17-04-2013, 04:38 PM
Convolutional interleaver for unequal error protection of turbo codes
[attachment=53132]
Abstract:
This paper describes construction of a convolutional interleaver as a block interleaver and discusses its application to turbo codes with equal and unequal error protection techniques. Based on simulations, different convolutional interleaver structures suitable for turbo codes with unequal error protection capability are suggested. Finally, based on conducted simulations the best method is selected.
INTRODUCTION
Compression has become the norm in mobile communications; currently speech dominates but video and general multimedia compression is fast becoming important. A vital characteristic of a compressed data stream is the unequal perceptual importance of the compressed parameters. In turn, this leads to unequal effects of errors during transmission. Thus Unequal Error Protection (UEP) has become an important part of design of channel encoders for multimedia content.
Amongst the known channel codes, turbo codes
have the highest performance in direct bit error rate
reduction terms. Turbo codes are produced by the
parallel concatenation of two or more convolutional
code generators separated by one or more
interleavers [1]. The reduction in BER resulting
from such codes is directly dependent on the design
of a suitable interleaver. In the literature, several
techniques for implementing UEP using turbo codes
have been proposed. In [2], single interleaver for all
levels and, in [3], one interleaver for each level are
suggested, while [4] discusses optimization of the
single interleaver solution.
CONVOLUTIONAL INTERLEAVER STRUCTURE
A convolutional interleaver consists of T parallel lines, with different delay in each line that represents interleaver period [5]. In general, each successive line has a delay which is M symbols durations higher than the previous line. The structure for M=1 and T=3 is illustrated in Figure 2.
INTERLEAVER FOR UNEQUAL ERROR
PROTECTION
For turbo codes with unequal error protection, the fixed interleaver period is considered for all levels, and, depending on the required protection level, the length of the data block at each level is altered. As an example, we consider three protection levels for data with length of 2048 bits. Specifications of these levels have been presented in Table 1. We assume that 2048 bits are split into 5 groups.
At the beginning of each group, 64 bits of data are allocated the protection level 1. Then, there are some blocks of length L=42 with the protection level 2, and the remaining blocks of length L=32 are allocated the protection level 3.
CONCLUSIONS
In this paper, we proposed a technique, which
initializes convolutional interleaver memories by the
insertion of stuff bits at the end data of blocks of
specific lengths. This process results in a semi-block
interleaver structure for turbo coding. The
performance of this new structure has been
compared with that of conventional block
interleavers and applied to unequal error protection
turbo codes. Two major methods were proposed and
tested. In each method, two schemes depending on
insertion of stuff bits before or after the tail bits were
investigated. The results show that a stuff bit
insertion before the tail bits in the interleaver with
variable period will produce the best performance.
Further work will consider finding a suitable
compromise between the turbo-code performance
and the stuff-bit redundancy.
[attachment=53132]
Abstract:
This paper describes construction of a convolutional interleaver as a block interleaver and discusses its application to turbo codes with equal and unequal error protection techniques. Based on simulations, different convolutional interleaver structures suitable for turbo codes with unequal error protection capability are suggested. Finally, based on conducted simulations the best method is selected.
INTRODUCTION
Compression has become the norm in mobile communications; currently speech dominates but video and general multimedia compression is fast becoming important. A vital characteristic of a compressed data stream is the unequal perceptual importance of the compressed parameters. In turn, this leads to unequal effects of errors during transmission. Thus Unequal Error Protection (UEP) has become an important part of design of channel encoders for multimedia content.
Amongst the known channel codes, turbo codes
have the highest performance in direct bit error rate
reduction terms. Turbo codes are produced by the
parallel concatenation of two or more convolutional
code generators separated by one or more
interleavers [1]. The reduction in BER resulting
from such codes is directly dependent on the design
of a suitable interleaver. In the literature, several
techniques for implementing UEP using turbo codes
have been proposed. In [2], single interleaver for all
levels and, in [3], one interleaver for each level are
suggested, while [4] discusses optimization of the
single interleaver solution.
CONVOLUTIONAL INTERLEAVER STRUCTURE
A convolutional interleaver consists of T parallel lines, with different delay in each line that represents interleaver period [5]. In general, each successive line has a delay which is M symbols durations higher than the previous line. The structure for M=1 and T=3 is illustrated in Figure 2.
INTERLEAVER FOR UNEQUAL ERROR
PROTECTION
For turbo codes with unequal error protection, the fixed interleaver period is considered for all levels, and, depending on the required protection level, the length of the data block at each level is altered. As an example, we consider three protection levels for data with length of 2048 bits. Specifications of these levels have been presented in Table 1. We assume that 2048 bits are split into 5 groups.
At the beginning of each group, 64 bits of data are allocated the protection level 1. Then, there are some blocks of length L=42 with the protection level 2, and the remaining blocks of length L=32 are allocated the protection level 3.
CONCLUSIONS
In this paper, we proposed a technique, which
initializes convolutional interleaver memories by the
insertion of stuff bits at the end data of blocks of
specific lengths. This process results in a semi-block
interleaver structure for turbo coding. The
performance of this new structure has been
compared with that of conventional block
interleavers and applied to unequal error protection
turbo codes. Two major methods were proposed and
tested. In each method, two schemes depending on
insertion of stuff bits before or after the tail bits were
investigated. The results show that a stuff bit
insertion before the tail bits in the interleaver with
variable period will produce the best performance.
Further work will consider finding a suitable
compromise between the turbo-code performance
and the stuff-bit redundancy.