K-means sequence extraction for high-dimensional QPSK constellations in a sub-THz photonics-aided 4600-m wireless system

Quadrature Phase Shift Keying (QPSK) has become one of the most widely used modulation techniques in modern communication systems due to its simplicity, robustness, and high spectral efficiency. However, as communication systems scale in complexity, the limitations of traditional QPSK in terms of spectral efficiency and noise resilience have become evident, particularly in high-speed communication environments. This paper proposes a novel time-domain multidimensional QPSK modulation scheme based on K-means sequence extraction to address these limitations. By extending the signal space in the time domain and using the K-means algorithm to extract a subset of constellation points in a multidimensional space, this method increases the Euclidean distance between constellation points, enhancing noise resilience. By combining with polarization multiplexing technology, true polarization-time-domain multidimensional modulation is achieved. It overcomes the limitations of conventional QPSK signals in probabilistic shaping. Unlike probabilistic shaping (PS) and geometric shaping (GS), our approach does not require modifications to the existing digital signal processing (DSP) architecture; it only requires the addition of sequence modulation and demodulation modules. Experimental validation conducted in a 125 GHz photonic-assisted 2 × 2 MIMO 4600-meter wireless system demonstrates that, at the same net data rate, the 100 Gbps KSE-QPSK signal achieves an optical power gain of 0.9 dB compared to conventional QPSK and maintains a bit error rate below the HD-FEC threshold of 4.7 × 10^-3 under high optical power conditions. This scheme shows great potential for high-speed and high-reliability communication systems, including next-generation 5G/6G wireless networks, optical communication systems, and other high-capacity communication applications.