Shuffled Decoding of Serial Concatenated Convolutional Codes

Shuffled decoding enables to accelerate the extrinsic information exchange during iterative decoding of concatenated codes. It has already been applied to parallel convolutional codes or low-density parity-check codes. In this article, we propose to apply shuffled decoding to serial concatenation convolutional codes. We take advantage of their systematic encoding to propose an efficient shuffled decoding scheme. Compared to a standard iterative decoding scheme, the convergence of our shuffled implementation is obtained within fewer iterations, each one requiring also less time to be completed. This convergence acceleration yields doubling the throughput. We finally show that doubling the throughput comes at a lower cost than doubling the hardware resources, making this shuffled scheme efficient in term of implementation. For instance, the memory usage is 29% more efficient thanks to our proposal than a baseline scheme, which significantly reduces the power consumption of hardware decoders.