Non-iterative optimal blind deconvolution and its application to machine condition monitoring

Adaptive blind deconvolution via convolutional neural networks for early fault detection in degraded gears under different speeds

ABSTRACT The multimedia courseware developed using Flash offers strong interactivity, small file size, and easy online distribution, making it widely used in education and online teaching. With Flash’s ActionScript language, interactive courseware can be created, including flexible control of multimedia elements such as audio. This study proposes an adaptive voice-programming recognition method using multimedia technology and Flash. Sound signals are collected through an intelligent microphone and processed with Butterworth filtering and wavelet denoising. An improved adaptive threshold endpoint detection algorithm is applied to extract valid voice segments, followed by non-Gaussian measurement to distinguish voice features. Normalised ASD, MFCC, and other features are extracted, and blind deconvolution is used to further refine signal characteristics. An optimised KNN classifier is then employed for signal feature classification to achieve voice programming. Experimental results show that the system reaches an average accuracy of 93.32% for identifying a single student’s voice source and 83.05% for programming control tasks, under fixed equipment location and stable external conditions. Additional tests considering different locations, distances, dataset sizes, and users further confirm the feasibility, effectiveness, and robustness of the proposed voice-programming model.

Deep constrained multi-frame blind deconvolution with learned regularizations

The precise ranging of ultra-wideband (UWB) fuzes relies on extracting time delay information from echo signals. However, ground multipath propagation effects induce a significant time-delay spread in the echo signals. This manifests as a channel impulse response (CIR) composed of numerous, closely spaced components, creating a challenging super-resolution problem that severely constrains the ranging accuracy and reliability of the fuze. Therefore, accurately estimating the CIR that characterizes these multipath structures from a single echo observation is crucial for the UWB fuze to perceive terrain structures and enhance ranging capabilities. This study proposes the following methods: (1) establishing an equivalent discrete multipath model(EDMM) of the ground to characterize the CIR; (2) proposing a sparse blind deconvolution(SBD) method via the ADMM-based framework under an asymmetric structured prior (ASP), which employs parametric projections to constrain the physical morphology of the unknown source signal, and designing a periodic sparse cluster projection operator to achieve super-resolution recovery of the discrete multipath structure of the channel h by enforcing the EDMM prior. Through three-variable robust decomposition, it actively separates dispersed clutter and enhances performance under low signal-to-noise ratio (SNR) conditions. Experimental results from both simulations and measured data demonstrate that the proposed algorithm exhibits excellent robustness and recovery accuracy in complex low-SNR scenarios, providing a foundational offline analysis method for understanding complex channel characteristics and guiding the development of improved real-time ranging algorithms.

In plasma physics research, laser-induced fluorescence (LIF) spectroscopy is a crucial diagnostic tool for determining plasma parameters. To retrieve particle parameter distributions utilizing Doppler shift and broadening characteristics, extraction of the Doppler-broadened spectral component from LIF spectra is essential, a process necessitating deconvolution techniques. Compared with conventional methods, blind deconvolution dispenses with the prior acquisition of the cold plasma spectral line shape, avoiding complex modeling of hyperfine structure and natural broadening, and remains independent of magnetic field strength data at the measurement location. This paper proposes a blind deconvolution algorithm based on Wiener filtering and least-squares minimization, enabling direct extraction of the pure Doppler-broadened spectrum from observed LIF spectra. We systematically investigate the influence of two regularization parameters on the deconvolution results and establish practical guidelines for parameter selection. Experimental validation with measured Hall thruster LIF spectra shows that, compared to conventional Gaussian deconvolution filters and maximum entropy methods, this algorithm achieves equivalent Doppler-broadened spectrum extraction accuracy while eliminating the need for spectral pre-modeling.

This article presents a review and comparative analysis of modern methods for correcting optical distortions in astrophotography acquired under conditions of atmospheric turbulence. The study investigates the physical and theoretical foundations of distortion formation, including atmospheric turbulence, optical aberrations, and noise inherent to the image acquisition process. The objective of this work is to systematize existing approaches and identify the most effective methods applicable to both amateur and professional astrophotography. The analysis covers image resolution restoration algorithms such as classical deconvolution (including the Richardson–Lucy algorithm and Wiener filtering), blind deconvolution, multi-frame processing, and neural network-based techniques. The results of the comparative analysis demonstrate that multi-frame algorithms and neural network approaches exhibit the highest efficiency under constrained computational resources and incomplete knowledge of the point spread function.

