Abstract: This paper proposes an algorithm for suppressing the multiple narrowband interferences using kernel independent component analysis (KICA) in direct sequence spread spectrum systems. Combining with the kernel methods the algorithm uses canonical correlations in a reproducing kernel Hilbert space (RKHS) as the contrast function. By initializing the demixing matrix using the pseudocode of the expected user, the algorithm realizes the separation of source signals and at the same time the expected user is obtained. As the restrictions are mutually independent of the spread spectrum signal and interferences, which can be easily satisfied in practice, the algorithm is useful for suppressing various NBIs. The simulation results verify the effectiveness of the algorithm.

Key words: narrowband interference, kernel methods, independent component analysis, canonical correlation