Orthogonal time-frequency space (OTFS) modulation has been shown to be robust in channels with high Doppler spreads. Conventional approach of OTFS signaling involves two steps, viz., conversion from delay-Doppler (DD) domain to time-frequency (TF) domain and then to time domain (TD) for transmission. A more direct approach converts the DD domain symbols to TD directly using the inverse Zak transform in one step, which is performance-wise better for large channel spreads. In this paper, we consider discrete Zak transform based OTFS (DZT-OTFS) and propose a deep learning based low-complexity channel estimation algorithm for fractional DD channels. The proposed approach learns the delay-Doppler matrix (DDM) through training rather than analytically computing it explicitly, and this drastically reduces complexity. A key novelty in the proposed approach is that learning is carried out in the TF domain for DD domain channel estimation. This is motivated by the observation that the values in the channel matrix in TF domain has a smaller swing compared to that in DD domain, which is more favorable for training. Simulation results show that the proposed TF learning based channel estimation achieves almost the same performance as that of a state-of-the-art algorithm in the literature but at a significantly lesser complexity, making the proposed approach practically appealing.


    Access

    Check access

    Check availability in my library

    Order at Subito €


    Export, share and cite



    Title :

    Delay-Doppler Channel Estimation in DZT-OTFS via Deep Learning in Time-Frequency Domain


    Contributors:


    Publication date :

    2024-06-24


    Size :

    1382231 byte





    Type of media :

    Conference paper


    Type of material :

    Electronic Resource


    Language :

    English