ViterbiNet: A Deep Learning Based Viterbi Algorithm for Symbol Detection

Abstract

Symbol detection plays an important role in the implementation of digital receivers. In this work, we propose ViterbiNet, which is a data-driven symbol detector that does not require channel state information (CSI). ViterbiNet is obtained by integrating deep neural networks (DNNs) into the Viterbi algorithm. We identify the specific parts of the Viterbi algorithm that are channel-model-based, and design a DNN to implement only those computations, leaving the rest of the algorithm structure intact.We then propose a meta-learning based approach to train ViterbiNet online based on recent decisions, allowing the receiver to track dynamic channel conditions without requiring new training samples for every coherence block. Our numerical evaluations demonstrate that the performance of ViterbiNet, which is ignorant of the CSI, approaches that of the CSI-based Viterbi algorithm, and is capable of tracking time-varying channels without needing instantaneous CSI or additional training data. Moreover, unlike conventional Viterbi detection, ViterbiNet is robust to CSI uncertainty, and it can be reliably implemented in complex channel models with constrained computational burden. More broadly, our results demonstrate the conceptual benefit of designing communication systems to that integrate DNNs into established algorithms.Parts of this work were accepted for presentation in the

Hardware-Limited Task-Based Quantization

“…One can also envision an online trainable task-based quantizer, which is capable of further tuning its hyperparameters in real-time to track dynamic environments, as in, e.g., [40]. For example, a communication receiver using a deep task-based quantizer for symbol detection, can exploit a-priori knowledge of pilot sequences as labels corresponding to inputs acquired in real-time.…”

Deep Soft Interference Cancellation for MIMO Detection

“…We do not assume that the channel is linear nor that the receiver knows the conditional probability measure p Y |S (·|·). Following the approach of [36], we design our network to implement interference cancellation in a data-driven fashion. In particular, our proposed receiver is based on the iterative SIC algorithm proposed in [6].…”

“…Here, we present a receiver architecture which implements iterative SIC in a data-driven fashion. Following the approach of [36], we wish to keep the overall structure of the iterative SIC algorithm, depicted in Fig. 2, while replacing the channel-model-based computations with dedicated suitable DNNs.…”

“…In our previous work [36] we proposed ViterbiNet, which is a data-driven implementation of the Viterbi algorithm for detecting symbols transmitted over channels with finite memory.…”

“…The resulting sequential training method allows the receiver to learn its symbol detection mapping from a smaller number of samples. The ability to accurately train with a relatively small number of training samples facilitates its applicability for online training, making the sequential approach attractive for dynamic environments in which the receiver has to frequently retrain, possibly by exploiting the inherent redundancy induced by digital communication protocols [36].…”