Research On The Decoding Algorithms Of Channel Codes In3GPP UMTS And LTE Systems

In the3rdGeneration Partnership Project (3GPP) Universal Mobile TelecommunicationsSystem (UMTS) and Long Term Evolution (LTE) systems, turbo codes, convolutional codesand short linear block codes have been applied as channel codes. In addition, the jointsource-channel coding (JSCC) technique has been applied implicitly in UMTS, such as voicecoding and transmission. In this paper, channel decoding algorithms of turbo codes,convolutional codes, short linear block codes and JSCC are studied, and implementabledecoding algorithms to these codes are provided.With the increment of frequency bandwides, and the development of advancedtechnologies such as MIMO, the throughput requirement of turbo decoders is becoming muchhigher. One of the essential techniques to achieve high throughput is the parallel decoding,which can make full utilization of hardware resources in parallel ways to save the decodingtime. Based on the conventional iterative turbo decoding principle, parallel decodingalgorithms are studied in this paper, which include: parallel decoding structure, parallel(de)interleaving by using ring interleaver bottleneck break for UMTS turbo code, and earlystopping. In addition, a simple scheme of sharing iteration resources among multiple codeblocks is provided, which improves the decoding performance without increasing hardwarecomplexity.A24-bit cyclic redundancy check (CRC) is always attached before a UMTS or LTE turboencoder for the purpose of error detection. When the code block size is big, the redundancytaken by CRC can be neglected. But when the code block size is small, the CRC redundancycannot be ignored. Based on the iterative turbo decoding, a CRC assisted turbo decodingalgorithm is proposed in this paper. In the proposed algorithm, the ordered statistics decoding(OSD) is performed if the turbo iterative decoding is unsuccessful. The log-likelihood ratios(LLRs) of different iterations are accumulated to overcome the oscillation of LLRs. In the OSD decoding process, the turbo code and the CRC code are taken as a whole, which meansthe generation/check matrix is a combination of those matrices of the turbo code and the CRCcode. A normalized Euclidean distance based error detection method is proposed to evaluatethe reliability of the decoding result. The whole CRC-assisted turbo-OSD decoding algorithmcan significantly improve the performance of turbo codes with short block sizes, whilekeeping the undetected error rates sufficient low to meet the requirements of practical servicessuch as VoIP in the wireless communication systems.A convolutional code is usually decoded using the Viterbi algorithm, while in this paper,we emphasis on the CRC aided Viterbi algorithm, parallel list Viterbi algorithm (PLVA). Weanalyze the principle of PLVA decoding, and make a detailed description on the PLVAdecoding algorithm, which includes the selection of local and the global survival paths, andthe trace back of global survival each path. In order to further enhance the coding efficiency, atail-biting convolutional code (TBCC) which does not have tail bits, is adopted by LTE. Wraparound Viterbi algorithm (WAVA) is an efficient decoding algorithm dedicated to TBCCs. Inthis paper, an algorithm with combination of WAVA and PLVA is analyzed. Comparing withthe conventional PLVA algorithm which utilizes convergence length, WAVA-PLVA has lowercomputation complexity when the code block size is small.An adaptive multi-rate (AMR) speech frame in UMTS includes three classes of bitstreams: Class A, B and C. The bits of Class A have more significant impact on speech qualitythan those of Class B and C. The bits of Class A are attached with CRC while bits of Class Band C have not CRC attachments. Furthermore, lower coding rate are applied on Class A thanthose on Class B and C. The purpose to do this is to make the bits of Class A more robust andreduce the transmission overhead of Class B and C. The unequal protection of Class A, B andC can be regarded as a kind of JSCC in broad sense. Based on the characteristics of sourceand channel coding of AMR speech in UMTS, we propose a novel and practical decodingscheme called ‘Dual-CRC’ based on PLVA to improve the speech quality, with little modification of the current system architecture. We also employ the well-known jointsource-channel decoding (JSCD) scheme together with LVA decoding to further improve thespeech quality. Simulation results show that the two proposed decoders, the Dual-CRC LVAand the combination of the LVA and the JSCD, can significantly improve the speech quality.To the short linear block codes, three maximum likelihood based algorithms, brute force,2A-point FHT based algorithm and32-point FHT based algorithm, are analyzed in this paper.By detailed comprising the number of all kinds of operations of different algorithms, it can beconcluded that the32-point FHT based algorithm has the lowest computation complexity.