Inverse Wavelet Research Articles

Crypto Trend Prediction Based on Wavelet Transform and Deep Learning Algorithm

Crypto currency price prediction is one of the biggest challenges for any prediction system owing to its volatility and dependence on many factors. Artificial intelligence powered by the development of sophisticated computing technology and developments in research has opened up a wide variety of applications and new possibilities to solve complicated problems. Extraction of appropriate features and input with features affecting the output is equally important in time series prediction problems. Wavelet transform is considered an effective method to split a non-stationary signal into uncorrelated components. This work attempts to extract the advantage of wavelets to extract features from an available data set and to predict the trend of the crypto price for next 16 days using the Bi-LSTM network. Historic data of crypto currencies available in Yahoo finance through the yFinance API with the python code is used as the primary data for this work. Due to the gaining importance of SHIB crypto due to its multiplication of value from its start, we have taken SHIB trend analysis as a case study with the use of a data set provided by Yahoo Finance API. The novelty of the proposed system lies largely on the combination of wavelet and deep learning system to improve the performance measures. The proposed systems evaluates the wavelet transform coefficients of the historic data available from the Yahoo Finance API and feed the wavelet transform values to the input of Bi-LSTM network to predict the wavelet transform coefficients for the next 16 days. Finally an inverse wavelet transform is used to get the price of the crypto currency for next 16 days. The results obtained shows that the model was able to estimate the trend in crypto price. The trend analysis shows whether there is an increase or decrease of prices in upcoming days and this could be useful to invest or to withdraw the funds to improve the returns. The proposed scheme resulted in lower Mean Square Error (MSE), Mean Average Error (MAE) and Root Mean Square Error (RMSE) compared with the currently available methods. The method could be useful for trend analysis of other crypto currencies as well. Though it is practically impossible to get the trend analysis from the price and volume history alone, the proposed scheme gives a reasonable estimate of the trend of the crypto price. The work may be extended in future to applications like prediction of agricultural product price, variation in climatic conditions, and availability of certain agricultural products like Onion in the market.

Enhancing short-term streamflow forecasting of extreme events: A wavelet-artificial neural network hybrid approach

Accurate short-term streamflow forecasting models are crucial for effective water resource management, enabling timely responses to extreme flood or drought events and mitigating potential socioeconomic damage. This study proposes robust hybrid Wavelet Artificial Neural Network (WANN) models for real-world hydrological applications. Two WANN variants, WANNone and WANNmulti, are proposed for short-term streamflow forecasting of extreme (high and low) flows at eight gauging stations within Brazil's Paraíba do Sul River basin. WANNone directly feeds both the original streamflow data and the decomposed components obtained through an À Trous wavelet transform into the ANN architecture. Conversely, WANNmulti utilizes separate ANNs for the original data, with the final streamflow estimate reconstructed via the inverse wavelet transform of the individual ANN outputs. The performance of these WANN models is then compared against conventional ANN models. In both approaches, Bayesian optimization is employed to fine-tune the hyperparameters within the ANN architecture. The WANN models achieved superior performance for 7-day streamflow forecasts compared to conventional ANN models. WANN models yielded high R2 values (>0.9) and low MAPE (4.8%–14.7%) within the expected RMSE range, demonstrating statistically significant improvements over ANN models (71% and 75% reduction in RMSE and MAPE, respectively, and 69% increase in R2). Further analysis revealed that WANNmulti models generally exhibited superior performance for low extreme flow predictions, while WANNone models achieved the highest accuracy for high extreme flows at most stations. WANN models' strong performance suggests their value for real-time flood warnings, enabling improved decision-making in areas like flood/drought mitigation and urban water planning.

A Seismic Inversion Method Based on Multi-Scale Super-Asymmetric Cycle-JNet Network

In order to improve the resolution and accuracy of seismic inversion, this study constructs a multi-scale super-asymmetric network (Cycle-JNet). In this model, wavelet analysis is used to capture the multi-scale data characteristics of well-seismic data, thereby improving the machine’s ability to learn details. Using the UNet neural network from Convolutional Neural Network (CNN), we modified the network structure by adding several convolution kernel layers at the output end to expand generated data, solving the problem of mismatched resolutions in well-seismic data, thus improving the resolution of seismic inversion and achieving the purpose of accurately identifying thin sandstone layers. Meanwhile, a cycle structure of Recurrent Neural Network (RNN) was designed for the secondary learning of the seismic data generated by JNet. By comparing the data transformed through inverse wavelet transform with the original data again, the accuracy of machine learning can be improved. After optimization, the Cycle-JNet model significantly outperforms traditional seismic inversion methods in terms of resolution and accuracy. This indicates that this method can provide more precise inversion results in more complex data environments, providing stronger support for seismic analysis.

