Near-Maximum-Likelihood Decoding Techniques with Reduced Complexity

降低复杂度的近最大似然解码技术

• 批准号: 9703844
• 负责人: Ilya Dumer
• 金额: $ 32.23万
• 依托单位: University of California-Riverside
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 1997
• 资助国家: 美国
• 起止时间: 1997-09-15 至 2001-08-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=9703844&HistoricalAwards=false
• 关键词: Near; Maximum; Likelihood; Decoding; Techniques

New coding applications are often limited by the huge decoding complexity required by the best possible maximum-likelihood (ML) decoding. Recently designed sub-optimal algorithms reduce the complexity exponent of ML decoding two times along with an arbitrarily small increase in decoding error probability. The project addresses the problem of designing more advanced algorithms that can further reduce the complexity exponent of ML decoding five to seven times. We are going to investigate the following topics. General study of sub-optimal algorithms. The goals are: (a) to obtain the explicit trade-offs between complexity and performance of near-ML decoding, (b) to generalize near-ML decoding for practically important channels, (c) to combine new near-ML decoding techniques with conventional trellis decoding. Design of new algorithms for near-ML decoding. The goals are: (a) to develop new presorting procedures which reduce the complexity exponent of ML decoding up to three times, (b) to design new random-search algorithms which reduce the latter exponent up to five times. Design of cascaded NML decoding algorithms. The goals are: (a) to develop cascaded near-ML decoding for concatenated codes, and to reduce the complexity exponent of ML decoding up to seven times, (b) to apply cascaded design for bursty and fading channels. Real-time software implementation of designed algorithms. The goal is to develop fast software algorithms for codes of lengths 50 to 100 used over the channels with a signal-to-noise ratio of about 1 dB.

新的编码应用通常受到最佳最大似然 (ML) 解码所需的巨大解码复杂性的限制。最近设计的次优算法将 ML 解码的复杂度指数降低了两倍，同时解码错误概率也略有增加。该项目解决了设计更先进算法的问题，这些算法可以进一步将 ML 解码的复杂性指数降低五到七倍。 我们将研究以下主题。 次优算法的一般研究。目标是：(a) 获得近 ML 解码的复杂性和性能之间的明确权衡，(b) 将近 ML 解码推广到实际重要的通道，(c) 将新的近 ML 解码技术与传统的网格解码相结合。 近机器学习解码的新算法设计。目标是：(a) 开发新的预排序程序，将 ML 解码的复杂性指数降低最多三倍，(b) 设计新的随机搜索算法，将后者指数降低最多五倍。 级联NML解码算法的设计。目标是：(a) 为级联代码开发级联近 ML 解码，并将 ML 解码的复杂性指数降低七倍，(b) 针对突发和衰落信道应用级联设计。 设计算法的实时软件实现。目标是开发用于通道上使用的长度为 50 至 100 的代码的快速软件算法，信噪比约为 1 dB。