Microphone array

A microphone array is a setup in which several microphones are placed side by side at set positions and their signals are processed together. If you compare the time differences in when a sound reaches each microphone, you can tell which direction it came from. When the signals gathered this way are combined, you can get less noise and a clearer voice than from a single microphone.

This setup developed to hear voices from far away well in things like meeting room speakers and voice assistant devices. In small devices where it is hard to make a microphone bigger, the need grew to get good performance by combining several of them. Recently, as signals are processed with AI, more research is aiming for an effect close to an array with fewer microphones.

When each microphone receives the same sound, the arrival time differs a little because of the difference in distance. If you calculate this difference, you can estimate which direction the sound comes from.

The method gives the signals time delays and weights so that sound from the direction you want is made stronger and sound from other directions is made weaker. This creates the effect of focusing on the side where the speaker is.

It is mainly used in the ceiling microphones of video meeting equipment, in smart speakers and in voice recognition in cars. Even in noisy spaces, it can follow the speaker's position and pick up the voice.

To apply it, the spacing and placement of the microphones must fit the features of the target sound, and the difference in sensitivity between microphones must be corrected. The size of the device and its power budget also have to be considered.

Putting on more microphones does not always give better sound. If the placement is bad or the calibration is poor, noise can actually get louder and the direction estimate can be wrong. Echoes and reflected sound in a room also lower performance.

A common misunderstanding is to judge performance from the number of microphones alone. Even with the same number, the results differ a lot with the signal processing method and the chip's performance, so you should also look at test results from the real environment where it will be used.

When several people talk in one room, having two ears lets you guess which side a voice comes from. With several microphones, the same principle lets a machine find the position of a sound more exactly.

Technologies that must take in speech accurately, such as speech recognition, video meetings and smart speakers, all rely on this setup. If you know that more microphones are not always better and that placement and processing matter too, you can understand product descriptions more accurately.