Multi-Channel Differencing Adaptive Noise Cancellation Based on Kernel-based Normalized Least-Mean-Square Algorithm

The performance of detection for weak signal of the sonar array is extremely attenuated by intricate noises. It is also difficulty how to acquire reference (or secondary) noise correlated with the primary noise. The multi-channel reference noises are obtained by differencing the accurately delayed outputs of array elements. Kernel-based normalized least mean square (KNLMS) algorithm is an efficient way used to online predict time series data with the advantages of simplicity and stability. This paper presents an algorithm that combines multi-channel differencing to obtain reference noise and KNLMS to adaptively cancel the unknown noise. The simulation results demonstrate that the performances of multi-channel differencing adaptive noise cancellation based on KNLMS are better than the conventional adaptive noise cancellation methods using the realistic lake experiment of sonar array noise data.