A Multi-Channel Partial-Update Algorithm for Sea Clutter Suppression in Passive Bistatic Radar China Academy of Electronics and Information Technology Speraker：Ma Yahui 2018.7

1 INTRODUCTION 2 SIGNAL MODEL 3 2D-PU-NLMS ALGORITHM 4 REAL-DATA EXPERIMENT 5 CONCLUSION

One feasible solution: 1. INTRODUCTION Problem: Sea clutters with Doppler-varying spectrum exert a notable negative impact on the detection performance, when a PBR is employed to detect sea-surface targets with low-velocity targets. One feasible solution: Conventional 1D (Range dimension) adaptive filters (such as NLMS) modulate the reference signal onto the Doppler dimension Multi-channel NLMS (MC-NLMS or 2D-NLMS) The added fully-updating counterparts in Doppler dimension induces high computational load partial-update NLMS (PU-NLMS) Our work: PU-NLMS+ 2D-NLMS Sea clutters Passive bistatic radars (PBRs) exploit the transmitted signals from non-cooperative illuminators of opportunity to perform detection and tracking of maneuvering targets, and thus has unique advantages in covert operation, immunity from various electronic countermeasures，and low cost compared with conventional active radars. PU-NLMS algorithms can reduce the computational complexity while yield approximate results with their fully-updating counterparts in terms of convergence rate and mean-squared error (MSE)

2 1 INTRODUCTION SIGNAL MODEL 3 2D-PU-NLMS ALGORITHM 4 REAL-DATA EXPERIMENT 5 CONCLUSION

2. SIGNAL MODEL Reference signal Received signal CAF: Noise the time delay and Doppler of a target in the PBR are extracted by examining the peaks of the CAF between the reference signal and the residual signal that contains the desired echo signal after clutter cancelation.

Diffcult to be cancelled 2. SIGNAL MODEL Targets Noise Received signal Direct signal Sea clutters 1D adaptive filters Wiener NLMS RLS …. Diffcult to be cancelled Passive bistatic radars (PBRs) exploit the transmitted signals from non-cooperative illuminators of opportunity to perform detection and tracking of maneuvering targets, and thus has unique advantages in covert operation, immunity from various electronic countermeasures，and low cost compared with conventional active radars.

1 INTRODUCTION 2 SIGNAL MODEL 3 2D-PU-NLMS ALGORITHM 4 REAL-DATA EXPERIMENT 5 CONCLUSION

3. 2D-PU-NLMS ALGORITHM 1D adaptive filter Multi-channel(2D) adaptive filter modulate the reference signal onto the Doppler dimension a filter with a wide notch and sharp edges The added fully-updating counterparts in Doppler dimension induces high computational load

3. 2D-PU-NLMS ALGORITHM Multi-channel(2D) NLMS Multi-Channel Partial-Update(2D-PU) NLMS Only select L out of N coefficients to update

Computational complexity 3. 2D-PU-NLMS ALGORITHM Computational complexity MC-NLMS KN complex multiplications are required for calculating one output point, and 2KN+K complex multiplications for updating K weight vectors is needed in MC-NLMS for one single iteration. 2D- Partial-Update (L<N) 2KL+K(N+1) complex multiplications are required in the 2D-PU-NLMS algorithm. MC-NLMS 2D-PU-NLMS 3KN+K 2KL+KN+K

1 INTRODUCTION 2 SIGNAL MODEL 3 2D-PU-NLMS ALGORITHM 4 REAL-DATA EXPERIMENT 5 CONCLUSION

4. REAL-DATA EXPERIMENT Para. value The carrier frequency 714 MHz bandwidth of the DVB 7.56 MHz baseband sampling rate 10 MHz. filter length 1000 The modulated frequencies stepsize It can be observed that the 2D-PU-NLMS has approximate results with the fully-updating counterparts in terms of convergence rate and mean-squared error (MSE) an experimental digital video broadcast (DVB)-based PBRwas installed on the roof of a building located in an eastern coastal city of China, Yantai. Learning curves for NLMS algorithm, MC-NLMS algorithm and 2D-PU-NLMS But little performance loss can achieve satisfactory performance in terms of the convergence rate and stability with a reduced computational complexity using fewer updated points.

4. REAL-DATA EXPERIMENT CAF results using NLMS algorithm   MC-NLMS 2D-PU-NLMS N=1000 L=600 L=500 L=300 SCNR(dB) 14.92 14.69 14.53 14.31 The desirable capability in clutter suppression as well as the reduced computational complexity can be both achieved by utilizing 2D-PU-NLMS algorithm CAF results using 2D-PU-NLMS algorithm with L=500

1 INTRODUCTION 2 SIGNAL MODEL 3 2D-PU-NLMS ALGORITHM 4 REAL-DATA EXPERIMENT 5 CONCLUSION

5. CONCLUSION a novel 2D-PU-NLMS algorithm is proposed, which is derived by extending the NLMS filter to a 2-D structure that accounts for both time delay and Doppler frequency. Compared with the state-of-the-art MC-NLMS filter, the proposed 2D-PU-NLMS is able to reduce the computational complexity while yields approximate interference cancelation performance in terms of convergence rate and MSE. The effectiveness of the proposed algorithm is verified using real radar data.