1. 1 Blind Channel Identification and Equalization in Dense Wireless Sensor Networks with Distributed Transmissions Xiaohua (Edward) Li Department of Electrical and Computer Engineering State University of New York at Binghamton xli@binghamton.edu http://ucesp.ws.binghamton.edu/~xli

2. 2 Outline Introductions on sensor network and blind equalization Cross-correlation-based blind equalization: a cooperative communication approach Cross-correlation and finite sample properties Simulations Conclusions

3. 3 1.1 Introduction: Sensor Network Wireless sensor network: dense, cooperative Sensor data and transmitted signals: highly cross-correlated Cooperation: enhance cross-correlation Multi-hop Wireless Sensor Network

4. 4 1.2 Introduction: Blind Equalization Blind channel identification and equalization in sensor networks –Mitigate multipath fading, inter-symbol interference –Remove training: save transmission energy and bandwidth, design convenience Especially helpful in wideband sensor networks, e.g., acoustic, video Need to compete with training-based methods in computational efficiency and robustness –Traditional blind methods not desirable –Need new blind methods

5. 5 1.3 Cooperative Equalization Observe: –Traditional blind methods: signals from different users are un-correlated –Sensor networks: signals among sensors are highly cross-correlated –Can we utilize cross-correlation to assist blind equalization? A new way of blind equalization based on cooperative communications –Passive cooperation: transmitting nodes do not cross- talk –Useful for general distributed networks

6. 6 2.1 System Model Sensor network Transmission block diagram of each sensor

7. 7 2.2 Cross-Correlation Assumption Source sequence cross-correlation  symbol sequence cross-correlation –By scrambling, cross-correlation among transmitted signals becomes highly structural: only one non- zero cross-correlation coefficient –Result: efficient/robust blind algorithms

8. 8 2.3 Received Signal Model A receiving node receives un-overlapped signals from transmitting sensors

9. 9 2.4 Blind Channel Estimation Computationally efficient, robust to ill-conditioned channels, optimal utilization of all received signals

10. 10 2.5 Blind Equalization Computationally efficient (linear), robust to ill- conditioned channels, fast convergence

11. 11 3.1 Cross-Correlation Property Find relation between (analog) source signal cross- correlation and (digitized) binary sequence cross- correlation Major results and simulation verification

12. 12 3.2 Finite Sample Effect Samples may be limited, samples contributing to cross-correlations are even more limited Find the relation among symbol amount, cross- correlation, and channel estimation MSE

13. 13 4.1. Channel Estimation Simulation Short Data Record Proposed: J=10 sensors. One packet (260 symbols) Training : 20% symbols for training Proposed blind method has near-training performance

14. 14 4.2 Blind Equalization

15. 15 4.3 Blind Channel Estimation Long Data Record Proposed: 10 sensors. 20dB SNR. 260 symbols/packet Training : 20% symbols for training New algorithms both have near-training performance

16. 16 4.4 Blind Equalization

17. 17 4.5 Convergence Property

18. 18 5. Conclusions Propose a new blind channel identification and equalization scheme for wireless sensor networks –Utilize cross-correlation among sensor signals –Have near-training performance, computation efficiency, and robustness to ill-conditioned channels A general approach of exploiting (passive) cooperative communications in distributed networks