TY - GEN
T1 - Low-redundancy bit interleaving
AU - Klein-Orbach, Alon
AU - Kochman, Yuval
N1 - Publisher Copyright: © 2016 IEEE.
PY - 2017/2/10
Y1 - 2017/2/10
N2 - We consider joint source-channel coding with bandwidth expansion, in a quadratic-Gaussian setting. We seek a scheme that will have favorable delay and robustness characteristics: it should be scalar (working on a single source sample), and closely follow the optimal curve of signal-to-distortion ratio (SDR) versus signal-to-noise ratio (SNR), asymptotically as the SNR grows. A natural candidate is bit interleaving, which was already proposed by Shannon in the context of bandwidth compression; however, in the case of expansion it suffers from severe noise amplification. A scheme proposed by Teherzadeh and Khandani introduces fixed bits in order to combat noise amplification, and has a multiplicative gap that is exponential the square root of the SNR from the ideal SDR-SNR curve. We propose an improved scheme that allocates the fixed bits in a source-dependent manner. This approach reduces the number of fixed bits needed, and thus has lower redundancy; we show that the multiplicative gap from the ideal curve is only linear in the SNR.
AB - We consider joint source-channel coding with bandwidth expansion, in a quadratic-Gaussian setting. We seek a scheme that will have favorable delay and robustness characteristics: it should be scalar (working on a single source sample), and closely follow the optimal curve of signal-to-distortion ratio (SDR) versus signal-to-noise ratio (SNR), asymptotically as the SNR grows. A natural candidate is bit interleaving, which was already proposed by Shannon in the context of bandwidth compression; however, in the case of expansion it suffers from severe noise amplification. A scheme proposed by Teherzadeh and Khandani introduces fixed bits in order to combat noise amplification, and has a multiplicative gap that is exponential the square root of the SNR from the ideal SDR-SNR curve. We propose an improved scheme that allocates the fixed bits in a source-dependent manner. This approach reduces the number of fixed bits needed, and thus has lower redundancy; we show that the multiplicative gap from the ideal curve is only linear in the SNR.
UR - http://www.scopus.com/inward/record.url?scp=85015252666&partnerID=8YFLogxK
U2 - 10.1109/ALLERTON.2016.7852339
DO - 10.1109/ALLERTON.2016.7852339
M3 - منشور من مؤتمر
T3 - 54th Annual Allerton Conference on Communication, Control, and Computing, Allerton 2016
SP - 969
EP - 973
BT - 54th Annual Allerton Conference on Communication, Control, and Computing, Allerton 2016
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 54th Annual Allerton Conference on Communication, Control, and Computing, Allerton 2016
Y2 - 27 September 2016 through 30 September 2016
ER -