Signal-to-noise ratio (SNR) estimation is a fundamental task of spectrum management and data transmission. Existing methods for SNR estimation usually suffer from significant estimation errors when SNR is low. This paper proposes a deep learning (DL) based SNR estimation algorithm using constellation diagrams. Since the constellation diagrams exhibit different patterns at different SNRs, the proposed algorithm achieves SNR estimation via constellation diagram recognition, which can be easily handled based on DL. Three DL networks, AlexNet, InceptionV1, and VGG16, are utilized for DL based SNR estimation. Experimental results show that the proposed algorithm always performs well, especially in low SNR scenarios.

The three-phase four-leg inverter has garnered significant attention for its advantages in addressing unbalanced loads and other factors. However, compared to traditional three-phase three-leg inverters, the introduction of additional branches and the occurrence of unbalanced three-phase currents increase the complexity of open-switch (OS) fault modes and types, posing challenges for fault diagnosis. In this article, a multiple OS fault diagnosis strategy is designed specifically for three-phase four-leg inverters operating under unbalanced loads. The method begins with the design of a novel interval sliding mode observer (SMO) that accurately and rapidly estimates phase currents through the use of an adaptive reaching law. By using the interval characteristics of the innovative interval SMO, the upper and lower bounds estimations of the observer are then utilized to design a fault detection parameter and its adaptive thresholds to ensure the robustness of the detection algorithm. In addition, single OS fault identification and multiple OS fault identification methods are developed. By combining the detection variables with the identification methods, the proposed approach enables the diagnosis of 36 different types of OS faults. Hardware-in-the-loop test results validate the efficient diagnosis of OS faults in three-phase four-leg inverters utilizing the proposed approach.

Seismic denoising is a core task of seismic data processing. The quality of a denoising result directly affects data analysis, inversion, imaging and other applications. For the past ten years, there have mainly been two classes of methods for seismic denoising. One is based on the sparsity of seismic data. This kind of method can make use of the sparsity of seismic data in local area. The other is based on nonlocal self-similarity, and it can utilize the spatial information of seismic data. Sparsity and nonlocal self-similarity are important prior information. However, there is no seismic denoising method using both of them. To jointly use the sparsity and nonlocal self-similarity of seismic data, we propose a seismic denoising method using sparsity and low-rank regularization (called SD-SpaLR). Experimental results showed that the SD-SpaLR method has better performance than the conventional wavelet denoising and total variation denoising. This is because both the sparsity and the nonlocal self-similarity of seismic data are utilized in seismic denoising. This study is of significance for designing new seismic data analysis, processing and inversion methods.

It is known that the spherical detector (SD) is very efficient for the multiple primary user spectrum sensing in cognitive radios (CRs). However, numerical results showed that the decision threshold obtained by the existing approximation methods has high accuracy in large sample scenarios, but it is not the case in small sample situations. In this letter, we develop SD for the practical applications with small observations. Here, two methods to calculate the decision threshold of the SD are proposed. In the first approach, the accurate expression, in the form of Meijer’s G-function, for the probability of false-alarm is derived. Based on this, the accurate threshold can be obtained via the numerical solving when a target false-alarm probability is given. In the second method, an approximation method is presented to lead to a simple expression for the decision threshold, which enables us to directly compute the threshold with low computation complexity and high accuracy. Simulation results verify the effectiveness of the two proposed methods.

Hadamard ratio detector (HRD) has been used for spectrum sensing scenarios with uncalibrated receivers in cognitive radios. In literature, the theoretical analysis for the probability of false-alarm and the decision threshold was carried out under the asymptotic conditions. Therefore, the obtained results were theoretically valid only when the number of receivers is small but the number of samples is very large. In this letter, we focus on the HRD in non-asymptotic scenarios with relatively small samples. A new analytical expression for the probability of false-alarm is firstly derived by using the cumulant generating function (CGF) method. Based on this, a computationally efficient method for determining the decision threshold is also provided. Simulation results validate the theoretical analysis.

The multiband detection problem in relatively small sample scenarios where the number of subbands is comparable to the number of samples in magnitude is described. Combined with the sequential hypothesis testing, an eigenvalue ratio based method is proposed for the multiband spectrum sensing (MSS). As the distribution of the new statistic when only noise is present can be precisely obtained by using the random matrix theory, the proposed method is able to reliably set the theoretical threshold, and outperforms the traditional MSS methods based on information theoretic criteria and principal component analysis, especially in the cases with small samples. Meanwhile, the proposed method belongs to a blind detection scheme and can be used in cases without prior knowledge of the primary signal, the channel and the noise. Simulation results show the superiority of the proposed method.