The multichannel active noise control (MCANC) approach is widely regarded as an effective solution to achieve a large noise cancellation zone in a complicated acoustic environment. However, the sluggish convergence and massive computation of traditional adaptive multichannel active control algorithms typically impede the MCANC system's practical applications. The recently developed selective fixed-filter method offers a way to decrease the computational load in real-time scenarios and enhance the reaction time. Nevertheless, this method is specifically designed for the single-channel ANC system and only considers the frequency information of the noise. This inevitably impacts the effectiveness of reducing noise from various directions, particularly in the MCANC system. Therefore, we proposed a spatial-frequency-based selective fixed-filter ANC technique that adopts the Bhattacharyya Distance Matching (SFANC-BdM). In our work, the BdM is a one-step spectra and is designed by calculating similarity of different data distribution. According to the most similar case, the corresponding control filter is then selected. By avoiding separately extracting the direction and frequency information, the proposed method significantly increases the algorithm's efficiency. Compared to the conventional SFANC method, it enables a more accurate filter choice and achieves better noise reduction.

Utilizing the difference in phase and power spectrum between signals and noise, the estimation of direction of arrival (DOA) can be transferred to a spatial sample classification problem. The power ratio, namely signal-to-noise ratio (SNR), is highly required in most high-resolution beamforming methods so that high resolution and robustness are incompatible in a noisy background. Therefore, this paper proposes a Subspaces Deconvolution Vector (SDV) beamforming method to improve the robustness of a high-resolution DOA estimation. In a noisy environment, to handle the difficulty in separating signals from noise, we intend to initial beamforming value presets by incoherent eigenvalue in the frequency domain. The high resolution in the frequency domain guarantees the stability of the beamforming. By combining the robustness of conventional beamforming, the proposed method makes use of the subspace deconvolution vector to build a high-resolution beamforming process. The SDV method is aimed to obtain unitary frequency matrixes more stably and improve the accuracy of signal subspaces. The results of simulations and experiments show that when the input SNR is less than −27 dB, signals of decomposition differ unremarkably in the subspace while the SDV method can still obtain clear angles. In a marine background, this method works well in separating the noise and recruiting the characteristics of the signal into the DOA for subsequent processing.

The selective fixed-filter active noise control (SFANC) method has gained popularity due to its high computational efficiency and fast response. However, it solely accounts for noise frequency information, while neglecting its spatial information, which undoubtedly undermines the noise reduction performance, especially for direction-varied noises. To address this limitation, we proposed the Frequency-Direction Aware Multichannel SFANC (FD-MCSFANC) method, which incorporates both frequency and direction information of primary noises for accurate filter selection in the multichannel ANC (MCANC) system. In our approach, a Convolutional Neural Network (CNN) is designed to classify noises based on both frequency components and Direction-of-Arrival (DOA), with the combined classification results determining selected control filters. Furthermore, a joint loss function based on multi-task learning is utilized to implement end-to-end training of the CNN. Numerical simulations show that the FD-MCSFANC method effectively attenuates noises with different frequencies and incident angles. Moreover, it responds much faster than traditional adaptive algorithms and achieves better global noise control performance than the Multichannel SFANC (MCSFANC) method.

Selective fixed-filter active noise control (SFANC) has emerged as a prominent and practical solution for noise suppression due to its fast response, high computational efficiency, and stability. However, its noise reduction performance still degrades when dealing with some dynamic noise scenarios where the filter switching cannot catch up with the variation of the primary noise source. To tackle this challenge, we proposed a novel approach to employ the dynamic factor graph (DFG) in predicting the direction and frequency bands of primary noise. Once the SFANC algorithm acquires this joint frequency and direction information in advance, it can generate a more suitable control filter for the coming noise in a shorter time. By involving the Bhattacharyya distance Matrix (BdM) to simultaneously render spatial and frequency information of the primary noise, the proposed method further facilitates the precision of the filter selection for multichannel ANC systems. The BdM can also be adaptive to non-uniformed array. Moreover, experimental results demonstrate the stability of BdM spectra in detecting joint frequency and directional information and showcase the method's ability to accurately predict noise variation and respond promptly. Comparative analysis against other popular algorithms validates the superiority of our approach in dealing with dynamic and moving noise sources. Therefore, these theoretical investigations and numerical simulations show that this proposed method can potentially be applied to cope with the dynamic noise in real-world scenarios.