Forward error correcting codes are used for digital communications over analog noisy channels. By transmitting redundant information, the reconstruction of the original messages is possible with exception of some residual error. A prime task of communications is to construct decoding methods which minimize this error under suitable assumptions. Here, we give an concise stochastic modelling of the abstract channel model for arbitrary binary linear block codes. Basing on a channel description by probability density functions, the exact mathematical treatment of all derived statements with stochastic tools becomes possible. Especially, the basic decoding techniques are classified and criterions for the construction of optimal decoding methods are given. Compared with the modelling in literature, generalized statements with arbitry continuous density functions are formulated and proofed. An optimal soft-output method for terminated convolutional codes is constructed and analyzed which is functionally comparable with the MAP algorithm. The given method is derived with a recursion formulation to minimize numerical operations. Numeric stability is achieved by a rescaling using the computation of the means of partial formulas. The soft-decision decoding of arbitrary binary linear block codes is treated by a new error optimal algorithm. Thereby, a specific adaptive code transformation on receiver side is used to transform the channel code such that a economical branch-and-bound method is applicable. The error optimal decoding in the transformed region allows for a distinct reduction of numerical steps in comparison to a branch-and-bound method operating on the original code tree. In summary, generalized issues for the error optimal decoding of arbitrary binary block codes are stochastically modelled and analyzed. For AWGN channels without specialization on code classes, an optimal soft-decision algorithm is entirely evaluated. Also, for the class of terminated convolutional codes, the soft-output decoding is completely described by an optimal method.
«Forward error correcting codes are used for digital communications over analog noisy channels. By transmitting redundant information, the reconstruction of the original messages is possible with exception of some residual error. A prime task of communications is to construct decoding methods which minimize this error under suitable assumptions. Here, we give an concise stochastic modelling of the abstract channel model for arbitrary binary linear block codes. Basing on a channel description by p...
»