Crossterm-free time-frequency representation exploiting deep convolutional neural network

• Published in: Signal processing 2022-03, Vol.192, p.108372, Article 108372
• Main Authors: Zhang, Shuimei, Pavel, Md. Saidur Rahman, Zhang, Yimin D.
• Format: Article
• Language: English
• Subjects: Crossterm mitigation; Deep neural network; Nonstationary signal; Time-frequency analysis
• ISSN: 0165-1684; 1872-7557
• DOI: 10.1016/j.sigpro.2021.108372

•We developed a deep neural network-based approach to generate crossterm-free time-frequency representations.•We show that, provided that the neural networks are adequately trained, the proposed method works robustly and provides significant performance improvement compared to existing time-frequency representation reconstruction algorithms.•We evaluated the generalization capability of the proposed method, including the effects of noise levels, amplitude difference, variation speed of the IFs, fading, number of signal components, and frequency spreading. Bilinear time-frequency (TF) analyses provide high-resolution time-varying frequency characterization of nonstationary signals. However, because of their bilinear natures, such TF representations (TFRs) suffer from crossterms. TF kernels, which amount to low-pass weighting or masking in the ambiguity function domain, are commonly used to reduce crossterms. However, existing fixed and adaptive kernels do not guarantee effective crossterm suppression and autoterm preservation, particularly for signals with overlapping autoterms and crossterms in the ambiguity function. In this paper, we develop a new method that offers high-resolution TFRs of nonstationary signals with desired autoterm preservation and crossterm mitigation capabilities, especially for signals with slowly time-varying instantaneous frequencies. The proposed method exploits a convolutional autoencoder network which is trained to construct crossterm-free TFRs. For the signals being considered, the proposed technique with properly trained networks offers the capability to outperform state-of-the-art TF analysis algorithms based on adaptive kernels and compressive sensing techniques.