Equalization of optical nonlinear waveform distortion using SVM-based digital signal processing

We proposed a novel nonlinear equalization scheme using a support vector machine (SVM) for optical fiber communication systems. The scheme works even when the received symbols are completely distorted on a constellation map. Whereas conventional SVM-based nonlinear equalization is performed on a two-dimensional (2D) in-phase-quadrature (IQ)-plane, our proposed scheme uses the SVM in a higher dimensional signal space after tapped delay lines. By employing the proposed scheme, we can enlarge the separation between the distorted symbols. We investigated the basic performance using simulated optical fiber transmission of 10-GSymbol/s 16QAM signals. The receiver sensitivity was improved by 2 dB at a BER of 10(-5).