AU2012241097A1 - A data fusion method for distributed sensor network - Google Patents
A data fusion method for distributed sensor network Download PDFInfo
- Publication number
- AU2012241097A1 AU2012241097A1 AU2012241097A AU2012241097A AU2012241097A1 AU 2012241097 A1 AU2012241097 A1 AU 2012241097A1 AU 2012241097 A AU2012241097 A AU 2012241097A AU 2012241097 A AU2012241097 A AU 2012241097A AU 2012241097 A1 AU2012241097 A1 AU 2012241097A1
- Authority
- AU
- Australia
- Prior art keywords
- statistics
- local
- local sensors
- sensors
- fusion
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Abandoned
Links
Landscapes
- Testing Or Calibration Of Command Recording Devices (AREA)
Abstract
Abstract A data fusion method for distributed sensor networks includes: receiving and storing N pulse data by local sensors; calculating R statistics based on the stored data by local sensors; transmitting R statistics to the fusion center by local sensors; calculating weighted R statistics based on local R statistics from local sensors; making detection decision based on weighted R statistics.The invention improves the system performance when the SNR gap of local sensors is large. TStart 210 Local sensors receive and store N pulse data 220 Local sensors calculated R statistics based on the stored data 230 Local sensors transmit R statistics to the fusion center. 240 Fusion center calculated global R statistics based on local R statistics from local sensors 250 Fusion center make detection decision based on global R statistics end Fig. 2
Description
1 A data fusion method for distributed sensor networks Field of the invention [0001] The present invention relates to a method for improving the system performance in distributed sensor networks, and more particularly, related to a distributed system which is composed of several local sensors and a fusion center. Background [0002] The deployment of multiple sensors for signal detection in an intelligent transportation application may substantially enhance system survivability. Due to the lack of the communication capacities available in early distributed systems, decisions were made in the local sensors and only the binary results were sent to the fusion center. This is also called the decentralized detection. Each local sensor transmits its binary decision to the fusion center where the global decision is obtained based on k-out-of-n fusion rule. However, the optimal fusion rule requires the prior knowledge of the probability of false alarm (PFA) and the probability of detection (Pd) at each local sensor, which is not always available in practice. To remove this restriction, a blind adaptive decision fusion was developed. It uses the estimated local PFA and Pd instead of true values. When the environment is stationary, the adaptive fusion can obtain performance similar to that of the optimal fusion. [0003] If the channel capacities between local sensors and the fusion center are large enough to allow local sensors to transmit nonbinary decisions or local statistics to the fusion center, better performance may be achieved. In such a system, the observations available at each local sensor are quantized to produce a multiple digit decision which is sent to the fusion center. The Sum rule performs as well as the optimum rule when SNRs happened to be identical in local sensors, and only has a tolerable loss when local SNRs differ significantly. However, when the SINRs of local sensors are different, the performance degradation of Sum rule will occur.
2 Summary of the invention [0004] A data fusion method for distributed sensor network is disclosed, in order to improve performance when the SNRs of local sensors are different. [0005] This system involves multiple distributed sensors spatially and observing the same range cell. All local sensors are linked to a central processor which fuses data received from all sensors. The observations at the local sensors are assumed to be conditionally independent. [0006] The steps of the invention are as follows: A. Local sensors receive and store N pulse data. B. Local sensors calculated local R statistics based on the stored data. C. Local sensors transmit local R statistics to the fusion center. D. Fusion center calculated global R statistics based on local R statistics from local sensors E. Fusion center make detection decision based on global R statistics [0007] In step A, suppose xi, is the data for ith sensor and nth pulse. [0008] In step B, the R statistics for ith sensor, ,., is calculated by: i N 2= Where Nis the pulse number and a,2 is the noise power. [0009] In step D, the weighted R statistics is calculated by weighting local R statistics by itself, which is as follows: L R = 52 1=1 Where L is the sensor number. [0010] In step E, the fusion center compares global R statistics with a threshold. If R is larger than the threshold, a target is detected.
