A Hybrid SFANC-FxNLMS Algorithm for Active Noise Control Based on Deep Learning

The selective fixed-filter active noise control (SFANC) method selecting the best pre-trained filters for various types of can achieve a fast response time. However, it may lead to large steady-state errors due inaccurate filter selection and lack adaptability. In comparison, filtered-X normalized least-mean-square (FxNLMS) algorithm obtain lower through adaptive optimization. Nonetheless, its slow convergence has detrimental effect on dynamic attenuation. Therefore, this paper proposes hybrid SFANC-FxNLMS approach overcome algorithm’s provide better reduction level than SFANC method. A lightweight one-dimensional convolutional neural network (1D CNN) is designed automatically select most suitable each frame primary noise. Meanwhile, FxNLMS continues update coefficients chosen at sampling rate. Owing effective combination two algorithms, experimental results show that rapid time, low error, high degree robustness.