Signal constellation design for Gaussian and non-Gaussian channels

[摘要] Optimum signal constellation design is crucial to the design of high-performance digital communication systems. Variation in signal set can dramatically change the system performance. Our goal is to design constellations, yield high performance, and satisfy power and waveform constraints. In our investigation, we use the Kullback-Leibler distance as a system performance index to design optimal signal sets for additive Gaussian and non-Gaussian noise channels. We optimize this distance, and therefore detector performance, for M-component signal sets defined in an N-dimensional signal space while constraining signal waveform properties. We obtained optimal signal sets for several noise distributions with the power of the signal spanning across all SNR. Our approach can be generally extended to design signal sets for various channel environments.

[发布机构] Rice University

[效力级别] Electrical engineering