Parallel progressive hierarchical turbo codes

This paper proposes a modified parallel turbo structure based on embedding the unequal error protection concepts in parallel concatenated convolutional codes. These concepts may be required for compressed data packets transmission whose sensitivities to channel errors are not uniformly distributed. We develop a new method for obtaining turbo gains for unequally protected classes of bits by using a parallel concatenation of encoders separated by increasing size interleavers. A progressive and hierarchical encoding of the decreasing priority bits of one data frame is performed through consecutive partial rate-1 dimensions projections. The result is one data frame whose bits are encoded in a number of dimensions proportional to their respective priority in terms of channel sensitivity to errors. The iterative decoding algorithm is modified in consequence; turbo gains are consistent for all priority subframes during iterations and a noticeable increased gain is obtained for lower level bits. Performances for a two priority data frame are given in comparison with the classical structure of parallel turbo codes and recursive systematic convolutional code.