Starting from obvious similarities in operations in LDPC decoding to common neuron models, we have started making this correspondence more obvious. There are manifold possibilities to bridge the two fields, where some applications like denoising offer very obvious links, others appear much more distant.
Posts Categorized / Research Interests
In contrast to the the well-known separation theorem, in the non-asymptotic regime, joint source-channel coding can have advantages. We design multi-edge-type LDPC codes for different source statistics.
As an alternative to Diffie Hellman’s key exchange protocol, physical layer key generation is discussed based on measurements of reciprocal channels. Due to statistically independent noise together with vector quantization, the keys might not be the same. Hence, the contribution discussed key reconciliation possibilities, either introducing guard bands or LDPC Slepian‐Wolf coding. We derive a
We investigate nonstationary disturbances in wireline and wireless systems. In wireline, we make us of the common-mode signal to cancel impulse noise from the differential mode or use the common mode to forecast error positions. In wireless, we model impulse noise on antenna arrays caused by ignition, electrical machines, MRI equipment, and alike. Furthermore, we
Work in xDSL focuses on problems of multicarrier transmission, like PAR reduction, time-domain equalization, bit-allocation. To name just a few outcomes of this research: First real capacity optimizing time-domain equalizer PAR reduction scheme which is widely used (many million installations) A bit-allocation scheme allowing for arbitrary SNR margins between priority
We designed UEP Turbo and LDPC codes based on puncturing and pruning. In Turbo codes, pruning and puncturing with time-varying patterns allow for easy variation of the rate and gain. UEP LDPC codes were designed based on a irregular check profile. The applied pruning procedure allowed rate and gain adaptation according to given requirements. UEP
We found that the iterative decoding of analog concatenated codes may be described as iterative projections realizing a least-mean-square solution. This allows for a very intuitive understanding of the iterative Turbo-like decoding procedures. Furthermore, we investigate the correction of bursty noise in additional background noise.
Network codes as a means of reducing the usage of links in a multicast scenario ask for different protection, since error propagation to the end node depends on error locations. Additionally, user data is of different importance and sensitivity. Both ask for suitable coding schemes combining aspects of rate-less and network codes. Different QoS
Our MIMO research focuses on the treatment of imperfect channel information and a possible realization of UEP properties just as in the case of multicarrier transmission. We consider multi-user bit loading, multi-user hierarchical modulation, LDPC coded hierarchical modulation.
Reserving spatial dimensions proved to be very efficient for PAR reduction. With a high numbers of transmit antennas, the corresponding rate loss can be neglected.
We investigate the properties of acoustical wave guides, especially water pipes. The goal will be low-rate data communications for control purposes.
In a joint project with life sciences, we investigate the communications and code properties of the DNA. Our share deals with reasoning of the codon/aminoacid encoding structure, the error-protection, and source coding properties, whereas Prof. Mushelishvili’s group concentrates more on the influences of the three dimensional orientation leading to analog control mechanisms in gene