DUIANet: A double layer U-Net image hiding method based on improved Inception module and attention mechanism

Image hiding is the process of hiding secret image into cover image, and revealing secret image from steganographic image. However, the quality of generated images and the ability to resist steganalysis detection of steganographic image can be further improved. We propose a novel double-layer U-Net image hiding method based on improved Inception and attention mechanism. The secret image undergoes Haar wavelet transform, hiding network hides the wavelet sub-band of secret image into cover image. We use improved Inception while cascading convolution and attention mechanisms to better extract and fuse information, resulting in high-quality steganographic image. In the revealing stage, Attention mechanism can help networks extract more features about secret image, then we using inverse Haar wavelet transform to revealing high-quality secret image. The experimental results show that the method can obtain high quality images and have good security. When hiding multiple images, the method still has good performance.

Improved DWT and IDWT architectures for image compression

In the recent era, a rapid development in the field of image processing has been observed. One of the important applications in image processing is compression. Several wavelet transform based image compression techniques have already been introduced. In this paper, Discrete Wavelet Transform (DWT) and Inverse Discrete Wavelet Transform (IDWT) based improved image compression and decompression techniques have been proposed by incorporating a scaling factor. The DWT and IDWT algorithms are implemented using folded architecture. To reduce the usages of hardware resources, a multiplier is recursively used. Image compression and decompression schemes based on proposed DWT and IDWT architectures are tested using four different image databases. The proposed technique provides better results in terms of bits per pixel, compression ratio, mean square error, peak-signal-to-noise ratio, normalized correlation coefficient and structural similarity index. FPGA based synthesis has been performed using Xilinx Vivado Synthesis tool in terms of slice LUTs, slice registers, clock frequency, delay and power. The synthesis results show that proposed DWT and IDWT architectures are amenable for image compression and decompression applications.

Robust Elastic Full-Waveform Inversion Based on Normalized Cross-Correlation Source Wavelet Inversion

The elastic full-waveform inversion (EFWI) method efficiently utilizes the amplitude, phase, and travel time information present in multi-component seismic recordings to create detailed parameter models of subsurface structures. Within full-waveform inversion (FWI), accurate source wavelet estimation significantly impacts both the convergence and final result quality. The source wavelet, serving as the initial condition for the wave equation’s forward modeling algorithm, directly influences the matching degree between observed and synthetic data. This study introduces a novel method for estimating the source wavelet utilizing cross-correlation norm elastic waveform inversion (CNEWI) and outlines the EFWI algorithm flow based on this CNEWI source wavelet inversion. The CNEWI method estimates the source wavelet by employing normalized cross-correlation processing on near-offset direct waves, thereby reducing the susceptibility to strong amplitude interference such as bad traces and surface wave residuals. The proposed CNEWI method exhibits a superior computational efficiency compared to conventional L2-norm waveform inversion for source wavelet estimation. Numerical experiments, including in ideal scenarios, with seismic data with bad traces, and with multi-component data, validate the advantages of the proposed method in both source wavelet estimation and EFWI compared to the traditional inversion method.

A multi-level wavelet-based underwater image enhancement network with color compensation prior

Due to the scattering of light and the influence of different water types, underwater images usually suffer from different type of hybrid degradation, e.g. color distortion, blurred details and low contrast. Existing underwater image enhancement methods are weak at handling hybrid degradation simultaneously, resulting in low quality results. Inspired by the fact that wavelet-based enhancement methods can correct color and enhance details in frequency domain and the color compensation prior can compensate missing color information in spatial domain, we design the Multi-level Wavelet-based Underwater Image Enhancement Network (MWEN) with the color compensation prior to enhance image in both frequency domain and spatial domain. Specifically, we integrate the multi-level wavelet transform and the color compensation prior into a multi-stage enhancement framework, where each stage consists of a Multi-level Wavelet-based Enhancement Module (MWEM), a Color Compensation Prior Extraction Module (CCPEM) and a color filter with prior-aware weights. The MWEM decomposes image features into low frequency and high frequency by a wavelet transform, and then enhances them by a low frequency enhancement branch and several high frequency enhancement branches, respectively. The low frequency reduces the color distortion of different water types using Instance Normalization for style transfer, while the high frequency enhancement enhances sparse details using a non-local sparse attention mechanism. After the inverse wavelet transform, the preliminary enhanced result by the MWEM is obtained. Then, the color filter whose weights are customized by the color compensation information extracted from the CCPEM dynamically is applied to output of the MWEM for color compensation. Such an operation enables network to adapt to hybrid degradation and achieve better performance. The experiments demonstrate MWEN outperforms existing UIE methods quantitatively and qualitatively.

