Reduced memory decoding schemes for turbo decoding based on storing the index of the state metric. Issue 12 (1st August 2014)

• Record Type: Journal Article
• Authors: Zhan, Ming; Wu, Jun; Wen, Hong
• Is Part Of: IET communications. Volume 8:Issue 12(2014)
• Page Start: 2095
• Page End: 2105
• Publication Date: 2014-08-01
• Subjects: turbo codes -- cache storage -- decoding -- binary codes -- convolutional codes -- error statistics; near optimal decoding scheme -- nonlinear estimation‐based decoding scheme -- max‐log‐MAP algorithm -- duo binary convolutional turbo code -- binary turbo code -- linear estimation‐based scheme -- BER simulation -- bit error rate simulation -- Jacobian logarithm -- index sequence -- forward state metrics -- compressing module -- regeneration module -- power dissipation -- SMC -- state metric cache size -- state metric index -- turbo decoding scheme -- reduced memory decoding schemes; Telecommunication systems -- Periodicals; Speech processing systems -- Periodicals; 621.38205
• DOI: 10.1049/iet-com.2013.0672
• Languages: English
• ISSNs: 1751-8628

Abstract

Abstract : In the implementation of turbo‐like decoder, the size of state metrics cache (SMC) has a predominant impact on the core area and the overall power dissipation. Different from previous reported decoding schemes, in the proposed decoding schemes, a compressing module and a regeneration module are added to the decoder. The compressing module sorts the forward state metrics from the minimum to the maximum, by which an index sequence and the corresponding increase metrics are calculated, and subsequently are stored in the SMC. In the regeneration module, the forward state metrics are estimated with the index sequence and the increase metrics that accessed from the SMC. With the cost of dummy calculation that is performed by the compressing and the regeneration modules, two decoding schemes are proposed. For an eight‐state turbo codes, the linear and the nonlinear estimation based decoding schemes reduce the SMC size by 62.5% and 57.5%, respectively. The bit error rate (BER) simulation is performed for both binary turbo code and duo binary convolutional turbo code, and shows BER of the linear estimation‐based scheme is superior to that of the enhanced max‐log‐MAP (the maximum a posteriori probability) algorithm, whereas BER of the non‐linear estimation‐based decoding scheme is very close to that of the near optimal decoding scheme.