Window Function And Linear Prediction Of Music Spectral Estimation Method Improvements

Among modern spectral estimation, it is commonly needed to estimate the frequency of sine wave with the disturbance from white noise. On the basis of the eigenvector decomposition of matrix, MUSIC method (Multiple Signal Classification) is broadly adopted, which possesses very high resolution. But the defect lies in its noise immunity performance has to be improved. Processing limited data, leads to the existence of the error.This thesis gives a brief introduction of the theory of spectral estimation, discusses the characteristics of the methods based on the eigenvector decomposition of matrix, such as Pisarenko and MUSIC. Then it proposes three modified method named WMUSIC, LMUSIC and LWMUSIC respectively, which combine Linear Prediction and Window function. With large numbers of emulational experiment, it proves the new method can provide better resolution, noise immunity and stability for frequency estimation than traditional MUSIC method. The result is referential to the research of spectral estimation to the sine wave signal with the disturbance from white noise.