Research And Application Of LT Code With Feedback

Digital fountain code is a class of rateless erasure code which is based on tanner graph. The encoded packets can be generated online. Thus it can adapt to the time-varying channel conditions. Digital fountain code is widely used in many aspects, such as multicast transmission, big data storage and multi-source download, with its simple efficient encoding and decoding algorithms and no intermediate feedbacks and so on. Therefore, digital fountain code has been attached great importance in channel coding field of communication gradually and has become a research hotspot. This paper focuses on the research of LT codes with feedback, in order to reduce the decoding redundancy.Firstly, this paper introduces the erasure channel model and the basic concepts and principles of digital fountain codes. The XOR coding algorithm and two kinds of decoding algorithms which are called GE decoding algorithm and BP decoding algorithm are presented in detail. Furthermore, we compare the advantages and disadvantages of the two kinds of decoding algorithms by simulation. Three kinds of degree distributions which are familiar in LT encoding process are introduced and compared. They are robust soliton distribution, ideal soliton distribution and all-at-once distribution. Then the analysis stresses the characteristic of robust soliton distribution.Secondly, considering the influence of repetitive coding and different selection of source packets, we propose an optimized scheme named unrepeatable maximum degree selection scheme which is based on a novel LT coding scheme with limited feedback packets. The ratio of successful decoding, the probability of successful decoding and the number of required feedback packets when successful decoding of these two schemes are compared by Matlab simulation. And then the effectiveness of the optimized scheme is verified.Finally, we analyze the common problem which exists in exsisting unequal error protection schemes based on LT code, namely we should spend more prices to restore lower prioritized packets, while keeping excellent decoding performance of prioritized packets. In order to solve this problem, the idea of feedbacks is applied to unequal error protection. Using traditional UEP-LT scheme as an example, UEP-LT scheme with feedbacks is proposed. This scheme improves the decoding performance of lower prioritized packets greatly. In addition, it is shown that when the number of restored source packets ’k ?0.1?y( y indicates the number of unrestored packets), the number of encoded packets which is required for restore all source packets is at least by simulation. With this conclusion, we propose a more suitable scheme for poor channel conditions and broadcast or multicast situations. That is UEP-LT scheme with a single feedback. Then we compare UEP-LT scheme with feedbacks and UEP-LT scheme with a single feedback by simulations.