Low-complexity, robust algorithm for sensor anomaly detection and self-calibration of microphone arrays

This study details a simple method for the detection of sensor degradation in microphone arrays and equalisation of the microphone gains. The described method is online in nature and based upon the long-term power spectra of the microphone signals of each array. The close proximity of the microphones within an array is essential for the application of the proposed detection and calibration method. The approach is tested on three sample degradation scenarios: complete sensor failure, non-linearity in the input and parasitic harmonic injection. While these scenarios are by no means exhaustive, they represent the most typical degradations in many applications. The results indicate that the developed approach is capable of detecting such failures. Further, the calibration factors computed using the proposed approach are similar to that obtained by established methods.