3 Brief description of the drawings [0011] Fig.1 is a schematic diagram of the distributed detection system in Parallel. [0012] Fig.2 is a schematic diagram of the data fusion method process [0013] Fig.3 is a schematic diagram of the performance of 10 pulses accumulation [0014] Fig.4 is a schematic diagram of the performance of 50 pulses accumulation Detailed description of the invention [0015] This invention is further exemplified with the use of the figures and the embodiments. [0016] A typical parallel network is shown in Fig.1. This system involves L sensors distributed spatially and observing the same range cell. LPi ... LPN which are denoted as 120 are local sensors. All sensors are linked to a central processor 110 which fuses data received from all sensors. The observations at the local sensors are assumed to be conditionally independent and the observation of the ith local sensors is denoted by x (i=1,..., L). Suppose xi, is the data for ith sensor and nth pulse. Thus the joint probability density function (PDF) ofxI, X2, ..., xL is in the following form. L f(x,x 2 ,...,XL)=1f1f(xI) i-1 [0017] Each local sensor forms the local result ui basing on its own observation xi and transmits ui to the fusion center. Then, the fusion center combines all ui, which can be binary decisions, local statistics or all raw observations data, to yield the global decision uo. [0018] The target echo is embedded in a white complex Gaussian noise whose power is unknown. Then the detection problem in the ith sensor is: x,=n, Ho x=s,+n, HI 4 Where ni, si is the noise and target signal at the ith local sensor respectively. [0019] According to the above assumptions, the conditional PDF of the single observation xi at the ith local sensor is f(xIHH)1= exp(-X) 2~i-o7 2-~i=1,2,..L 1 exp(- Xi (x, | H ) = 2o- +,) 2 where o.2 and A2 represent the noise power and SNR at the ith local sensor, respectively. [0020] The method process is shown in Fig.2. In step 210, Local sensors receive and store N pulse data. [0021] In step 220, local sensor calculated local R statistics r based on received data: I N -2 YZ(X" Xi*,) -i _ [0022] In step 230, local sensors send t to the fusion center. [0023] In step 240, the global R statistic is calculated by the fusion center, which is called Wsum(weighted sum) method: L R = Y 2 1=1 Where L is the number of local sensors. [0024] the Wsum fusion rule, which weights these local R statistics differently, is to solve the problem that detection performance drops. This extension is important especially in case of multiple pulses for the increased difference between local SNRs. This new rule does not sum up the R statistic directly but weights them by themselves. L Hi RW=Zr 2 T =
HO
5 where T is the threshold decided by the desired PFA of the system. [00251 The Wsum rule changes the linear fusion of R statistics to the quadratic fusion, which enlarges the gap of local R statistics, thus making the bigger R statistics acquire more importance. With the reasonable assumption that local sensors with higher SNR will send bigger R statistics, we could expect Wsum rule outperform the Sum fusion when local SNRs are quite different. Besides, the Wsum rule doesn't need the local SNRs either, so it is also a nonparametric fusion rule and can be used conveniently. [0026] In step 250, the fusion center compares the R with a threshold. The threshold is decided by the desired PFA of the system. If R is larger than the threshold, a target is detected. Because the Wsum rule is quartic power of observation and the SNR is the ratio of quadratic power of target signal to noise power. We define the total SNR of the Wsum rule as follows: E(rw,,um 1, L 1 L SNu= Er_|H)-1= - (1+2j-12 -Z_ E(rwm Ho) L L i [0027] The Wsum rule can provide higher total average SNR than that given by the sum fusion rule. [0028] Now we consider a parallel networks involving 3 local sensors. Other conditions are the same as before. The pulse number in Fig.3 and Fig.4 are 10 and 50 respectively. Fig.3 shows the performance of the two rules with Ai=2db and 22=3. The improvement of the Wsum fusion rule over the Sum fusion rule is prominent for 22 and 23 both being low. In Fig.4, the similar results to the case of two sensors with 50 pulses are depicted. [0029] From the above simulation results, it can be seen that when there's significant difference among local SNRs, the Sum rule degrades a lot. The Wsum method weights the R statistics with itself, so it emphasizes the use of the data with higher SNRs and mitigates the negative influence of local statistics with lower SNRs. [0030] What is stated above is just an example of the embodiments of this invention, and is not meant to be restricting on the scope of the present invention. Any equivalent modification or 6 adjustment of the scope of the invention falls within the scope of the present invention.
Claims (5)
1. A data fusion method for distributed sensor networks, comprising: Receiving and storing N pulse data by local sensors; Calculating R statistics based on the stored data by local sensors; Transmitting R statistics to the fusion center by local sensors; Calculating weighted R statistics based on local R statistics from local sensors; Making detection decision based on weighted R statistics.
