In this paper, we present the concept of non-uniform channel polarization and a scheme for rate-compatible polar code construction for any codeword length using additive white Gaussian noise (AWGN) channels and the successive cancellation (SC) decoder. A Non-Uniform Polarization technique based on the Gaussian Approximation (NUPGA) is developed and an efficient rate-compatible design technique is devised to choose the best channels for transmission by a process of re-polarization of the codeword with the desired length. Simulations illustrate the proposed NUPGA design against existing rate-compatible techniques.
I. INTRODUTIONPolar codes, introduced by Arikan [1], were proved to achieve the symmetric capacity of binary input symmetric discrete memoryless channels (B-DMCs) under a Successive Cancellation (SC) decoder as the codelength goes to infinity. Polar codes are based on the phenomenon of channel polarization. The channel polarization theorem states that, as the codelength N goes to infinity, a polarized bit-channel becomes either a noiseless channel or a pure noise channel. The information bits are transmitted over the noiseless bit-channels and the pure noise bit-channels are set to zero (frozen bits). The construction of a polar code involves the identification of channel reliability values associated to each bit to be encoded. This identification can be effectively performed for a code length and a specific signal-to-noise ratio.Among the most well-known code construction techniques are the Bhattacharyya-based approach of Arikan [1], the Density Evolution (DE) schemes of Mori [2], [3] and Tal [4], the Gaussian Approximation (GA) technique of [5] and the Polarization Weigth (PW) algorithm [6]. The Bhattacharyya parameter-based approach that was proposed by Arikan along with Monte-Carlo (MC) simulations to estimate bit channel reliabilities [1]. The DE method proposed in [2] and [3] can provide theoretical guarantees on the estimation accuracy at a high computational cost. Extending the ideas of [2][3], Tal [4] devised two approximation methods by which one can get upper and lower bounds on the error probability of each bit-channel efficiently. Since DE includes function convolutions, its precision is limited by the complexity. A bit-channel reliability estimation method for additive white Gaussian noise (AWGN) channels based on GA of DE has been proposed in [5], giving accurate results with limited complexity. The PW algorithm [6] is a recent attempt to study