to mic spacing and frequency, there are advanced techniques to shift the problem into the frequency domain. The processing can take place in the frequency domain, or alternatively, the audio can be broken into sub-bands, and processed band by band. Multiple mics (more than 2) can improve the gain, and allow for asymmetrical solutions. DSPWidgets can provide DSP consulting on solutions for beamforming royalty free. Simple sum solutions are easy to construct and port, and also solutions with simple delay. We can provide consulting to show you how vary the delay and compute the sums of the delay. We can also provide consultation on the placement of the microphones. |