Search

An object-oriented algorithm applicable to range super-resolution estimation of linear frequency modulation continuous wave (LFMCW) car sensors is proposed. By utilizing prior information on target distribution pre-extracted by group segmentation, the complicated target estimation problem is reduced to a simple minimization problem, which is then solved by a Hopfield neural network. Analysis shows that the prior information helps to decrease computational complexity and enhance resolution. Both simulation and experimental results have demonstrated the superiority of this algorithm over other super-resolution algorithms such as MUSIC and ME(AR).

An efficient implementation of DOA estimation is proposed for DS-CDMA systems with smart antennas. We suggest that the received signal is first decoded to enhance its SNR and reduce its complexity, then a rough estimation is done to simplify the ANN structure before data is processed by a backpropagation artificial neural network (BP ANN). Simulation results show this method is more precise and stable and can endure lower SNR than many other algorithms.