#### DMCA

## Channel-Aware Decision Fusion in Distributed MIMO Wireless Sensor Networks: Decode-and-Fuse vs. Decode-then-Fuse

Citations: | 13 - 8 self |

### Citations

465 |
Digital communication over fading channels: A unified approach to performance analysis, NewYork:
- Simon, Alouini
- 2000
(Show Context)
Citation Context ...upper) “communication” bound [5], i.e. PE0 under the assumption of ideal sensors, i.e. (PD,PF)=(1, 0). In this case the bound is represented by the symbol error probability of a N branch MRC combiner =-=[32]-=-, that is P comm = E0 ( )N 1−μ ∑N−1 ( ) N−1+l 1+μ 2 l=0 l ( 2 )l √ ,where SNR μ � SNR+N . Simulations show that all the rules, except for MRC, present a unimodal behaviour, i.e. there exists a finite ... |

371 |
Fundamentals of Statistical Signal Processing, Volume I: Estimation Theory.
- Kay
- 1993
(Show Context)
Citation Context ...Fig. 2a, the DFC aims at detecting the presence of the target directly from the received signal vector without any intermediate step to decode the transmitted vector. A. Optimum Rule The optimal test =-=[24]-=- for the considered problem can be formulated as [ ] H=H1 ˆ p(y|H1) Λopt � ln ≷ γ (2) p(y|H0) ˆH=H0 2 Note that in case of an absence/presence task, where H0 is less probable, On-Off Keying (OOK) can ... |

370 |
A comparison of optimal and sub-optimal MAP decoding algorithms operating in the log domain,” in
- Robertson, Villebrun, et al.
- 1995
(Show Context)
Citation Context ...-AND-FUSE ... 2979 A similar expression was derived for MIMO beamforming and combining systems in [26]. D. Max-Log Rule Let us first recall the Max-Log approximation known from turbo-codes literature =-=[27]-=- ln ( L∑ ℓ=1 Bℓe Aℓ ) ( L∑ = ln ≈ ℓ=1 e Aℓ+ln(Bℓ) ) max ℓ∈{1,2,...,L} [Aℓ +ln(Bℓ)] (7) where Ai ∈ R and Bi ∈ R +. The approximation in Eq. (7) is accurate when one of the terms in the sum ∑L Aℓ ℓ=1 Bℓ... |

272 | On maximum-likelihood detection and the search for the closest lattice point - Damen, Gamal, et al. - 2003 |

199 |
Optimal data fusion in multiple sensor detection systems,”
- Chair, Varshney
- 1986
(Show Context)
Citation Context ...ing stages. Firstly, the decoding block computes an estimate of x, denoted ˆx in the following, from y. Finally, the global decision ˆ H is taken on the basis of ˆx using the Chair-Varshney (CV) rule =-=[28]-=-, i.e. the optimal fusion rule for noiseless channels, whose expression is given by K∑ ( PD,k ΛCV = ûk ln PF,k k=1 ) +(1− ûk)ln ( 1 − PD,k 1 − PF,k (8) ) , (9) where ûk � ˆxk+1 2 , k ∈ K. Note that wh... |

56 | Channel-aware distributed detection in wireless sensor networks,”
- Chen, Tong, et al.
- 2006
(Show Context)
Citation Context ... to the low channel SNR. This implies that when afixed(low)PF0is imposed, as in typical Neyman-Pearson test [24], the MRC is a more attractive choice than EGC, in juxtaposition with the PAC case [2], =-=[3]-=-. The difference comes from the impossibility of vector z in Eq. (5) to sum the contributions of all the sensors coherently. PD0 vs (SNR)dB: In Fig. 4 we show, for the presented rules, PD0 as a functi... |

55 | Equal gain transmission in multipleinput multiple-output wireless systems,”
- Love, Heath
- 2003
(Show Context)
Citation Context ...(6)CIUONZO et al.: CHANNEL-AWARE DECISION FUSION IN DISTRIBUTED MIMO WIRELESS SENSOR NETWORKS: DECODE-AND-FUSE ... 2979 A similar expression was derived for MIMO beamforming and combining systems in =-=[26]-=-. D. Max-Log Rule Let us first recall the Max-Log approximation known from turbo-codes literature [27] ln ( L∑ ℓ=1 Bℓe Aℓ ) ( L∑ = ln ≈ ℓ=1 e Aℓ+ln(Bℓ) ) max ℓ∈{1,2,...,L} [Aℓ +ln(Bℓ)] (7) where Ai ∈ ... |

