Robustness analysis of adaptive null-forming scheme for speech enhancement

The robustness of adaptive null-forming scheme with two microphones for speech enhancement was evaluated. The analysis reveals that the algorithm is sensitive to errors in the assumed signal model when the Signal-to-InterferenceRatio is high. However, when the Signal-to-Interference Ratio is low, the algorithm is robust to the phase mismatch between the two microphones. The magnitude mismatch decreases the amount of interference reduction, while causes little target distortion. Compared with the mismatch, the noise level of microphones affects the performance of the algorithm less. Constrain of filter weight or calibration of the microphones will improve the robustness of the adaptive null-forming algorithm.