Microphone Calibration for Multi-Channel Signal Processing

The application of microphone arrays and adaptive beamforming techniques promises significant improvements compared to systems operating with a single microphone as the spatial properties of the sound field are exploited. However, for real-world applications beamformers imply the risk of severe signal degradation due to mismatched microphones. Microphone mismatch naturally arises from production tolerances as well as from aging effects in the long run. The influence of microphone mismatch on beamforming techniques is barely addressed in literature. Mostly microphone mismatch is simply considered equivalent to deviations in the steering direction of a beamformer. In fact both of these imperfections may cause the well-known signal cancellation effect which usually comes along with a growth of the magnitudes of the filter coefficients. However, there are major differences as it will be pointed out in this chapter.

This chapter gives an overview on calibration techniques with fixed as well as with adaptive filters. After a short introduction to beamforming techniques we will point out the problem of mismatched microphones by theoretical investigations under the assumption of a diffuse noise field. In the systematic investigations of adaptive beamformers an unknown characteristic was discovered which could be denoted as "pattern cancellation effect". In a third section different calibration techniques are introduced. These techniques are evaluated under more practical conditions by conducting Monte Carlo simulations on the basis of measured microphone characteristics. Here, the limits of the different calibration techniques are pointed out. In the following section, we will focus on systems which calibrate the microphones automatically. A class of efficient techniques for self-calibration is presented and compared to existing methods. Such self-calibration methods perform a calibration adaptively in the background during normal operation of the system and therefore save the need for an additional costly calibration procedure. The performance of the calibration techniques is examined using practical real-world scenarios.