A GMM-based non-uniform quantization scheme for improving low-resolution IMDD-UFMC system performance

Abstract

In this paper, a Gaussian mixture model (GMM) based nonuniform quantization scheme with low-bit resolution is firstly presented in optical intensity-modulated direct detection universal filter multicarrier (IMDD-UFMC) system for increasing system performance. By utilizing the Expectation-Maximum (EM) iterative algorithm of GMM model, the optimal quantized level can be directly obtained, and then is used to realize the nonuniform quantization for reducing the digital-to-analog converter (DAC) bit resolution of UFMC signals. In this processing, the extra estimation operation of signals’ probability density distribution (PDF) is not involved, which can reduce the complexity of the common non-uniform quantization scheme compared to the method based on the PDF estimation. And, the proposed scheme is further validated in a 10 Gb/s IMDD-UFMC transmission system through a 20 km fiber based the commercially OptiSystem15.0 software. And, all the results are obtained by this numerical simulation. Results verified that, in the cases of 3 ∼ 5 bit, 2.7%∼7.8% the error vector magnitude (EVM) and at least 1-dB received sensitivity improvements can be achieved compared to the uniform quantization schemes.