An Improved Fast Independent Vector Analysis Algorithm for Enhanced Efficiency

ABSTRACT This paper proposes an improved fast independent vector analysis (FastIVA) algorithm. The proposed algorithm replaces the second‐order derivative with the original function and its first derivative, incorporating an adaptive scaling factor to effectively reduce the number of iterations and computation time. The experiments are conducted based on a hybrid dataset constructed from the PNL Non‐Speech Corpus and the ATR Japanese Speech Database. Experimental results demonstrate that the improved algorithm achieves separation performance comparable to the traditional FastIVA while significantly reducing both the iteration count and runtime.