Method and apparatus for decoding turbo-encoded code sequence

Abstract

The method starts by initializing the forward recursive probability function vector α0 and the backward recursive probability function vector βN. Next, transition probability matrices Γ(Rk) and Γi(Rk) are determined from the received codes of the sequence R1N. Next, the value of the vector αk corresponding to the received Rk is determined from Γ(Rk). While determining αk, multiple multiplication operations on Γ(Rk) and Γ1(Rk) are performed in parallel. After receiving the complete code sequence R1N, all the values ​​of the backward recursive probability vector βk (here, k = 1, 2, ..., N-1) are determined in parallel by using the results of the matrix multiplication operation, and also about The logarithmic likelihood ratio of each decoded bit dk (here, k=1, 2, . . . , N). The circuit uses a set of regular matrix operations to perform continuous decoding processing in parallel.

Description

technical field [0001] The present invention relates to error correction (error correction) decoding for digital signals received via channels containing noise, and more particularly to a maximum induction (MAP) decoding method of TURBO codes based on parallel matrix processing. Background technique [0002] In mobile wireless communication, multi-channel fading, inter-symbol interference and thermal noise caused by electronic equipment can cause many transmission errors. Channel coding techniques are general methods used to obtain high quality signals in receivers. The gist of the most popular method of channel coding techniques has been pointed out by Proakis in "Digital Communications" published by McGraw-Hill International Editions in 1989. [0003] In recent years, Berrou et al. first described the newly differentiated binary parallel continuous recursive organization convolution in the 1993 International Conference on IEEE Communications "Near S...

AI Technical Summary

Problems solved by technology

[0006] However, there are serious shortcomings in the existing MAP method, therefore, it is difficult to achieve low-cost VLSI implementation

Of most concern: the complex calculations necessary for forward-backward iterations, which are necessary for MAP-based methods, result in decoding delays and large storage capacities

Almost all leading-edge MAP decoders and convolutional code decoders (refer to, for example, Comatlas of France, Chateaubourg published "CAS 5093 Turbo-Code Codec, data sheet", August 1994, Eastlake, Ohio, USA EfficientChannel Coding, Inc. published "ECC Turbo product code technology" in March 1998, and Small World Communication in Australia and Adelaide published "MAPO4 and MAPO4A 16 State MAP Decoders" in April 1998) requires considerable complexity

Images