Research On Weak Signal Detection Algorithms Based On Wavelet Transform And Higher Order Cumulant

Weak signal detection is very important field to every country. It not only has commercial value, but also has more important military value in the national defense. Measurement of weak signals has an extremely wide range of applications, such as radar, sonar, communication, vibration measurement, fault diagnosis, physics and others. With the development of the modern signal proeessing technology, there are some new menthods presented inrecent years, such as chaos, higher order cumulant, neural networws etc. In this paper, weak signal detection with the Gaussian noise and natural noise is researched based on wavelet transform and higher order cumulant.First of all, the concepts, definitions, properties of higher order cumulant and wavelet transform are introduced. The advantage of higher order cumulants in signal detection and features of signal and noise in wavelet transform is analyzed in detail.Secondly, to solve the shortcomings of traditional threshold function in wavelet denosing, an new thresholding function is proposed based on characteristics of soft-threshold function and hard-threshold function. The new function can obtain better threshold which is between the hard threshold and soft threshold by adjusting the parameters. It not only overcomes the segmentation of hard-threshold function effectively, but also solves the constant deviation of soft-threshold function. The simulation results show that the algorithm is feasible and effective.Finally, in order to improve real-time of fourth-order cumulants in weak signal detection and solve low calculation speed of fourth order cumulants, a new methods based on one-dimensional slice of the fourth-order cumulants for detecting weak signals is presented. One-dimensional slice of the fourth-order cumulants is obtained by the recursive formula algorithm. It can improve the detection performance and save the computational time as the same time. The simulation results show that the algorithm is effective.