Abstract
Polar codes are recently invented promising forward error correction codes, which are considered in dierent global standards including the fth-generation wireless technology (5G).The length of a polar code is in order of power of two. Moreover, the performance of short-length polar codes is not as competitive as other codes. These features limit their usage in services (in particular, in video services), which are highly demanded in the current and next generation of broadband networks and require dierent error protection levels. In order to construct an arbitrary length polar code, puncturing or shortening technique can be used, which removes a number of bits from the codeword. In general, nding an optimum puncturing algorithm is not an easy task. This is mainly evident for long length codes or codes require a high number of punctured bits. Alternatively, the multi-kernel technique provides exible length codes but at the expense of high complexity on their design.
This thesis presents an ecient puncturing technique for polar codes. It is con-
ducted by dening sub-codes of the main polar code and implementing a reliable
puncturing algorithm for each sub code. It also represents a new scheme of serially concatenated codes, whose lengths are the product of the length of constituent polar codes. Two puncturing algorithms are presented for these codes, where bits are removed from one or both constituted codes. As an interactive decoding scheme is applied for them, the decoder properly recovers the punctured information in dierent iterations.
More exibility on the design of polar based serially concatenated codes is achievable by utilizing a convolutional interleaver, which inserts a number of redundant bits to the codeword obtained from the outer code. Due to the known feature of redundant bits, it is possible to construct a variable-length polar code, which outperforms other polar codes formed by puncturing, shortening or multi-kernel techniques.
| Date of Award | Jul 2020 |
|---|---|
| Original language | English |
| Supervisor | Sina Vafi (Supervisor) & Erwin Chan (Supervisor) |