Title:Performance of Hierarchical Sparse Detectors for Massive MTC

Abstract:Recently, a new class of so-called \emph{hierarchical thresholding algorithms} was introduced to optimally exploit the sparsity structure in joint user activity and channel detection problems. In this paper, we take a closer look at the user detection performance of such algorithms under noise and relate its performance to the classical block correlation detector with orthogonal signatures. More specifically, we derive a lower bound for the diversity order which, under suitable choice of the signatures, equals that of the block correlation detector. Surprisingly, in specific parameter settings non-orthogonal pilots, i.e. pilots where (cyclically) shifted versions interfere with each other, outperform the block correlation detector. Altogether, we show that, in wide parameter regimes, the hierarchical thresholding detectors behave like the classical correlator with improved detection performance but operate with much less required pilot subcarriers. We provide mathematically rigorous and easy to handle formulas for numerical evaluations and system design. Finally, we evaluate our findings with numerical examples and show that, in a practical parameter setting, a classical pilot channel can accommodate up to three advanced pilot channels with the same performance.