46 | Fusion of decisions transmitted over Rayleigh fading channels in wireless sensor networks,”
- Chen, Varshney
- 2006
(Show Context)
Citation Context ... discuss recent related work and focus on results needed for the work in this paper. Sub-optimal rules for PAC scenario, presenting only issues (i) and (ii), were designed in [1], [2], [4], [8]–[10], =-=[14]-=-. More specifically, the optimal rule was compared to Maximum Ratio Combining (MRC), Chair-Varshney Maximum-Likelihood (CV-ML), Equal Gain Combining (EGC) and Max-Log. MRC and CV-ML fusion rules appro... |

41 |
Wireless sensors in distributed detection applications
- Chamberland, Veeravalli
- 2007
(Show Context)
Citation Context ... B. Related Work A vast literature is present on DF, still growing in the context of WSNs. Several tutorial papers and books have been published on the topic, providing extensive references [3], [11]–=-=[13]-=-. Here we briefly discuss recent related work and focus on results needed for the work in this paper. Sub-optimal rules for PAC scenario, presenting only issues (i) and (ii), were designed in [1], [2]... |

39 |
An efficient generalized sphere decoder for rank-deficient MIMO systems,”
- Cui, Tellambura
- 2005
(Show Context)
Citation Context ...alytically demonstrated (in low and/or high SNR regime) for CV-ML, Max-Log and MRC over MIMO channel, in perfect analogy to the PAC case. Also, we formulate efficient Generalized Sphere Decoder (GSD) =-=[23]-=- implementation to tackle exponential complexity of the first two rules. • We show that CV-MMSE outperforms CV-ML in low-tomoderate SNR regime, since it accounts for the correlation among decisions in... |

36 |
Decentralized detection over multiple-access channels,”
- Duman, Salehi
- 1998
(Show Context)
Citation Context ...d-Ratio (LLR) in [4]. The rules considered in [9] have also been derived and compared in the context of sensors differential encoding [8]. DF in distributed detection over MAC was first considered in =-=[15]-=-, where an algorithm for the design of local-sensor quantizers was derived for the optimal fusion rule. Also, several advanced transmission schemes have been proposed for MAC. In [16], [17] a fully-lo... |

36 | Asymptotic detection performance of type-based multiple access in sensor networks,”
- Mergen, Naware, et al.
- 2005
(Show Context)
Citation Context ...Multiple Access (TBMA), has been proposed for distributed detection in [20], [21]. The scheme relies only on statistical CSI and presents limited bandwidth requirements. However, it has been shown in =-=[22]-=- that TBMA suffers from significant loss of performances in i.i.d. zero-mean fading channels. DF over MIMO channels was firstly proposed in [7], focusing on J-Divergence optimal power allocation under... |

30 | Fusion of censored decisions in wireless sensor networks,” - Jiang, Chen - 2005 |

21 | Sensing reality and communicating bits: A dangerous liaison,” - Gastpar, Vetterli, et al. - 2006 |

20 | Soft-input soft-output single tree-search sphere decoding.
- Studer, Bolcskei
- 2010
(Show Context)
Citation Context ... (xk|H1) k=1 (12) The computation of Eq. (12) can be easily performed through a double search with GSD (one for each hypothesis) or with a more efficient single search, following the same approach in =-=[30]-=-. In both cases the complexity of Max-Log is always higher than CV-ML, that is n1 >n2. Detailed results on the complexity reduction deriving from the GSD implementations of minimum distance searches c... |

19 |
Fusion of decisions transmitted over fading channels in wireless sensor networks
- Chen, Jiang, et al.
- 2002
(Show Context)
Citation Context ...sensor communicates through a parallel access channel (PAC), which has to be implemented through time, code or frequency division schemes, since the wireless channel is “naturally” a broadcast medium =-=[1]-=-–[4]. Recently it has been suggested to exploit the wireless medium as a multiple-access channel (MAC) for DF while coping with presence of intrinsic interference [5]. Furthermore, the problem of deep... |

