Research On Harmonic Frequency Estimation Based On Non-parameter Methods

Frequency estimation of harmonic signals is an important part of modern signal processing in the information science and is widely used in the research of radar, sonar, space navigation, mobile communications, voice recognition, biomedicine, vibration engineering, remote sensing, geography reconnaissance and spectroscopy, etc. It is getting more and more important with the development of industry, agriculture, national defence and science. It attracts much attention from many researchers in the world. Its development can reflect the science and technology level of a country to some extent.There are many methods for harmonic frequency estimation, such as parameter methods based on a linear modeling, nonparameter methods based on a non-modeling and entropy spectra estimation, etc. All the methods have their own advantages. For example, parameter methods have higher frequency estimation precision for a shorter numeric sequence with higher Signal-to-Noise, Fourier analysis among non-parameter methods is more widely used because of its faster estimation speed, and the modern frequency estimation based on a signal space decomposion has higher estimation precision even under the condition of lower Signal-to-Noise. This dissertation studies the technologies of harmonic frequency estimation focusing on non-parameter methods based on the harmonic signal model. Its main content includes:(1)The primary factors that affect the performance of Fourier transform are studied, then, data error explanation of spectra leackage, the non-uniform characteristic measurement of a numeric sequence and some improved harmonic frequency estimation methods such as FFT non-uniform sampling realization and a sine frequency estimation based on the FFT rotational invariance are proposed. (2)The algorithm characteristics and difference are compared definitely between ZFFT and CZT by theoretical analysis and experiment simulation. Furthermore, corresponding improved solutions are discussed for their disadvantages.(3)Some new improved frequency estimation methods such as fast MUSIC algorithm based on a FFT finite search, MTLS-ESPRIT algorithm and D-ESPRIT algorithm are proposed while discussing the fundmental theory of modern high precision frequency estimation technologies based on a signal space decomposion.(4)For MUSIC algorithm, chapter 5 discusses the harmonic frequency detection of a weak signal in noise background based on multi-layer auto correlation, noise modulation technology, FFT rotational invariance, or cross-four-order-cumulant.In all, the technology of harmonic frequency estimation is an important application science. It is summed up and renewed focusing on non-parameter estimation methods in the dissertation.