Colloquium - Details
You will receive information about presentations in time if you subscribe to the newsletter of the Colloquium Communications Technology.
Master-Vortrag: Optimized Receiver Systems Preserving Maximum Relevant Mutual Information
RocioMartin Lima
Freitag, 10. September 2021
11:00 Uhr
virtueller Konferenzraum
In modern communication systems the receiver-sided signal processing often results in bottlenecks. In particular, high-speed computations in channel decoders are demanding, but also crucial to achieve high data rates. Therefore, reduced complexity decoding algorithms for the widely used low-density parity-check (LDPC) codes are developed in this thesis. State-of-the-art LDPC decoders use costly mathematical operations and high-precision log-likelihood ratios in iterative message passing algorithms. Both typically results in high decoder complexity.
In contrast, this thesis proposes to design check node operations that decode directly using the bit-level representation of strongly quantized integer messages. Like this, small logical circuits that only use simple bit-level operations (e.g. XOR) on the passed messages can replace demanding check node operations. At the same time, only a small number of bits per message is required. An information bottleneck inspired design method is applied to learn suitable logical circuits for decoding. This method involves genetic optimization algorithms.