16 | Asymptotically optimal decentralized type-based detection in wireless sensor networks
- Liu, Sayeed
- 2004
(Show Context)
Citation Context ...he basis only of statistical CSI. An alternative and novel communication scheme, based on the method of types, called Type-Based Multiple Access (TBMA), has been proposed for distributed detection in =-=[20]-=-, [21]. The scheme relies only on statistical CSI and presents limited bandwidth requirements. However, it has been shown in [22] that TBMA suffers from significant loss of performances in i.i.d. zero... |

15 | Optimal Power Allocation for Distributed Detection Over MIMO Channels in Wireless Sensor Networks.
- Zhang, Poor, et al.
- 2008
(Show Context)
Citation Context ...DF while coping with presence of intrinsic interference [5]. Furthermore, the problem of deep fading was addressed with multiple antennas at the DFC in order to ameliorate the fusion performances [5]–=-=[7]-=-. This choice demands only further complexity on DFC side and does not affect simplicity of sensors implementation. The result is a communication over a “virtual” Multiple-Input Multiple-Output (MIMO)... |

14 |
PAC vs. MAC for decentralized detection using noncoherent modulation,”
- Berger, Guerriero, et al.
- 2009
(Show Context)
Citation Context ...D0(γ)] P (H1)+PF0(γ)P (H0)} , (14) γ with Λ representing the decision statistics of a generic fusion rule. In the following figures, for comparison purposes, we report the (upper) “observation bound” =-=[5]-=-, i.e. the optimum performances over noiseless channel, given by: P obs D0 = P obs F0 = K∑ i=Kγ K∑ i=Kγ ( K i ) (PD) i (1 − PD) K−i , (15) ( ) K (PF ) i i (1 − PF ) K−i . (16) where Kγ is a discrete t... |

13 |
Distributed Detection and Data Fusion, 1st ed
- Varshney
- 1996
(Show Context)
Citation Context ...able. B. Related Work A vast literature is present on DF, still growing in the context of WSNs. Several tutorial papers and books have been published on the topic, providing extensive references [3], =-=[11]-=-–[13]. Here we briefly discuss recent related work and focus on results needed for the work in this paper. Sub-optimal rules for PAC scenario, presenting only issues (i) and (ii), were designed in [1]... |

11 |
Nonorthogonal transmission and noncoherent fusion of censored decisions,”
- Yiu, Schober
- 2009
(Show Context)
Citation Context ...oding and DF rule, are compared with the optimum. Also, in [18] a CDMA protocol is considered, but no decision is performed locally by the sensors, since they employ an amplify-and-forward scheme. In =-=[19]-=-, the idea of the direct sequence spread spectrum is exploited to perform a communication based on the On-Off Keying (OOK) modulation with censoring; at the DFC the optimal rule and sub-optimal altern... |

10 | Large system decentralized detection performance under communication constraints,”
- Jayaweera
- 2005
(Show Context)
Citation Context ... is proposed: DF exploiting instantaneous CSI is performed and sub-optimal alternatives, based on the partitioning of transmitted symbols decoding and DF rule, are compared with the optimum. Also, in =-=[18]-=- a CDMA protocol is considered, but no decision is performed locally by the sensors, since they employ an amplify-and-forward scheme. In [19], the idea of the direct sequence spread spectrum is exploi... |

9 | Distributed detection over fading MACs with multiple antennas at the fusion center
- Banavar, Smith, et al.
- 2010
(Show Context)
Citation Context ...@unina2.it). Digital Object Identifier 10.1109/TWC.2012.061912.112049 1536-1276/12$31.00 c○ 2012 IEEE Fig. 1: The Decision Fusion model in presence of a (virtual) MIMO channel. (CSI) is often assumed =-=[6]-=-, [7]. Design of channel-aware fusion rules and corresponding high performances motivates this assumption [2], [4], [8], [9], fulfilled in many scenarios 1 .For the same reason, channel-aware fusion r... |