Blind watermarking scheme for medical and non-medical images copyright protection using the QZ algorithm

The widely used watermarking technique is to hide the watermark information through adding process, which means the watermark is added with the host image to realize embedding. The constraints of these addition-based watermarking systems mainly are non-blind, they are vulnerable to forgery attacks and have the false-positive problem. To solve these problems, a novel blind watermarking scheme based on the QZ algorithm for medical and normal images is first presented in this paper. In the presented scheme, we take the watermark and the host image as two inputs of the QZ algorithm to realize embedding, rather than the adding process. The outputs of the QZ algorithm include four parts: two upper quasitriangular matrices and two unitary matrices. Then the three parts of the outputs are synthesized by the vector synthesis and inverse discrete wavelet transform-singular value decomposition (DWT-SVD) to obtain the watermarked host image. Due to the characteristics of the QZ algorithm, our scheme is blind. The performance is tested and compared by applying our scheme to both medical and non-medical images. The experimental results show that our scheme is robust against various image-processing attacks, forgery attacks, and false-positive problem. And from the comparison results, the proposed scheme outperforms the existing addition-based schemes in terms of robustness against the Median filter, Low-pass filter, Gaussian blur attacks, and forgery attacks, while maintaining good invisibility.

Optical single-channel cryptosystem based on the non-negative matrix factorization and face biometric in cyan–magenta–yellow–black color space

In this paper, an optical color single-channel asymmetric cryptosystem based on the non-negative matrix factorization (NMF) and a face biometric in cyan–magenta–yellow–black (CMYK) space is proposed. To the best of our knowledge, this is the first time that NMF has been introduced into optical color image encryption. In the proposed cryptosystem, the color image in CMYK space is first decomposed into four color channels: C, M, Y, and K. By performing NMF operations on the four color channels, the four basic and sparse matrices can be obtained, respectively, which achieves asymmetry and saves computational resources. The four basis matrices can be used as private keys, and the four coefficient matrices are synthesized by the inverse discrete wavelet transform for subsequent encryption. Finally, the synthesized image is encoded with double random phase encoding based on phase truncation (PT). Compared with the existing PT-based cryptosystems, our cryptosystem can improve security against a special attack. In addition, the chaotic random phase mask is generated by a face biometric, which is noncontact and unique. Numerical simulation results are shown to verify the feasibility and robustness of our cryptosystem. Further, the proposed cryptosystem can be extended to encrypt multiple images conveniently.

Hybrid technique for crop image compression using discrete wavelet transform and sparse singular vector reconstruction

In this paper, a hybrid compression technique, based on discrete wavelet transform (DWT) and singular vector sparse reconstruction (SVSR), is proposed. In the initial phase, an original image has been divided into four sub-band by using the DWT technique. After that, SVSR is utilized for boosting the compression ratio and to improve reconstruction using piecewise cubic Hermite interpolation (PCHIP). Inverse discrete wavelet transform (IDWT) process has been applied to obtain a compressed crop image. The first time, this developed technique has been applied for compression. The performance of proposed hybrid techniques has been assessed in terms of several evaluation parameters, which are compression ratio (CR), peak signal-to-noise ratio (PSNR), structural similarity index measure (SSIM), and mean square error (MSE) on ten crop image data sets. It is observed from the results that the proposed technique is quite effective in compression of different types of crop images. Further, the experimental results show that the proposed hybrid method achieved significantly higher compression over the existing techniques. The proposed technique with Haar wavelet provides 412.5% and 94.97% higher CR as compared to Singular Value Decomposition (SVD), and SVSR respectively with desirable image quality. Further, other wavelets Db4 provide 396.48% and 88.88% and Bior3.5 also provides 306.74% and 54.74% higher CR as compared to SVD, and SVSR respectively with similar image quality.

Fine Perceptive GANs for Brain MR Image Super-Resolution in Wavelet Domain.

