Data Detection for Oversampled OTSM System with Low-bit ADCs

Orthogonal time sequency multiplexing (OTSM) has emerged as a promising modulation scheme for wireless communication in high-mobility scenarios. It achieves a bit error ratio (BER) performance comparable to orthogonal time frequency space (OTFS) modulation at a lower complexity, as the Walsh-Hadamard transform (WHT) avoids multiplication operations. However, existing system models typically assume ideal infinite-precision analog-to-digital converters (ADCs), which entail high power consumption and complex hardware implementation. Therefore, we investigate an oversampled OTSM system with low-bit ADCs to achieve an attractive trade-off between data detection performance and power consumption. We first derive the input-output relationship of the new system under oversampling and analyze the sparsity of the channel matrix in the time domain. Next, we introduce a three-layer generative stochastic network (3L-GSN) structure to reformulate the data detection problem. The first two layers are dedicated to implementing noise whitening operation for solving the noise correlation problem caused by oversampling, while the last layer characterizes the low-bit quantization in a more precise probabilistic manner. To solve the data detection problem within the 3L-GSN structure, we propose a data detector based on the modified time-domain multi-layer vector approximate message passing (MTD-ML-VAMP) algorithm, leveraging the discrete characteristic of the OTSM signal to improve detection performance. Our simulation results demonstrate the superior BER performance of our proposed data detector compared to traditional detectors.