Abstract: In this paper, a high resolution technique for detecting the presence of frequency components of a narrow-band signal in the broad-band correlated noise is examined. It is based on the fact that the correlation interval of a broad-band noise is much smaller than that of a narrow-band signal. When the time lag of the autocorrelation function of the input is larger than the correlation interval of the broad-band noise, the effect of broad-band noise is negligible. Therefore, the measured autocorrelation function of large time lag can be used to estimate the frequency components of the narrow-band signal. The basic procedures of the estimation is presented in the paper. Firstly, the autocorrelation function of large time lag is interpolated to the section of small time lag with complex exponential algorithm and the estimation of the autocorrelation function of small time lag for a narrow-band signal is obtained. The whole autocorrelation function is thus constructed from the measured section of large time lag and the estimated section of small time lag. Secondly, the autocorrelation function is extrapolated by using the maximum entropy method to obtain the weighting coefficients of the high-order prediction filter. Then the frequency, power and bandwidth of the components of the narrow-band signal can be estimated accurately. In order to yield high resolution, the high-order autocorrelation matrix is used, the number of the operations for the estimation of frequency and power is tremendous. An efficient method for significantly reducing the number of operation is suggested. In this method, frequency components of stronger power, instead of all components, are used to estimate the frequency accurately and the power approximately. Finally, a series of computer experiments is performed. As an example, the results for the parameter estimation of the line spectrum of sinusoids in broad-band noise are given. The computer simulation shows that spectral resolution of new method presented in this paper is higher than conventional spectral analysis.