| Turbo code, which utilizes the condition of random coding and decoding of Shannon noisy channel coding theory, has obtained the performance near to Shannon theoretic limit. It is one of the best channel coding schemes in all presented coding methods, and has been accepted as one of the third generation communication system standards and of the deep space communication system. The principle of turbo code has been applicated in several related fileds. However, the high computational complexity and the ieal premise have restricted its further applications. The practical turbo decoding scheme has to give consideration to some practical issues, e.g. the decoding performance, the computational complexity as well as the the system robustness.Turbo decoding over the additive white Gaussian noise (AWGN) channel and the Rayleigh fading channel, as well as turbo equalization over two typical ISI (Inter-symbol Interference) channels have been deep studied in this thesis. Several practical approaches for turbo decoding and turbo equalization have been proposed.A novel method based on the correction function is proposed, which could evaluate the performance of the turbo decoding algorithm as well as the signal-to-noise (SNR) sensitivity. A practical turbo decoding scheme over the AWGN channel has been given by considering the computational complexity, the decoding performance and the SNR sensitivity. The proposed scheme does not need to estimate the channel SNR. It is insensitivity to SNR offsets, which results as a robust turbo decoding (RTD) method over the AWGN channel.It is proved that for turbo decoding over the Rayleigh fading channel, the Log-MAP algorithm with the channel reliability factor proposed by Frenger, is equivalent to an SNR underestimation. The mathematical expression for this underestimation is given. As the underestimation value tends to a constant when the SNR goes to infinity, a novel method for estimating the channel reliability factor with a constant SNR offset is proposed. The proposed method performs as well as Frenger's, and does not need to estimate the variance of the fading factor and the variance of the estimate noise, which could greatly lower the risk of performance degradation caused by the parameter estimate offsets.It is proved that the Max-Log-MAP algorithm is insensitivity to all parameters of the Rayleigh fading channel, and this is still held when the Rayleigh fading channel is not perfectly estimated. By scaling the soft output the Max-Log-MAP decoder, the decoding performance of the sub-optimal algorithm could be greatly improved without any increase of computational complexity. The proposed decoding scheme does not require any channel parameter, which is also an RTD technique over the Rayleigh fading channel. Two simplified methods for linear turbo equalization are proposed. The minimum mean squared error (MMSE) equalizer could be viewed as a non-stationary AWGN channel. For the linear turbo equalization, the estimate (?)_k has a linear relationship with the soft output of the equalizer over both the moderate-loss and the high-loss ISI channels. As a result, the non-stationary AWGN channel could be approximately considered as a stationary one. Two simplified methods, named as the moment method and the experiential method, are given to estimate the mean and the variance of the stationary AWGN channel. The soft output of the MMSE equalizer could be directly estimated from the estimate (?)_k . The proposed approaches perform as well as the convetional method, and reduce the computational complexity.A novel approach for analyzing the SNR sensitivity is given, which is based on a statistical model. The classical method is proved as a special case when the noise of the SNR estimate is 0. However, from a practical point of view the SNR should be supposed to be unbiased. The function between the variance of the noise estimate and the variance of the SNR estimate is given. The effects of the biased and unbiased estimate, as well as the estimate noise to the maximum a posteriori (MAP) algorithms are studied. Finally, a practical scheme for turbo equalization is given by considering the computational complexity, the decoding performance as well as the SNR sensitivity.
|
PDF Full Text Request