9 | Bayesian fusion performance and system optimization for distributed stochastic gaussian signal detection under communication constraints,” Signal Processing,
- Jayaweera
- 2007
(Show Context)
Citation Context ... the results observed. We consider a WSN with N = 1, 2 (corresp. Fig. 6 and 7) antennas at DFC and a channel (SNR)dB =15. The latter assumption clearly represents a total power constraint (TPC) [18], =-=[31]-=- on the WSN, i.e. the individual channel SNR for the kth sensor is scaled as SNRk = SNR K . We also report the (upper) “communication” bound [5], i.e. PE0 under the assumption of ideal sensors, i.e. (... |

4 | Multiple-symbol differential decision fusion for mobile wireless sensor networks
- Lei, Schober
(Show Context)
Citation Context ...) is difficult due to exponential functions with large dynamic range especially for moderate-to-high channel SNRs KN/σ 2 w ≫ 1. This becomes a quite severe requirement for fixed point implementations =-=[8]-=-, [9], [19]. All the proposed sub-optimal rules instead present numerical stability for realistic SNR values, although they require a different degree of system knowledge and they also differ in compu... |

3 |
Optimal bayesian data fusion and lowcomplexity approximations for distributed DS-CDMA wireless sensor networks in Rayleigh fading,”
- Jayaweera
- 2005
(Show Context)
Citation Context ...st considered in [15], where an algorithm for the design of local-sensor quantizers was derived for the optimal fusion rule. Also, several advanced transmission schemes have been proposed for MAC. In =-=[16]-=-, [17] a fully-loaded code-division multiple access (CDMA) communication between sensors and the DFC is proposed: DF exploiting instantaneous CSI is performed and sub-optimal alternatives, based on th... |

2 |
distributed detection strategies for wireless sensor networks
- “Optimal
- 2004
(Show Context)
Citation Context ...is only of statistical CSI. An alternative and novel communication scheme, based on the method of types, called Type-Based Multiple Access (TBMA), has been proposed for distributed detection in [20], =-=[21]-=-. The scheme relies only on statistical CSI and presents limited bandwidth requirements. However, it has been shown in [22] that TBMA suffers from significant loss of performances in i.i.d. zero-mean ... |

1 |
aware decision fusion in wireless sensor networks
- “Channel
- 2004
(Show Context)
Citation Context ... The Decision Fusion model in presence of a (virtual) MIMO channel. (CSI) is often assumed [6], [7]. Design of channel-aware fusion rules and corresponding high performances motivates this assumption =-=[2]-=-, [4], [8], [9], fulfilled in many scenarios 1 .For the same reason, channel-aware fusion rules for coherent, noncoherent, and differential modulation were already proposed for PAC [2], [4], [8]–[10].... |

1 |
Robust suboptimal decision fusion in wireless sensor networks
- Jiang, Misra, et al.
- 2005
(Show Context)
Citation Context ...or communicates through a parallel access channel (PAC), which has to be implemented through time, code or frequency division schemes, since the wireless channel is “naturally” a broadcast medium [1]–=-=[4]-=-. Recently it has been suggested to exploit the wireless medium as a multiple-access channel (MAC) for DF while coping with presence of intrinsic interference [5]. Furthermore, the problem of deep fad... |

1 |
Max-Log decision fusion in wireless sensor networks
- “Coherent
- 2010
(Show Context)
Citation Context ...usion model in presence of a (virtual) MIMO channel. (CSI) is often assumed [6], [7]. Design of channel-aware fusion rules and corresponding high performances motivates this assumption [2], [4], [8], =-=[9]-=-, fulfilled in many scenarios 1 .For the same reason, channel-aware fusion rules for coherent, noncoherent, and differential modulation were already proposed for PAC [2], [4], [8]–[10]. Unfortunately,... |

1 |
Improved low-complexity fusion receivers for synchronous DS-CDMA sensor networks
- Dyer, Natarajan
- 2007
(Show Context)
Citation Context ...sidered in [15], where an algorithm for the design of local-sensor quantizers was derived for the optimal fusion rule. Also, several advanced transmission schemes have been proposed for MAC. In [16], =-=[17]-=- a fully-loaded code-division multiple access (CDMA) communication between sensors and the DFC is proposed: DF exploiting instantaneous CSI is performed and sub-optimal alternatives, based on the part... |