Microphone Channel Frequency Response Calibration Using Circular Shifting and Spatio-Temporal Prediction

Microphone channel frequency response calibration (MCFRC) can compensate for frequency response mismatches among channels and has attracted increasing attention in recent years, which is accompanied by booming applications of microphone arrays. State-of-the-art MCFRC techniques translate the frequency response calibration (FRC) into the optimization of calibration filter coefficients and have shown remarkable performance. However, a compromise between inter-microphone distance and passband bandwidth shrinkage still exists, which limits the wide application of those techniques to a great extent. To address this problem, an MCFRC method using circular shifting and spatio-temporal prediction is proposed in this article. First, all microphones from uncalibrated channels are deployed in a rotating symmetric arrangement, and circular shifting is conducted, making each channel experience all different predetermined positions. Then, for each channel, a spatio-temporal prediction technique is utilized to separately enhance those segments containing received calibration signals at different positions. Next, the summation of all enhanced segments for each channel is calculated, which can almost ensure that differences among microphone channel (MC) outputs are only caused by channel frequency response mismatches. Finally, the Newton algorithm is employed to design the calibration filter coefficients corresponding to each channel. Simulation results reveal that the proposed approach considerably outperforms existing calibration methods across various signal-to-noise ratio (SNR), reverberation time, and inter-microphone distance conditions. Meanwhile, the passband bandwidth shrinkage problem is well handled. Real-world experiments also verify its effectiveness.