[Back]

S. Schwarz, M. Rupp:
"Predictive Quantization on the Stiefel Manifold";
IEEE Signal Processing Letters, 22 (2015), 2; 234 - 238.

In this letter, we consider time-varying complex-valued $n \times m$ matrices ${\bf H}[k]$ ($m \leq n$) and proposea predictive quantizer for the eigenvectors of the Gramian ${\bf H}[k]{\bf H}{[k]^{\rm H}}$, whichoperates on the associated compact Stiefel manifold. The proposed quantizerexploits the temporal correlation of the source signal to provide high-fidelityrepresentations with significantly reduced quantization codebook size comparedto memoryless schemes. We apply the quantizer to channel state informationquantization for limited feedback based multi-user MIMO, employing regularizedblock-diagonalization precoding. We demonstrate significant rate gains comparedto block-diagonalization precoding using Grassmannian predictive feedback.

Adaptive quantization, Grassmann manifold, limited feedback, multi-user MIMO, Stiefel manifold

http://dx.doi.org/10.1109/LSP.2014.2354258

Project Head Markus Rupp:
Mobile Access Research 2015