Feature Extraction Of Underwater Acoustic Signals Based On The Cross-components Of Time-Frequency Distribution

Underwater target echoes are typically nonstationary signals and Time-Frequency analysis is effective to deal with nonstationary signals. When the echoes have multi-components there are much cross-components in the Time-Frequency distribution (TFD), so it is not suitable for the feature extraction of the echoes. Cross-components suppression is always the intractability in the studies of TFD and the difficulty in signal processing we should overcome. On the other hand, the cross-components have much information of two (or more) signal components, so it is suitable for signal detection. This paper studies mainly on the cross-components in the underwater target echoes signal processing.The main work and originality in this paper can be summarized as below: 1.Studies on the cross-components in the Wigner-Ville distribution of multi-component signals. We present Adaptive Gaussian Kernel Time-Frequency distribution according to the feature of cross-components in the ambiguity domain. But it is not suitable for online implementation or for tracking signal components with characteristic that change with time, a new modified algorithm is present based on the adaptive short-time kernel. Simulations show that adaptive short-time kernel time-frequency distribution has better performance than anyother time-frequency distribution not only in the time-frequency resolution but also in the cross-components suppression.2.Studies on the Wigner-Hough transform in the suppression of cross-components, a modified algorithm is present based on Adaptive Gaussian Kernel Time-Frequency distribution, this algorithm not only suppresses the cross-components but also has good performance in the low SNR.3. A novel modular learning strategy is present based on the cross-component in the WVD and neural network for the detection of a target signal of interest in a nonstationary oceanic environmemt and this strategy makes no assumptions on the environmemt. Simulations show that this new method has much superiority over matched filter.