Maximum-likelihood detection for MIMO systems based on differential metrics

Ming-Xian Chang, Wang Yueh Chang

Research output: Contribution to journalArticle

8 Citations (Scopus)

Abstract

The multiple-input multiple-output (MIMO) system makes efficient use of spectrum and increases the transmission throughput in wireless communications. The sphere decoding (SD) is an efficient algorithm that enables the maximum-likelihood (ML) detection for the MIMO system. However, the SD algorithm has variable complexity, and its complexity increases rapidly with decreasing signal-To-noise ratio (SNR). In this paper, we propose a novel ML detection algorithm for the MIMO system based on differential metrics.We define the differential metrics and derive the associated recursive calculation. We then give the indicative functions, which can be used to possibly find some ML-detected bits of the initial sequence. The indicative functions are further applied to implement an efficient tree search for ML detection. The proposed algorithm does not need QR decomposition and matrix inversion. The tree search process needs only the additive operation, while the number of multiplications before the tree search is constant. Our algorithm can achieve the exact ML detection as the SD algorithm. Unlike the SD algorithm, the complexity of our algorithm reduces with decreasing SNR, whereas at high SNR, the complexity is nearly constant.We also give the convergence analysis for the SD and proposed algorithms, and the simulation verifies our analysis. For the proposed algorithm, the number of necessary memory is constant during the tree search, and the implementation by parallel processing is possible. The soft output of ML-detected bits can also be generated in our algorithm.

Original languageEnglish
Article number7912332
Pages (from-to)3718-3732
Number of pages15
JournalIEEE Transactions on Signal Processing
Volume65
Issue number14
DOIs
Publication statusPublished - 2017 Jul 15

Fingerprint

Maximum likelihood
Decoding
Signal to noise ratio
Trees (mathematics)
Throughput
Decomposition
Data storage equipment
Communication
Processing

All Science Journal Classification (ASJC) codes

  • Signal Processing
  • Electrical and Electronic Engineering

Cite this

@article{2a2845ae826a4c61a21f226376611788,
title = "Maximum-likelihood detection for MIMO systems based on differential metrics",
abstract = "The multiple-input multiple-output (MIMO) system makes efficient use of spectrum and increases the transmission throughput in wireless communications. The sphere decoding (SD) is an efficient algorithm that enables the maximum-likelihood (ML) detection for the MIMO system. However, the SD algorithm has variable complexity, and its complexity increases rapidly with decreasing signal-To-noise ratio (SNR). In this paper, we propose a novel ML detection algorithm for the MIMO system based on differential metrics.We define the differential metrics and derive the associated recursive calculation. We then give the indicative functions, which can be used to possibly find some ML-detected bits of the initial sequence. The indicative functions are further applied to implement an efficient tree search for ML detection. The proposed algorithm does not need QR decomposition and matrix inversion. The tree search process needs only the additive operation, while the number of multiplications before the tree search is constant. Our algorithm can achieve the exact ML detection as the SD algorithm. Unlike the SD algorithm, the complexity of our algorithm reduces with decreasing SNR, whereas at high SNR, the complexity is nearly constant.We also give the convergence analysis for the SD and proposed algorithms, and the simulation verifies our analysis. For the proposed algorithm, the number of necessary memory is constant during the tree search, and the implementation by parallel processing is possible. The soft output of ML-detected bits can also be generated in our algorithm.",
author = "Ming-Xian Chang and Chang, {Wang Yueh}",
year = "2017",
month = "7",
day = "15",
doi = "10.1109/TSP.2017.2698411",
language = "English",
volume = "65",
pages = "3718--3732",
journal = "IEEE Transactions on Signal Processing",
issn = "1053-587X",
publisher = "Institute of Electrical and Electronics Engineers Inc.",
number = "14",

}

Maximum-likelihood detection for MIMO systems based on differential metrics. / Chang, Ming-Xian; Chang, Wang Yueh.

In: IEEE Transactions on Signal Processing, Vol. 65, No. 14, 7912332, 15.07.2017, p. 3718-3732.

Research output: Contribution to journalArticle

TY - JOUR

T1 - Maximum-likelihood detection for MIMO systems based on differential metrics

AU - Chang, Ming-Xian

AU - Chang, Wang Yueh

PY - 2017/7/15

Y1 - 2017/7/15

N2 - The multiple-input multiple-output (MIMO) system makes efficient use of spectrum and increases the transmission throughput in wireless communications. The sphere decoding (SD) is an efficient algorithm that enables the maximum-likelihood (ML) detection for the MIMO system. However, the SD algorithm has variable complexity, and its complexity increases rapidly with decreasing signal-To-noise ratio (SNR). In this paper, we propose a novel ML detection algorithm for the MIMO system based on differential metrics.We define the differential metrics and derive the associated recursive calculation. We then give the indicative functions, which can be used to possibly find some ML-detected bits of the initial sequence. The indicative functions are further applied to implement an efficient tree search for ML detection. The proposed algorithm does not need QR decomposition and matrix inversion. The tree search process needs only the additive operation, while the number of multiplications before the tree search is constant. Our algorithm can achieve the exact ML detection as the SD algorithm. Unlike the SD algorithm, the complexity of our algorithm reduces with decreasing SNR, whereas at high SNR, the complexity is nearly constant.We also give the convergence analysis for the SD and proposed algorithms, and the simulation verifies our analysis. For the proposed algorithm, the number of necessary memory is constant during the tree search, and the implementation by parallel processing is possible. The soft output of ML-detected bits can also be generated in our algorithm.

AB - The multiple-input multiple-output (MIMO) system makes efficient use of spectrum and increases the transmission throughput in wireless communications. The sphere decoding (SD) is an efficient algorithm that enables the maximum-likelihood (ML) detection for the MIMO system. However, the SD algorithm has variable complexity, and its complexity increases rapidly with decreasing signal-To-noise ratio (SNR). In this paper, we propose a novel ML detection algorithm for the MIMO system based on differential metrics.We define the differential metrics and derive the associated recursive calculation. We then give the indicative functions, which can be used to possibly find some ML-detected bits of the initial sequence. The indicative functions are further applied to implement an efficient tree search for ML detection. The proposed algorithm does not need QR decomposition and matrix inversion. The tree search process needs only the additive operation, while the number of multiplications before the tree search is constant. Our algorithm can achieve the exact ML detection as the SD algorithm. Unlike the SD algorithm, the complexity of our algorithm reduces with decreasing SNR, whereas at high SNR, the complexity is nearly constant.We also give the convergence analysis for the SD and proposed algorithms, and the simulation verifies our analysis. For the proposed algorithm, the number of necessary memory is constant during the tree search, and the implementation by parallel processing is possible. The soft output of ML-detected bits can also be generated in our algorithm.

UR - http://www.scopus.com/inward/record.url?scp=85021745487&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85021745487&partnerID=8YFLogxK

U2 - 10.1109/TSP.2017.2698411

DO - 10.1109/TSP.2017.2698411

M3 - Article

AN - SCOPUS:85021745487

VL - 65

SP - 3718

EP - 3732

JO - IEEE Transactions on Signal Processing

JF - IEEE Transactions on Signal Processing

SN - 1053-587X

IS - 14

M1 - 7912332

ER -