Magnetic resonance (MR) imaging plays an important role in clinical and brain exploration. However, limited by factors such as imaging hardware, scanning time, and cost, it is challenging to acquire high-resolution MR images clinically. In this article, fine perceptive generative adversarial networks (FP-GANs) are proposed to produce super-resolution (SR) MR images from the low-resolution counterparts. By adopting the divide-and-conquer scheme, FP-GANs are designed to deal with the low-frequency (LF) and high-frequency (HF) components of MR images separately and parallelly. Specifically, FP-GANs first decompose an MR image into LF global approximation and HF anatomical texture subbands in the wavelet domain. Then, each subband generative adversarial network (GAN) simultaneously concentrates on super-resolving the corresponding subband image. In generator, multiple residual-in-residual dense blocks are introduced for better feature extraction. In addition, the texture-enhancing module is designed to trade off the weight between global topology and detailed textures. Finally, the reconstruction of the whole image is considered by integrating inverse discrete wavelet transformation in FP-GANs. Comprehensive experiments on the MultiRes_7T and ADNI datasets demonstrate that the proposed model achieves finer structure recovery and outperforms the competing methods quantitatively and qualitatively. Moreover, FP-GANs further show the value by applying the SR results in classification tasks.

Small Target Detection Algorithm Based on Wavelet Domain Features and Weighted Sample Allocation Strategy

Small target detection using UAV aerial photography has emerged as a popular research topic. The resolution of small target images is low and the background is complex. In this paper, utilizing a single-stage target detection algorithm ATSS as its foundation, we propose a small target network detection model SDW-Net that combines deformation convolution with wavelet domain feature enhancement by weighting the samples from both classification and regression tasks, and combining the strong local feature representation ability of discrete wavelet transform. Firstly, the original features are obtained by using discrete wavelet transform, and the weighted wavelet features are obtained through the deformable convolution block and the spatial channel attention unit; secondly, the weighted features of the regular domain are obtained through the inverse wavelet trans-form; finally, the original feature and the wavelet features are fused to enrich the details of the small target image and sent to the detection head for output. According to the experimental findings, when compared to the ATSS algorithm, the proposed detection model achieved an average increase in prediction accuracy of 7.2%, and a recall rate increase of 3.7%, and the problem of missing detection and false detection can be alleviated, which can provide a new technical solution for small-sized target detection scenarios.

Quantum analysis of squiggle data

Squiggle data is the numerical output of DNA and RNA sequencing by the Nanopore next generation sequencing platform. Nanopore sequencing offers expanded applications compared to previous sequencing techniques but produces a large amount of data in the form of current measurements over time. The analysis of these segments of current measurements require more complex and computationally intensive algorithms than previous sequencing technologies. The purpose of this study is to investigate in principle the potential of using quantum computers to speed up Nanopore data analysis. Quantum circuits are designed to extract major features of squiggle current measurements. The circuits are analyzed theoretically in terms of size and performance. Practical experiments on IBM QX show the limitations of the state of the art quantum computer to tackle real life squiggle data problems. Nevertheless, pre-processing of the squiggle data using the inverse wavelet transform, as experimented and analyzed in this paper as well, reduces the dimensionality of the problem in order to fit a reasonable size quantum computer in the hopefully near future.

End-to-end wavelet block feature purification network for efficient and effective UAV object tracking

Recently, unmanned aerial vehicle (UAV) object tracking tasks have significantly improved with the emergence of deep learning. However, owing to the object feature pollution caused by motion blur, illumination variation, and occlusion, most of the existing trackers often fail to precisely localize the target in the complex real-world circumstances. To overcome this challenge, we present a novel wavelet block feature purification network (WFPN) for efficient and effective UAV tracking. WFPN is mainly composed of downsampling network through wavelet transforms and upsampling network through inverse wavelet transforms. To be specific, the downsampling network performs discrete wavelet transform (DWT) to reduce interference information and preserve original feature details, while the upsampling network applies inverse DWT (IDWT) to reconstruct decontaminated feature information. Additionally, a novel sequential encoder is introduced to achieve a better purification effect. Finally, a pooling distance loss is devised to improve the purification effect of DWT downsampling network. Extensive experiments show that our WFPN achieves promising tracking performance on three well-known UAV benchmarks, especially on sequences with feature pollution. Moreover, our method runs at 33.2 frames per second on the edge platform of Nvidia Jetson AGX Orin, which is suitable for UAVs with limited onboard payload and computing capability.

