Klaus Hueske

An adaptive and complexity reduced decoding algorithm for convolutional codes and its application to digital broadcasting systems

Convolutional codes are still prevalent in digital broadcasting systems, like DVB-T, DVB-H, DAB, DMB and DRM. This paper presents an adaptive and complexity reduced decoding algorithm for CCs. The proposed algorithm is based on the earlier proposed block syndrome decoder (BSD) and features a significant reduction of decoding effort depending on the current channel conditions. Additionally it is capable of adapting its output bit error rate to a given threshold to achieve a further reduction of decoding complexity. For this purpose, an SNR estimation is incorporated to achieve an adaption of decoding performance to a given bit error rate threshold in dynamic SNR environments. This is especially useful for the widely used serially concatenated coding schemes, which is demonstrated by means of the terrestrial DVB system.

Turbo Equalization for Receivers with Unreliable Buffer Memory

In this paper the effect of unreliable buffer memory on Turbo equalization is investigated. Under the assumption that the resulting bit errors are uniformly and independently distributed on receive and LLR buffer output, the effect of unreliable memory on the channel capacity is analyzed. It is further shown, that extrinsic information transfer and bit error rate (BER) performance of a conventional Turbo Equalizer are heavily degraded. Based on these observations a fault tolerant (FT) Turbo Equalizer is derived, which considers the memory error characteristics by using a modified transition metric for the involved MAP equalizer and decoder. It is demonstrated, that the FT Turbo Equalizer can effectively compensate memory errors and thus yields a significantly improved BER performance.

Adaptive low complexity MAP decoding for turbo equalization

Turbo equalization is a powerful method to iteratively detect and decode convolutionally encoded data that is corrupted by inter symbol interference (ISI) and Gaussian noise. It is based on the exchange of reliability information between the equalizer and the decoder, which is typically some sort of maximum a posteriori (MAP) decoder. While the number of remaining errors in the received sequence decreases during the iteration process, the computational effort for decoding remains unchanged in each iteration. In this paper a syndrome based MAP decoder is proposed, that is capable of reducing the computational decoding effort during the iteration process without significantly influencing the convergence behavior.

Syndrome based block decoding of convolutional codes

A block processing approach for decoding of convolutional codes is proposed. The approach is based on the fact that it is possible for Scarce-State-Transition decoding and syndrome decoding to determine the probability of a certain trellis state before the actual decoding happens. This allows the separation of the received sequence into independant blocks with known initial and final states, thus making overlapping or modifications of the encoder or the information stream unnecessary. The proposed scheme offers potentials for both parallelization and reduction of power consumption.

Hard decision based low SNR early termination for LTE Turbo decoding

In this paper a syndrome based Low SNR early termination (Low SNR ET) scheme for Turbo decoding is presented. The scheme is based only on hard decision binary computations and is therefore easy to realize. While Low SNR ET is a general principle that can be applied in various scenarios, the focus of this work is on its application in the Long Term Evolution (LTE) system. It is shown that Low SNR ET is in particular very effective for typical LTE scenarios and that the proposed scheme can reduce the decoding iterations by about one third.

Ov-OFDM: A reduced PAPR and cyclic prefix free multicarrier transmission system

Orthogonal Frequency Division Multiplexing (OFDM) enables high data rate transmissions over frequency selective fading channels. Its beneficial qualities are a low implementation complexity and the possible application of power loading schemes to increase data throughput. Drawbacks of OFDM are the requirement for a cyclic prefix (CP) to avoid interference between OFDM symbols and the high peak to average power ratio (PAPR). The CP causes transmission overhead and hence reduces throughput, while a high PAPR requires the use of less efficient linear amplifiers. The CP overhead can be reduced by increasing the OFDM symbol length, which will, however, increase processing delay and PAPR. In conventional OFDM the symbol length directly determines PAPR and transmission overhead. In this paper we use a CP free OFDM transmission system that allows shortening the OFDM symbol length to reduce the maximum PAPR without transmission overhead trade-off. The proposed system, called Ov-OFDM, is based on overlapping MMSE frequency domain equalization to remove interference between OFDM symbols.

Optimized channel coding schemes for a guard period free transmission system

Guard periods are widely used to avoid inter block interference in wireless communication systems. The independent processing of each data block enables computational efficient data estimation, however, the necessary insertion of guard periods produces a transmission overhead and reduces the throughput of the respective transmission system. By using overlapping techniques a block-wise data estimation can be realized without using guard periods. With this modification, either the throughput of the system can be increased or the code rate of the forward error correction code can be reduced. Based on the latter option we will propose two optimized channel coding schemes for the overlapping based system. The first uses a reduced code rate while the second exploits the characteristic error distribution of the overlapping based data estimator. The bit error performance of the resulting transmission systems is finally compared to a common cyclic prefix based transmission system.

Analysis of a reduced complexity generalized Minimum Mean Square Error detector

The generalized Minimum Mean Squared Error (GMMSE) detector has a bit error rate performance, which is similar to the MMSE detector. The advantage of the GMMSE detector is, that it does not require the knowledge of the noise power. However, the computational complexity of the GMMSE detector is significantly higher than the computational complexity of the MMSE detector. Using the circular approximation of the Toeplitz structured system matrix an approximate GMMSE detector can be derived, whose computational complexity is only slightly higher than MMSE, i.e. only an iterative gradient descent algorithm based on the inversion of diagonal matrices is additionally required. The performance of this approximate GMMSE detector is analysed by investigating the noise enhancement due to the conditioning of the system matrix (Toeplitz) and its circular approximation. Simulation results for fading multipath channels are given and the computational complexity is briefly discussed.

Parallel Block Signal Processing in High Speed Wireless Communication Systems

Block transmission systems are used for high data rate transmissions over wireless channels. The block structure can be obtained by inserting guard periods, which are used to avoid inter block interference between the transmitted data blocks. However, these guard periods produce a transmission overhead and therefore reduce the throughput of the transmission system. By using so-called overlapping techniques, a block processing structure can be realized without inserting guard periods. In this paper we will use these techniques for two major processing tasks, the data estimation and the error correction Viterbi decoding. By avoiding the overhead in terms of guard periods and initialization/termination sequences for con- volutional coding, the throughput of the system can be increased significantly, while the bit error performance is only slightly decreased. Furthermore the overlapping approaches enable a block-wise parallel implementation of the receivers processing tasks. Depending on the used system parameters (block length, interleaving length, decoding block length), different levels of parallelization can be achieved.

Improving the Performance of a Recurrent Neural Network Convolutional Decoder

The decoding of convolutional error correction codes can be described as combinatorial optimization problem. Normally the decoding process is realized using the Viterbi Decoder, but also artificial neural networks can be used. In this paper optimizations for an existing decoding method based on an unsupervised recurrent neural network (RNN) are investigated. The optimization criteria are given by the decoding performance in terms of bit error rate (BER) and the computational decoding complexity in terms of required iterations of the optimization network. To reduce the number of iterations and to improve the decoding performance, several design parameters, like shape of the activation function and level of self-feedback of the neurons are investigated. Furthermore the initialization of the network, the use of parallel decoders and different simulated annealing techniques are discussed.