“…Faced with the requirements of the ultra-low latency, high spectrum efficiency, high transmission rate and business diversity in the future networks, researchers and practitioners in related fields have exerted a tremendous fascination on a number of alternative single-carrier or multicarrier modulation techniques [31], including CP-Discrete Fourier Transform spread OFDM (CP-DFT-s-OFDM) [32], filter bank multicarrier (FBMC) [33][34][35][36][37][38][39], generalized frequency division multiplexing (GFDM) [40,41], universal filtered multicarrier (UFMC) [42][43][44], and filtered-orthogonal frequency division multiplexing (F-OFDM) [45][46][47], and so forth. Apart from typical waveform design methods where waveform parameters are obtained manually, with the development of machine learning, waveform design methods with data-driven models have attracted increasing attention [48,49]. These modulation waveforms have their own merits and drawbacks in 5G communication scenarios, respectively.…”