2. The method of claim 1, wherein distributed sensor networks includes several local sensors and one fusion center which makes global decision based on the input of local sensors.
3. The method of claim 1, wherein local sensor calculates R statistics based on the stored data by local sensors according to the following formula: I N -=2 YZ(xi'x4) Ui n-l xin is the data for ith sensor and nth pulse, N is the pulse number, 5 is the R statistics of ith sensor. o2 is the noise power of the ith sensor.
4. The method of claim 1, wherein the weighted R statistics is calculated by weighting local R according to the following formula: L R =yr2 1=1 Where L is the sensor number.
5. The method of claim 1, wherein Fusion center making detection decision comprises: Comparing global R statistics with a threshold. If R is larger than the threshold, a target is detected.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
AU2012241097A AU2012241097A1 (en) | 2012-10-14 | 2012-10-14 | A data fusion method for distributed sensor network |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
AU2012241097A AU2012241097A1 (en) | 2012-10-14 | 2012-10-14 | A data fusion method for distributed sensor network |
Publications (1)
Publication Number | Publication Date |
---|---|
AU2012241097A1 true AU2012241097A1 (en) | 2014-05-01 |
Family
ID=50548837
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
AU2012241097A Abandoned AU2012241097A1 (en) | 2012-10-14 | 2012-10-14 | A data fusion method for distributed sensor network |
Country Status (1)
Country | Link |
---|---|
AU (1) | AU2012241097A1 (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104778094A (en) * | 2015-04-09 | 2015-07-15 | 北京羽乐创新科技有限公司 | Fused data error-correcting method and device |
-
2012
- 2012-10-14 AU AU2012241097A patent/AU2012241097A1/en not_active Abandoned
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104778094A (en) * | 2015-04-09 | 2015-07-15 | 北京羽乐创新科技有限公司 | Fused data error-correcting method and device |
CN104778094B (en) * | 2015-04-09 | 2018-11-23 | 北京羽乐创新科技有限公司 | A kind of fused data error correction method and device |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN103929259B (en) | A kind of many bit decision cooperation adaptive spectrum cognitive methods based on degree of belief in cognitive ofdm system | |
CN104253606B (en) | Receiving circuit | |
Gao et al. | On remote estimation with multiple communication channels | |
Zheng et al. | Cooperative spectrum sensing based on SNR comparison in fusion center for cognitive radio | |
Gao et al. | A robust and efficient cooperative spectrum sensing scheme in cognitive radio networks | |
AU2012241097A1 (en) | A data fusion method for distributed sensor network | |
Liu et al. | An adaptive double thresholds scheme for spectrum sensing in cognitive radio networks | |
Nurellari et al. | Optimal quantization and power allocation for energy-based distributed sensor detection | |
Mitsuya et al. | Improvement of sequential-test-based cooperative spectrum sensing systems in band limited control channels | |
Althunibat et al. | Energy-efficient reporting scheme for cooperative spectrum sensing | |
Hu et al. | Minimum transmission delay via spectrum sensing in cognitive radio networks | |
TWI695598B (en) | Adaptive device and method for wireless network | |
CN110677204B (en) | Frequency spectrum sensing time optimization method with intelligent interference | |
Alam et al. | Optimizations of cooperative spectrum sensing with reporting errors over myriad fading channels | |
Sreedharan et al. | Spectrum sensing using distributed sequential detection via noisy reporting mac | |
Ganguly et al. | A new algorithm for distributed nonparametric sequential detection | |
Du et al. | Performance evaluation for the two-stage cooperative spectrum sensing scheme in cognitive radio | |
US11808916B2 (en) | Precipitation monitoring using point to point radio links | |
Song et al. | Multi-sensor-based aperiodic least-squares estimation for networked systems with transmission constraints | |
CN107078978B (en) | Signal compensation method and device | |
Zhao et al. | Prior probability-aided secure cooperative spectrum sensing | |
Gao et al. | Distributed detection with quantization based on Neyman-Pearson criterion | |
JPH05145445A (en) | Automatic equalizer for digital microwave communication system | |
KR20110129592A (en) | Apparatus for cognitive radio and method thereof | |
Yu et al. | On optimal operating characteristics of sensing and training for cognitive radios |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
MK1 | Application lapsed section 142(2)(a) - no request for examination in relevant period |