Optimal data-based kernel estimation of evolutionary spectra

Complex demodulation of evolutionary spectra is formulated as a two-dimensional kernel smoother in the time-frequency domain. First, a tapered Fourier transform, y/sub v/(f, t), is calculated. Then the log-spectral estimate, is smoothed. As the characteristic widths of the kernel smoother increase, the bias from the temporal and frequency averaging increases while the variance decreases. The demodulation parameters, such as the order, length, and bandwidth of spectral taper and the kernel smoother, are determined by minimizing the expected error. For well-resolved evolutionary, spectra, the optimal taper length is a small fraction of the optimal kernel halfwidth. The optimal frequency bandwidth, w, for the spectral window scales as w/sup 2/ approximately lambda / tau , where tau is the characteristic time and lambda /sub F/ is the characteristic frequency scalelength. In contrast, the optimal halfwidths for the second stage kernel smoother scales as h approximately 1/( tau lambda /sub F/)/sup 1/(p+2)/ where p is the order of the kernel smoother. The ratio of the optimal-frequency halfwidth to the optimal-time halfwidth is determined. >

[1]  Ingrid Daubechies,et al.  The wavelet transform, time-frequency localization and signal analysis , 1990, IEEE Trans. Inf. Theory.

[2]  D. Thomson.  Quadratic-inverse spectrum estimates: applications to palaeoclimatology , 1990, Philosophical Transactions of the Royal Society of London. Series A: Physical and Engineering Sciences.

[3]  H. Müller,et al.  Variable Bandwidth Kernel Estimators of Regression Curves , 1987 .

[4]  Florencio I. Utreras,et al.  On Generalized Cross-Validation for Multivariate Smoothing Spline Functions , 1987 .

[5]  Mark Asch,et al.  Frequency Content of Randomly Scattered Signals I , 1990 .

[6]  L. Cohen,et al.  Time-frequency distributions-a review , 1989, Proc. IEEE.

[7]  U. Grenander,et al.  Statistical analysis of stationary time series , 1958 .

[8]  D. Slepian.  Prolate spheroidal wave functions, fourier analysis, and uncertainty — V: the discrete case , 1978, The Bell System Technical Journal.

[9]  C. L. Mallows,et al.  Linear processes are nearly Gaussian , 1967 .

[10]  Guy Melard,et al.  CONTRIBUTIONS TO EVOLUTIONARY SPECTRAL THEORY , 1989 .

[11]  W. Härdle,et al.  How Far are Automatically Chosen Regression Smoothing Parameters from their Optimum , 1988 .

[12]  H. Müller.  Nonparametric regression analysis of longitudinal data , 1988 .

[13]  I. G. Zurbenko,et al.  Spectral Analysis of Non-stationary Time Series , 1991 .

[14]  D. Cox.  MULTIVARIATE SMOOTHING SPLINE FUNCTIONS , 1984 .

[15]  W. Härdle.  Applied Nonparametric Regression , 1992 .

[16]  G. Wahba,et al.  Smoothing Spline ANOVA with Component-Wise Bayesian “Confidence Intervals” , 1993 .

[17]  F. Hlawatsch,et al.  Linear and quadratic time-frequency signal representations , 1992, IEEE Signal Processing Magazine.

[18]  G. Wahba.  Automatic Smoothing of the Log Periodogram , 1980 .

[19]  D. Thomson,et al.  Spectrum estimation and harmonic analysis , 1982, Proceedings of the IEEE.

[20]  G. Wahba.  Spline models for observational data , 1990 .

[21]  C. J. Stone,et al.  Optimal Global Rates of Convergence for Nonparametric Regression , 1982 .

[22]  W. Härdle,et al.  Applied Nonparametric Regression , 1991 .

[23]  P. Sheng,et al.  FREQUENCY CONTENT OF RANDOMLY SCATTERED SIGNALS II: INVERSION , 1990 .

[24]  M. Priestley.  Evolutionary Spectra and Non‐Stationary Processes , 1965 .

[25]  George Papanicolaou,et al.  Frequency Content of Randomly Scattered Signals , 1991, SIAM Rev..