Metal components such as rolling bearings are prone to wear, cracks, and defects in harsh environments and long-term use, leading to performance degradation and potential equipment failures. Therefore, detecting surface cracks and other defects in rolling bearings is of great significance for ensuring equipment reliability and safety. However, traditional signal decomposition methods like EEMD and FEEMD suffer from residual noise and mode mixing issues, while deconvolution algorithms such as CYCBD are sensitive to parameter settings and struggle in high-noise environments. To mitigate the susceptibility of fault signals to background noise interference, this paper proposes a fault feature extraction method based on fast complementary ensemble empirical mode decomposition (FCEEMD) and adaptive maximum second-order cyclostationarity blind deconvolution (ACYCBD). Firstly, we propose FCEEMD, which effectively eliminates the residual noise of ensemble empirical mode decomposition (EEMD) and fast ensemble empirical mode decomposition (FEEMD) by introducing paired white noise with opposite signs, solving the problems of traditional decomposition methods that are greatly affected by noise, having large reconstruction errors, and being high time-consuming. Subsequently, a new intrinsic mode function (IMF) screening index based on correlation coefficients and energy kurtosis is developed to effectively mitigate noise influence and enhance the quality of signal reconstruction. Secondly, the ACYCBD model is constructed, and the hidden periodic frequency is detected by the enhanced Hilbert phase synchronization (EHPS) estimator, which significantly enhances the extraction effect of the real periodic fault features in the noise. Finally, instantaneous energy tracking of bearing fault characteristic frequency is achieved through Teager energy operator demodulation, thereby accurately extracting fault state features. The experiment shows that the proposed method accurately extracts the fault characteristic frequencies of 164.062 Hz for inner ring faults and 105.469 Hz for outer ring faults, confirming its superior accuracy and efficiency in rolling bearing fault diagnosis.

Real-time multi-frame blind deconvolution with Nesterov-like optimizer based on GPU

Partial volume effect (PVE) arises from the limited spatial resolution of positron emission tomography (PET) scanners, causing significant quantitative biases that hinder accurate metabolic activity assessment. To address these problems, we proposed an unsupervised deep residual compensation model (U-DRCM) for PET partial volume correction (PVC). U-DRCM first predicted an initial blur kernel for the PVE-affected PET image based on a conditional blind deconvolution module (CBD module). Then, a conditional residual compensation module (CRC module) was introduced to compensate for the error caused by inaccurate blur kernel prediction. The whole model is unsupervised which only needs a single patient's PET image as the training label and the corresponding MR image as the network input. The performance of U-DRCM was evaluated against several established PVC approaches, including Richardson-Lucy (RL), reblurred Van-Cittert (RVC), iterative Yang (IY), neural blind deconvolution (NBD), and deep convolutional neural network (DeepPVC) using both simulated BrainWeb phantom and real clinical datasets. In the simulation study, U-DRCM consistently outperformed competing methods across multiple quantitative metrics, achieved a higher peak signal-to-noise ratio (PSNR), an improved structural similarity index (SSIM), and a lower root mean square error (RMSE). For the real clinical study, U-DRCM delivered substantial improvements in standardized uptake value (SUV) and standardized uptake value ratio (SUVR) across various brain volumes of interest (VOIs). Experimental results show that U-DRCM effectively mitigates the impact of PVE, resulting in high-quality PVC PET images with enhanced brain visualization.

Blind deconvolution is a widely studied vibration analysis tool for extracting fault-related information from vibration signals. However, existing mainstream blind deconvolution methods struggle to extract fault information from strongly random, noisy, and non-smooth vibration signals. To address this issue, this study proposes a novel blind deconvolution method, named maximum reweighted local kurtosis deconvolution (MRLKD), which effectively extracts fault pulses from vibration signals under time-varying speed conditions. First, a novel evaluation metric, reweighted local kurtosis (RLK), is introduced, which exhibits strong robustness against outliers and noise while remaining unaffected by rotational speed variations. Second, the optimal filter is iteratively computed by maximizing the RLK objective function, enabling the extraction of transient pulses from the time-domain signal. Finally, the filtered signals are resampled from the time domain to the angular domain, and fault features are extracted through envelope order spectrum analysis, thereby facilitating fault diagnosis under time-varying speed conditions. The effectiveness and reliability of the proposed method are verified through simulations and experimental signals, compared with existing techniques. The results demonstrate the advantages of the proposed method.

Multidimensional fast nonlinear blind deconvolution network for bearing compound features extraction.

This study proposes a novel framework to enhance inner race fault features in servo motor bearings by acquiring rotary encoder-derived instantaneous angular speed (IAS) signals, which are obtained from a servo motor encoder without requiring additional external sensors. However, such signals are often obscured by strong periodic interferences from motor pole-pair and shaft rotation order components. To address this issue, three key improvements are introduced within the cyclic blind deconvolution (CYCBD) framework: (1) a comb-notch filtering strategy based on rotation domain synchronous averaging (RDA) to suppress dominant periodic interferences; (2) an adaptive fault order estimation method using the autocorrelation of the squared envelope spectrum (SES) for robust localization of the true fault modulation order; and (3) an improved envelope harmonic product (IEHP), based on the geometric mean of harmonics, which optimizes the deconvolution filter length. These combined enhancements enable the proposed improved CYCBD (ICYCBD) method to accurately extract weak fault-induced cyclic impulses under complex interference conditions. Experimental validation on a test rig demonstrates the effectiveness of the approach in enhancing and extracting the fault-related features associated with the inner race defect.

