Researches On Some Key Problems Of Turbo Codes

Turbo codes are parallel concatenated convolutional codes originally, the encoder of Turbo codes consists of two same encoders and one interleaver. The first encoder encodes the information sequence directly, while the second encoder encodes the sequence which are the scrambled version (through an interleaver) of those informations entering the first encoder. Block Turbo codes and hybrid concatenated codes are presented now. The scopes of Turbo codes are expanded. The study of Turbo codes includes the designs of the encoders, designs of the interleavers, and the researches on soft-in soft-out iterative decoders. Turbo codes have a very wide foreground. There are many new laws to study. However, so far, reports on the research of Turbo codes are still insufficient.Firstly, the research backgrounds, the developments and trends of concatenated codes are introduced. Some problems during the research are pointed out. The skeleton and academic value of this paper are illuminated. The construction principles of the encoder and the decoder of Turbo codes are also presented, basic theories about this paper are introduced, and the advantages and the disadvantages of Turbo code performances are analyzed as well.Secondly, the maximum a posterior (MAP) is discussed and illuminated. On the base of MAP, a newly serial decoder construction is provided, which is used to improve the parallel decoder construction whose iterative gain need to be guessed withsimulated annealing methods. The serial decoding method is proved using Pearl's propagation algorithm in artificial intelligence. Its iterative gain is always one. The complexity of serial decoding method is reduced. The result obtained by the serial decoding method is compared with the result obtained by the parallel decoding method. The performance curves getted by these two methods are the same.The techniques of encoding and decoding of parallel concatenated extended Hamming codes are studied, and the trellises of extended Hamming codes from the trellises of Hamming codes are obtained. The performances of Hamming codes and extended Hamming codes of parallel concatenation are analyzed. Then the codes are obtained, which can operate with Eb/No about IdB away from theoretical limit in the medium to high coding rate range. Simulation results are provided. This supplements the existing methods, such as the parallel concatenated convolutional codes which are most suitable for rates up to 1/2 and the parallel concatenated Hamming codes which are most suitable for rates higher than 2/3. The influences of different interleavers and of the different interleaver lengths on the performances of parallel concatenated codes are studied. The interleaver is obtained, which is most fit for parallel concatenated extended Hamming codes.The techniques of encoding and decoding of parallel concatenated Hadamard codes of low rate are studied, and reduced input fast Hadamard transform and reduced output fast Hadamard transform are investigated. The simple and efficient decoding methods based on the fast Hadamard transformation are obtained,and a kind of codes are gained, which is fitted for mobile speech systems where short frame length is mandatory but relatively high BER (Bit Error Rate), e.g., 10-2~10 3, is tolerable. The code whose frame length is 200 bits and BER is 10"3 with Eb/No=OAdB and BER is 10"2 with Eb/No = -0.4dB is found out.Finally, because of inavoidable floor response of Turbo codes, the hybrid concatenation methods are presented, which make use of extended Hamming codes as outter codes and parallel concatenated Hadamard codes as inner codes and make them in serial concatenation mode. The coding technique, decoding technique and linking technique of hybrid concatenated codes are investigated. In order to get good performance in both low Eb/No and high Eb/No the switch techniques are adopted. A kind of codes are obtained, which are fitted for mobile data systems and whose Eb/No is 1.7dB with about 200 bits of frame length in BER =10"5 and Eb/No is l.ldB with about 280 bi...