Blind source separation (BSS) is a critical task in untangling non-stationary signals without prior information. This paper extensively explores diverse time-frequency analysis (TFA) methods within BSS systems over the past decade. It underscores the pivotal role of TFA in dealing with non-stationary signals by characterizing their attributes across time and frequency domains. This approach provides a comprehensive understanding of signal dynamics that surpasses conventional techniques focusing solely on temporal or spectral domains. The paper delves into various TFA methods, investigating their influencing factors and aiding researchers in selecting relevant techniques aligned with their objectives. Furthermore, it comprehensively reviews contemporary research, categorizing BSS algorithms into three classes. The role of commonly used TFA methods in each class is systematically evaluated, identifying their strengths and limitations during different separation stages. The paper addresses challenges in implementing BSS algorithms, particularly in under-determined systems with fewer mixing channels than source signals. It highlights the central role of TFA in overcoming these challenges and enhancing separation outcomes.
The estimation accuracy of the mixed matrix is very important to the performance of the underdetermined blind source separation (UBSS) system. To improve the estimation accuracy of the mixed matrix, the sparsity of the mixed signal is required. The novel fractional domain time–frequency plane is obtained by rotating the time–frequency plane after the short-time Fourier transform. This plane represents the fine characteristics of the mixed signal in the time domain and the frequency domain. The rotation angle is determined by global searching for the minimum L1 norm to make the mixed signal sufficiently sparse. The obtained time–frequency points do not need single source point detection, reducing the calculation amount of the original algorithm, and the insensitivity to noise in the fractional domain improves the robustness of the algorithm in the noise environment. The simulation results show that the sparsity of the mixed signal and the estimation accuracy of the mixed matrix are improved. Compared with the existing mixed matrix estimation algorithms, the proposed method is effective.
Instantaneous frequency estimation plays a crucial role in signal processing; however, existing methods face challenges in achieving high precision, especially when dealing with non-stationary signals and low signal-to-noise ratio environments. To address this, this paper proposes a novel algorithm that combines phase change rate with modulation maxima to enhance the performance of instantaneous frequency estimation. The algorithm synergistically utilizes both time and frequency domain information, overcoming the limitations of traditional methods and significantly improving estimation accuracy and robustness. Through comparative experiments in various signal-to-noise ratio environments, we validate the superior accuracy of the proposed algorithm, particularly demonstrating outstanding performance in low signal-to-noise ratio conditions. This affirms the algorithm's practical superiority, providing an effective approach to enhance accuracy in signal identification.