Method for updating microphone configuration in audio super-resolution

Audio super-resolution is a technique by which a high-resolution signal is reconstructed from a low-resolution input. The low-resolution input can be a set of signals captured by multiple microphones at a low sampling rate. In such a case, the microphone configuration might affect the resultant reconstruction performance. This study is related to a method for updating the microphone configuration to keep it from becoming ill-conditioned in a super-resolution problem. We specifically examine the condition number of the design matrix and attempt to reduce it by optimally updating the configuration in an empirical but effective manner. To serve as an alternative to the condition number, a simple measure is defined: the sum of the inverse of the distances between microphones. Based on this measure, the microphone and direction to move are selected. Computer simulations show that the method reduces the worst condition number in many cases. Moreover, results show that when it is applied to signal processing of audio super-resolution, it can improve the estimation performance by 90.4% on average.