Research On Frequency Estimation Algorithm For Sinusoidal Signal Under Strong Noise

Extracting the frequency of sinusoidal signal accurately in strong noise is a fundamental and well-studied problem in signal processing and communications. Currently,under strong noise background sinusoidal signal frequency estimation has been successfully applied in radar detection, speech signal processing, sonar targeting, signal recovery in communication systems, the bridge vibration detection ,biomedical detection,as well as electronic communications technology. Therefore, the research of sinusoidal signal frequency estimation has grate theoretical significance and application value.Frequency is the most important parameter and the most essential characteristic of sinusoidal signal, frequency estimation of signal processing is a classic issue. We analyzed Several types of sinusoidal signal frequency estimation algorithms in this paper, including classical DFT algorithm maximum likelihood estimation, PHD algorithm, MC algorithm,TSA algorithm, RIFE Fourier coefficients interpolation iterative algorithm, and compared these algorithms in theory and computation respectively, Simulation results showed the relationship between frequency estimation mean square error and SNR and compared the mean square error with Rao-Cramer lower bound. Finally, basing on the analysis and summarizes of various algorithms, we made corresponding improvements. The main research work and conclusions are as follows:1. Proposing a new algorithm which is based on the segmented nature of sinusoidal signal LP auto-correlation frequency estimation. The algorithm solves the shortcomings which estimation performance and computation can't be taken into account in the two-step self-correlation frequency estimation algorithm (TSA algorithm), the performance of the algorithm approximation TSA2 in the case of a small amount of calculation, the problem of the increasing amount of calculation after the bringing up of the TSA1. Then, we conducted the simulation experiments with the segmented auto-correlation algorithm and used results to analyze the effectiveness of the algorithm.2. Proposing a FFT-based sinusoidal signal frequency estimation algorithm. By analyzing the performance of rife algorithm and Fourier coefficients iterative algorithm, we find that rife algorithm is simple but accuracy is not high, the Fourier coefficients iterative algorithm needs two iterations to meet accuracy requirements, and each iteration need heavy computational effort. Combining with the advantages and disadvantages of these two algorithms, an improved high precision frequency estimation algorithm is presented in this paper. and take simulations and experiments on this algorithm,the results demonstrate that the property of this method is relative good.