Research On Doa Estimation And Tracking Approach Based On Monte Carol Method

The direction-of-arrival (DOA) estimation of any signal of interest is an important aspect of signal processing which finds a wide range of application in radar, communication, sonar and other national and military area. However, the classical DOA estimation methods can only deal with static DOA and the precision and resolution needs to be increased. As a result, this paper began with the data model of array signal, developed the estimation and tracking model based on Bayesian theorem, solved the model using Mental Carle method and particle filter and finally settled the problem of DOA estimation and tracking. The main results and achievements can be summarized as follows:1. The data model of array signal was studied in this paper. On the basis of existing array signal model, the model constructed in time domain was mainly studied, which is applicable to wideband signals as well as narrow band signals. The relationship between model error and bandwidth as well as time range was discussed and the model was extended to any array set and time varying DOA condition. Theoretical analysis shows that the error increases as the bandwidth expands2. Mental Carle sampling method was studied deeply. Based on the existing Monte Carlo sampling methods including dependent Markov chain Metropolis-Hastings (MH) algorithm and random walks MH algorithm, a new mixed sampling method was presented and its constringency was proved. The new method constructs evaluation function according to the samples and the single sample method is selected adaptively dependent on it. Theoretical analysis shows the new method can improve the convergence rate and estimation precision.3. Particle filter was studied deeply and improved. In order to realized joint source number and DOA tracking, RJMCMC(Reversible Jump Markov Chain Monte Carlo) was introduced and improved. The improved RJMCMC can avoid the confusion of DOA sequence. Then, the improved RJMCMC was introduced to particle filter and improved the performance of particle filter. The improved particle filter can not only realize joint state and model order estimation, but also avoid sequence confusion.4. The Bayesian DOA estimation approach based on the new mixed sampling method was studied, which can improve the convergence rate, resolution and precision. Based on the array signal data model and Bayesian theorem, the posterior of unknown parameters was demonstrated, the irrespective parameter was integrated and finally the posterior of DOA was got. Then, taken the posterior parameter as desired distribution of markov chain, samples was largely produced using mixed sampling method. Finally, we got DOA estimation using these samples. Theoretical analysis and simulation results show that the method presented in this paper can greatly reduce calculation burden of Bayesian estimation, improve convergence rate and precision, and can deal with correlative signal.5. The DOA tracking approach based on the improved particle filter was studied which can realize joint source number and DOA tracking as well as integrated DOA tracking and beamforming. Based on the array signal data model, the tracking equation was established by incidence signal estimation. Then, the improved particle filter was used , the particles was updated in every iteration, and finally the time varying DOA and source number was estimated. Theoretical analysis and simulation results show that method presented in this paper can realize joint source number and DOA tracking and tracking precision is close to CRLB on the high SNR. Besides, it can realize integrated DOA tracking and beam forming.