Blind Deconvolution Based on Combined Kurtosis Indicator and Its Application for Fault Detection of Rotating Machinery

Coherent multipath arrivals in oceanic waveguides generate complex interference patterns that degrade the performance of conventional beamforming (CBF) on horizontal arrays, often causing azimuthal bias and sidelobes-especially in the endfire direction. These limitations stem from grazing-angle multipath propagation in the vertical plane. As a physics-based alternative, this study proposes striation-based beamforming (SBF), which leverages the waveguide invariant. SBF consists of three steps: (1) estimating the time-domain Green's function of an unknown broadband source via ray-based blind deconvolution, (2) resampling this Green's function along striation slopes (relative frequency shifts) aligned to the first arrival, and (3) applying CBF to the resampled Green's function. Anchoring to the first arrival-corresponding to the minimum grazing angle-allows SBF to reduce azimuthal bias and sidelobes. This study also introduces a simplified variant, termed CB1, which applies CBF exclusively to the first arrival, bypassing the resampling step. CB1 offers substantial computational savings while achieving performance comparable to SBF. Experimental results using a bottom-mounted horizontal array confirm CB1's effectiveness in tracking the azimuth of a ship of opportunity in shallow water.

Double beamforming leverages two vertical arrays (at the source and receiver) and a matrix of Green's functions to estimate acoustic launch and reception angles of eigenrays in the ocean. This study extends the technique to a more practical configuration involving a vertical receiver array and an unknown broadband source, such as a passing ship. The proposed method consists of two main steps: (1) blind deconvolution to estimate the Green's function between the vertical receiver array and the source, and (2) extrapolation of this function to neighboring ranges using waveguide invariant theory to synthesize a horizontal array centered on the source. Experimental results using broadband noise (200-900 Hz) from a ship of opportunity in shallow water demonstrate the viability of this approach. The estimated launch and reception angles closely match model-based eigenray predictions, with errors under 1°.

Enhancing temporal resolution in ground penetrating radar data through sparse blind deconvolution with the Smoothed One Over Two (SOOT) algorithm and Gabor Deconvolution (GD)

Objects behind a diffuser are observed with varying blur depending on their location and the viewpoint, making the recovery of the clear scene challenging. Unlike defocus or motion blur, the blurring caused by a diffuser is a phenomenon that depends on the three-dimensional geometry of the scene, including the shape of the diffuser. This paper proposes a method to deblur the scene behind the diffuser from the observations from multiple viewpoints through inverse rendering. By representing the scene with 3D Gaussian primitives, the blurry images from an arbitrary viewpoint can be rendered efficiently. By optimizing to reproduce the observations from each viewpoint, we can obtain a Gaussian distribution representing a consistent, shared, clear scene, along with the blur parameter of the diffuser. Comparative experiments have been done to show that our method outperforms conventional techniques based on radiance fields and blind deconvolution. We also demonstrate that the proposed method improves the accuracy of the downstream task of text recognition.

This study extends our previous work on image deconvolution, where we modelled the process as a two-player static game. In the initial work, and without prior knowledge of the original image or the point spread function, player one aimed to recover a clean image, while player two estimated the point spread function using fractional-order derivatives. In this extended version, both players define their objective functions using fractional-order derivatives, leading to a more robust approach. We prove the existence of a Nash equilibrium for this game and introduce an alternating optimization algorithm to ensure convergence. Extensive experiments show that the proposed method outperforms our previous work and other techniques, with significant improvements in image quality metrics such as Peak Signal-to-Noise Ratio and Structural Similarity Index Measure. This enhanced framework offers a novel and effective approach to the challenging problem of blind image deconvolution in image processing.

Optical blur is an inherent property of any lens system and is challenging to model in modern cameras because of their complex optical elements. To tackle this challenge, we introduce a high-dimensional neural representation of blur-the lens blur field-and a practical method for acquiring it. The lens blur field is a multilayer perceptron (MLP) designed to (1) accurately capture variations of the lens 2D point spread function over image plane location, focus setting and, optionally, depth and (2) represent these variations parametrically as a single, sensor-specific function. The representation models the combined effects of defocus, diffraction, aberration, and accounts for sensor features such as pixel color filters and pixel-specific micro-lenses. To learn the real-world blur field of a given device, we formulate a generalized non-blind deconvolution problem that directly optimizes the MLP weights using a small set of focal stacks as the only input. We also provide a first-of-its-kind dataset of 5D blur fields-for smartphone cameras, camera bodies equipped with a variety of lenses, etc. Lastly, we show that acquired 5D blur fields are expressive and accurate enough to reveal, for the first time, differences in optical behavior of smartphone devices of the same make and model.