A shrinkage-free approach for fusing GRACE-based total water storage changes with models using wavelet multiresolution analysis

The observations from Gravity Recovery and Climate Experiment (GRACE) satellites provide total water storage changes (TWSC) at ∼300 km or larger spatial scales. To obtain information at a higher spatial resolution, suitable for management purposes, previous studies combined GRACE TWSC maps with information from hydrological models via assimilation techniques (e.g. CLSM-DA), or with external observations via scaling and machine learning approaches. Both techniques may introduce shrinkage errors, a reduction in amplitude of long spatial wavelength signals, and distorting water budgets at basin scale best observed by GRACE. Here, we present a fusion approach based on wavelet multiresolution analysis to combine long-wavelength GRACE TWSC observations with short-wavelength measurements of TWSC from the GLDAS Noah model in the contiguous United States (CONUS). We decompose TWSC maps into building blocks at different spatial wavelengths, examine their statistical characteristics, and combine complementary components at wavelet coefficient levels. The fused TWSC is then obtained by inverse wavelet transform. A spectral analysis indicates that the fused TWSC comprises frequency content that balances spectral characteristics of both input datasets. The fused TWSC dataset possesses enhanced details in the spatial domain, while it accurately quantifies the water budget and its long-term spatial trend at basin scale during 2003–2015, showing a good agreement with GRACE estimates. Independent validation against elastic hydrologic loading deformation measured at ∼2000 GNSS stations across CONUS shows similar overall performance for GRACE and fused datasets, while it outperforms GRACE in modeling groundwater storage changes in CONUS when compared to CLSM-DA.

Electroencephalographic Characterization by Covariance Analysis in Men with Parkinson’s Disease Reveals Sex- and Age-Related Differences

Parkinson’s disease (PD) is the fastest growing neurological disease associated with ageing; its symptomatology varies between sexes. Several quantitative electroencephalography analyses have been used to study the early stages and progression of PD. In this study, we aim to characterize the brain activity by considering the five brainwaves in an eyes-closed resting state, using covariance wavelet analysis (CWA) of electroencephalographic records (EEGs) to analyze the influence of sex and age. To effectively eliminate artifacts from the EEG dataset and extract pertinent brain activity, we employ the inverse wavelet analysis. EEGs from men with PD were divided into two age groups (PD < 60 and PD > 60 years old) with their respective age-matched controls (CL). Brain activity patterns in frequency and power domains were analyzed with the CWA. Main frequency profiles, global wavelet curves, power anomalies, and power per brainwave were used to illustrate the CWA patterns. Power anomalies were used to generate anteroposterior power gradients. In PD < 60 men, frequency and power for the α brainwave decreased, while the δ brainwave increased. The θ brainwave increased and was dominant over the α brainwave in PD > 60 men. The anteroposterior power gradient in PD < 60 men had a positive slope, but it was negative in CL. In both PD and CL > 60 men, the anteroposterior gradient was negative. In PD > 60 men, the θ brainwave increased and became dominant. Men with PD had twice the θ brainwave increase. An inverse relationship between α and δ brainwaves was detected in a PD < 60 sex comparison. A conventional EEG spectral analysis using CWA indicated significant differences in brain activity patterns in the PD/CL groups affected by sex and age, yielding previously unknown information.

Image compression-hiding algorithm based on compressive sensing and integer wavelet transformation

In this paper, a three-dimensional chaotic system is proposed. Based on this chaotic system and two-dimensional compressive sensing, an asymmetric visually meaningful image compression-hiding algorithm is presented. Firstly, in the keystream generation stage, a novel parameter transformation model is constructed to pick up the feature information from the plain image as the plaintext key. Then, Rivest-Shamir-Adleman algorithm is employed to encrypt the plaintext key into the ciphertext key seen as public key. Before generating the initial values for the chaotic system, a new initial value getting model is designed to transform both the plaintext and the ciphertext keys. After solving the chaotic system, the keystream is produced which is then used in the image encryption process. Secondly, in the compression and encryption phase, a novel keystream pre-processing model is built to generate new sequences with a confusion performed on the plain image. Then, a newly constructed measurement matrix is designed to do two-dimensional compressive sensing on confusing the image to get measurements. Before obtaining the cipher image, a double diffusion operation is applied on these measurements. Thirdly, in the image hiding stage, the carrier image is performed by integer wavelet transformation to obtain coefficient matrices. Then, the cipher image is decomposed in decimal, getting the ones, tens and hundreds of pixels to form three bit matrices, of which are embedded into the three medium-high coefficient matrices of integer wavelet transformation, respectively. Finally, after performing inverse integer wavelet transformation, the carrier image containing secrets, i.e., visually meaningful encrypted image, is obtained. Experimental results also show that at a low compression ratio of 0.25, the normalized correlation coefficient between the original plain image and the recovered image is almost equal to one, while the peak signal-to-noise ratio between the carrier image containing secrets and its original carrier can reach as high as 42 dB. In addition, the proposed image compression-hiding algorithm performs good ability consideing the brute force attack and the cropping attack.

