Technology Study On Cross-Coupled Multi-Channel Adaptive Active Noise Cancellation

Submarine noise is the main interference of passive sonar system, which decreases the SNR at the input of the passive sonar, resulting in the deduction of sonar ranging. Therefore, reduction of submarine noise is of great importance. The reduction of submarine noise is a systematic project, requiring enormous financial support as well as series of techniques which have not yet been resolved fundamentally up till now so that researches on the techniques of adaptive active submarine noise canceling is meaningful and valuable for practical application. Applying the acoustical principles, our research subject is to find out a new adaptive active noise canceling technique, which will retain the signal energy during the noise canceling process. This technique is creative and much more difficult compared with conventional active adaptive noise canceling.AANC(Adaptive Active Noise Cancellation) works on the principle of destructive interference between the sound fields generated by the original "primary" sound source and that due to other "secondary" sources, whose acoustic outputs can be controlled by controllers, which are usually adaptive filters. The key is the configuration of adaptive filters and the corresponding algorithm in AANC systems.It's great for the effect of the signal components in the reference input on the performance of the AANC systems. In this paper, the effect of the signal components combined with the reference noises on the performance of the AANC systems has been studied comprehensively. Mathematical expressions of the signal-to-noise density ratio, signal distortion and noise power spectrum at the adaptive active noise canceller output have been derived under the condition of no constraints for the adaptive solution. By introducing Modular Squared Coherence(MSC) function to the AANC systems, the method for evaluating the processing gain of the AANC systems is proposed in the paper.The line noise is a primary low frequency component of self-noise in submarines. In this paper, a hybrid active periodic noise canceller based on Chebyshev Polynomials has been introduced, and it has been compared with the direct form and the parallel form. Using chebyshev polynomials assures reliable reference inputs to a certain extent. The frequency difference between any two successive sinusoidal components is effectively increased in the hybrid form, as compared to the direct implementation technique. The rate of convergence of each adaptive filter can be significantly improved. In this paper, an AANC using a pulse train has been introduced to reduce the line interference. It has been shown that the total computational cost decreases if the AANC is combined with the Delayed-X LMS algorithm, and the excellent convergence performance has been obtained.The continuous power spectrum noise is wideband noise. First, the multi-channel filtered-x recursive-least-squares(MFXRLS) algorithms are introduced in time domain in this paper. Secondly, the frequency domain multi-channel filtered-X recursive least squares based on linear convolution(MLFFXRLS) algorithm has been proposed. It can be seen from the theoretical analysis and computer simulations that the MLFFXRLS algorithm will be have a much lower computational load than the MFXRLS algorithm, and the MLFFXRLS algorithm will be have a lower computational load than the MFXLMS algorithm for large channels numbers. In this paper, a delayless subband multi-channel system for active broadband self-noise canceling is also introduced, which has computational and convergence speed advantages comparable to a conventional subband design but without introducing any delay in the cancellation path.Finally, we analyze the real data using the proposed algorithms in the paper. In the case of uncorrelated signal and noise, whose bandwidths are both 2kHz, it can be shown from the processing that the adaptive processing gain of 21 dB was obtained, applying the MLFFXRLS algorithm, under the conditions of 2 control actuators, 2 error sensors and the perfect transfer functions of the primary paths and secondary paths to a certain extent if SNRis lower than -45dB in the reference input. The results validate the effectiveness of the proposed algorithms in the practice.