Search

Viterbi decoders are widely employed along with convolutional encoders to provide an excellent error correction probability in digital wireless transmissions. To achieve lower error rates, the constraint length (k) of the encoder, has to be defined at higher values. However, higher (k) results in increasingly complex implementations of the Viterbi algorithm. As a result, the power consumption of the Viterbi decoder will increase exponentially which is to the disadvantage of many wireless communication devices. Many reduced complexity decoding techniques of the Viterbi decoder presented in the past focused on algorithm and architecture specific levels. Most of the studies concentrated on the Add-Compare Select Unit (ACSU) of the decoder, due to its nature of repetitive processing; also add to the complexity of the process. This paper presents a review of several variations of the Viterbi algorithm performed in the ACSU. The comparison was made among several different algorithms with the most optimal algorithm presented in this paper. A combined algorithm approach is proposed at the end of the review for future improvement.

Cross-entropy (CE) based stopping criteria is known for its early termination capability of the turbo decoding iterations without scarifying its performance, in terms of bit error rate (BER). This advantage, however, performs well in high signal-to-noise ratio (SNR) albeit requiring infinite iterations at low SNR, thereby resulting in more unwanted computations and decoding delays. In this paper, a simple turning point (TP) measurement from the BER and average iteration number (AIN) graphs of the Genie stopping criterion has been introduced. The motivation is twofold; to terminate the iteration at the early stage in low SNR and maintaining the same level of performance of the existing method in high SNR, by using the proposed TP measurement. From the simulation results, the enhanced CE-based stopping criteria improve the early termination in low SNR and maintain the BER performance for various frame sizes compared to existing stopping criteria.

Most of the turbo decoding stopping criteria assume perfect channel reliability available at the receiver since they require a threshold based on signal-to-noise ratio (SNR) information. However, operational environments can vary according to frame sizes, code structures and channel reliability requirements, further aggravating the difficulty in threshold determination and convergence or non-convergence detection. This paper aims to provide a robustness analysis of improved minimum descriptive length (IMDL) stopping criterion peculiar to these factors: the frame sizes, the channel reliability requirements, the code structures, and the code rate. The robustness analysis is benchmarked to the well-known Genie stopping criterion based on the average iteration number (AIN) and bit error rate (BER) performance.

Measurement of reliability (MOR) stopping criterion is able to terminate early in the low and high signal-to-noise ratio (SNR) while maintaining the bit error rate (BER) performance. However, the performance of MOR is only based on one code structure and hence, the robustness of MOR is still unknown in turbo iterative decoding. Thus, this paper will test the robustness of MOR based on the following parameters: frame size, code structure, channel reliability and code rate. Then, we analyse and compare the average iteration number (AIN) and the BER performance of MOR with the benchmark stopping criterion known as Genie to determine the robustness of MOR. From the analysis, MOR has a BER degradation for low code rate. MOR also fails to perform well if the corret channel reliability is not available at the receiver and this results a large degradation in BER performance. However, MOR has close performance to Genie in terms of BER for various frame sizes, code structures and high code rate with the assistance of correct channel reliability. MOR is also able to save AIN at low SNR as compared to Genie and this can reduce delay and complexity of turbo codes.

The Cross-Entropy-based (CE-based) stopping criteria advantage is able to terminate early in the high signal-to-noise ratio (SNR) while maintain the bit error rate (BER) performance. Unfortunately, the criteria fails to cope with low SNR region and make the decoder iterates until maximum or infinite iteration. This paper proposed an early termination technique at low SNR for the CE-based stopping criteria using the decoding threshold derived by the measurement of reliability (MOR) at low SNR. In the simulation results and analysis, we compare the average iteration number (AIN) and the bit-error rate (BER) performance between the proposed combination methods with the existing CE-based stopping criteria. From the results, the combination method capable to reduce the AIN at low SNR with minimum one AIN while maintaining the AIN at high SNR as the traditional method. This significant reduction could reduce delay and complexity of existing CE-based stopping criteria while maintaining the BER performances.

The invention of turbo codes has attracted many researchers to explore various fields regarding turbo codes since it provides better error rate performance compared to the existing codes. Good error rate performance gives a penalty to the complexity of the codes. It includes the complexity of decoding algorithm and iterative decoding. This paper reviews the history of turbo codes and its structures. This paper also discusses the turbo decoding stopping criteria algorithm and analyses the performance of fixed and cross-entropy (CE) based stopping criteria. From the results, both criteria fail to terminate early in low SNR. However, CE-based stopping criteria outperform the fixed stopping criterion at high SNR and able to save more iteration and delay. This leads to an energy saving preservation while maintaining the performance of turbo codes.

Space-Time Trellis Code (STTC) can achieve both the diversity and coding gains. To maximize the advantages of STTC, two design criteria for slow Rayleigh fading channels will be used: i.e. the rank and determinant criteria. This paper focuses on the determinant criteria, which involves the evaluation of the generator matrix G. Evaluation is improved by pruning the search process earlier, which is made possible by estimating the initial upper bound prior to the search. In order to reduce the search complexity, the initial upper bound will be calculated at the minimal cycle. Comparatively, it can reduce the search space by 25.8%.

With the rapid growth of multicore processors, memory optimisation is necessary in improving the usage of cache memories. This includes efforts in improving the performance of data fetching from memory. Current prefetching algorithms depend on usage, such that items, which are not frequently used, will be removed. This can cause potential delay when infrequent items are removed, even when they are needed by many other processors in the near future. To alleviate this limitation, the hierarchical analytical scheduling (HAS) model is proposed in this paper. The model works by determining the relative importance of data during cache replacement policy. HAS is based on the concept of hierarchical temporal memory (HTM), in which the scheduling is derived on the priority and similarity of data from the aspect of space and time. Implemented with OCTAVE, a simulation of the model is implemented to analyse its behavior in a hypothetical cache management scenario.

In this paper we present a new visualisation approach in the effort of improving the evaluation strategy of space-time trellis code (STTC) generator matrix G. To our knowledge, although visualisation is widely used to handle a variety of problems, it has never been employed specifically to solve complexity problems that are related to generator matrix G evaluation. Most approaches are either mathematically or algorithmically inclined. As such, they tend to offer a series of refinement that enhances the current available method, but do not provide fresh insight on the problem at hand. By comparing it with the enhancement strategy that was discovered via the normal approach (i.e., by analysing algorithm) it was discovered that visualisation had inspired an entirely different pruning technique that outperformed the common approach by 20%.