An Image Encryption Algorithm Based on Improved Hilbert Curve Scrambling and Dynamic DNA Coding.

As an effective method for image security protection, image encryption is widely used in data hiding and content protection. This paper proposes an image encryption algorithm based on an improved Hilbert curve with DNA coding. Firstly, the discrete wavelet transform (DWT) decomposes the plaintext image by three-level DWT to obtain the high-frequency and low-frequency components. Secondly, different modes of the Hilbert curve are selected to scramble the high-frequency and low-frequency components. Then, the high-frequency and low-frequency components are reconstructed separately using the inverse discrete wavelet transform (IDWT). Then, the bit matrix of the image pixels is scrambled, changing the pixel value while changing the pixel position and weakening the strong correlation between adjacent pixels to a more significant correlation. Finally, combining dynamic DNA coding and ciphertext feedback to diffuse the pixel values improves the encryption effect. The encryption algorithm performs the scrambling and diffusion in alternating transformations of space, frequency, and spatial domains, breaking the limitations of conventional scrambling. The experimental simulation results and security analysis show that the encryption algorithm can effectively resist statistical attacks and differential attacks with good security and robustness.

Ocular Artifact Removal Method Based on the Wavelet and ICA Transform

The electroencephalogram is a promising tool used to unravel the mysteries of the brain. However, such signals are often disturbed by ocular artifacts caused by eye movements. In this study, Independent Component Analysis and Wavelet Transform based ocular artifact removal method, which does not need reference signals, is proposed to obtain signals free from ocular artifacts. With our proposed method, firstly, the ocular artifact regions in the time domain of the signal are detected. Then the signal is decomposed into its components by independent component analysis and independent components containing artifacts are detected. Wavelet transform is only applied to these components with artifact. Zeroing is applied to the parts of the wavelet coefficients obtained as a result of the wavelet transform corresponding to the ocular artifact regions in the time domain. Finally, the clean signal is obtained by inverse Wavelet transform and inverse Independent Component Analysis methods, respectively. The proposed algorithm is tested on a real data set. The results are given in comparison with the method in which the zeroing is applied to the classical independent components. According to the results, it is seen that most of the signal is not affected by the zeroing and the neural part of the EEG signals is successfully preserved.

Fault Diagnosis of Train Wheelset Bearing Roadside Acoustics Considering Sparse Operation with GA-RBF

Trackside acoustic signals are useful for non-contact measurements as well as early warnings in the diagnosis of train wheelset bearing faults. However, there are two important problems when using roadside acoustic signals to diagnose wheel-to-wheel bearing faults; one is the presence of strong interference from strong noise and high harmonics in the signal, and the other is the low efficiency of bearing fault identification caused by it. Therefore, from the viewpoint of solving the two problems, a sparse operation method is proposed for denoising and detuning the modulation of the roadside acoustic signal, and a machine learning classifier with a Genetic Algorithm (GA)-optimized Radial Basis Neural Network (RBFNN) is proposed to improve the rate at which the features of roadside acoustic signal faults are recognized. Firstly, the background noise is filtered out from the Doppler-corrected acoustic signal using the Sparse Representation method, and the inverse wavelet transform is reconstructed into a noiseless signal. Secondly, the interference high-harmonic signal in the signal is filtered out using the Resonant Sparse Signal Decomposition (RSSD) method. Then, the GA is selected to optimize the parameters of the RBF neural network and build a fault diagnosis model. Finally, the extracted acoustic signal feature set is trained on the network model, and the trained model is used for testing. In summary, the sparse operation on the roadside acoustic signal processing and the GA-RBFNN diagnosis model were verified as being very effective in the diagnosis of roadside acoustic train wheel pair faults through the simulation experiment.