28 results about "Inverse discrete wavelet transform" patented technology

The invention discloses a semi-supervised underwater image enhancement system and method based on wavelet transform and a diffusion model. The semi-supervised underwater image enhancement system comprises an image decomposition module used for obtaining a low-frequency component and a high-frequency component of an image to be enhanced by using discrete wavelet transform; the low-frequency diffusion module is used for gradually and sequentially denoising the randomly generated pure noise image by using a trained neural network model in a semi-supervised underwater enhanced image model through taking the low-frequency component as condition guidance to obtain an enhanced low-frequency component; the high-frequency repairing module is used for obtaining an optimized high-frequency component; and the inverse discrete wavelet transform module is used for fusing the optimized high-frequency component and the enhanced low-frequency component through inverse discrete wavelet transform to obtain a reconstructed enhanced image. According to the method, the wavelet transform is fused into the diffusion model, and the image enhancement process is optimized by utilizing the semi-supervised learning model and cooperatively utilizing the annotated data and the annotated data, so that the problem that the data scale is limited is solved, and the calculation efficiency in the image enhancement process is improved.

The invention relates to an infrared and visible light image frequency domain diffusion fusion method, and belongs to the technical field of image processing. Comprising the following steps: inputting an infrared image and a visible light image into a potential space encoder to form original bimodal features; a Gaussian noise is randomly sampled, the Gaussian noise and the original bimodal features are input into a wavelet denoising network together to be decomposed into a low-frequency component and three high-frequency components, the bimodal high-frequency components are fused through FMM, and the bimodal low-frequency components are fused through FMM at the lowest scale; carrying out inverse discrete wavelet transform on the fused low-frequency component and high-frequency component to obtain estimated noise, and then carrying out noise removal on the estimated noise to obtain de-noising features; carrying out iterative optimization and feature updating to obtain fusion features; and inputting the fusion features into a potential space decoder to obtain a fusion image. The fusion quality of the infrared and visible light images can be comprehensively improved.

The invention discloses a super-resolution structured light reconstruction method based on a wavelet enhanced CNN-Transformer structure. The super-resolution structured light reconstruction method comprises the following steps of: firstly, carrying out wavelet enhancement on a CNN-Transformer structure; according to the method, image reconstruction is realized based on a WECT network adopting a coding-decoding architecture; the method comprises the following steps: firstly, extracting initial features through shallow convolution, performing two-stage down-sampling by using discrete wavelet transform, and respectively inputting results into a convolutional neural network and a Swin Transform module to perform parallel feature extraction and fusion; secondly, performing two-stage up-sampling through inverse discrete wavelet transform, and performing jump connection with shallow layer features; finally, pixel shuffling up-sampling operation is introduced, and residual jump connection between the input image and the output image is established; according to the method, high-resolution details can be efficiently recovered while the structure reduction precision is ensured, the image definition and contrast are enhanced, and the application value and practicability of the SIM image in scenes such as biological imaging and tissue analysis are improved.

The invention relates to the technical field of image encryption, and discloses an image steganography method and system based on dynamic latent variables and frequency domain embedding, and the method comprises the steps: obtaining a hash value and a plurality of secret keys of a to-be-encrypted color plaintext image; constructing a hyperchaotic system, and iteratively generating a plurality of key streams; a to-be-encrypted color plaintext image is decomposed into R, G and B channel matrixes, and three ciphertext sequences are obtained based on the three channel matrixes; the method comprises the following steps: initializing a noise image, determining a time step by using a secret key, and executing reverse denoising of a denoising diffusion implicit model (DDIM) with a fixed step length to obtain an intermediate latent variable; based on the intermediate latent variable and the three ciphertext sequences, obtaining a dense frequency domain-containing sub-band, and performing inverse discrete wavelet transform on the low-frequency sub-band and the dense frequency domain-containing sub-band to obtain a dense latent variable; and executing DDIM denoising of the time step by taking the secret-containing latent variable as a starting point to obtain a secret-containing image. By adopting the method, the security and visual quality of image steganography can be improved, and high embedding capacity is ensured.

The invention discloses an offshore scene segmentation method based on condition invariant semantic segmentation, and belongs to the field of computer vision and semantic segmentation. According to the method, style mutual conversion of a source domain image and a target domain image is realized through a data loading stage, and a multi-style view pair is generated; a CISS network model is constructed, the CISS network model is composed of a MiT-B5 encoder and a context sensing decoder, the encoder extracts high-level features insensitive to visual conditions by means of discrete wavelet transform, and the decoder achieves feature reconstruction without information loss through inverse discrete wavelet transform and generates a category probability graph; designing a joint loss function including pixel-level cross entropy loss and feature-level invariance loss, and balancing loss weight in combination with a dynamic weighting strategy; and training and optimizing the model by adopting an AdamW optimizer in combination with a linear learning rate warm-up strategy. Experiments show that the method can effectively cope with visual condition changes in a real marine environment, and the robustness of the semantic segmentation model in different marine environments is improved.

The invention relates to the technical field of image information security, and discloses a controllable image steganography method and system based on a steganography diffusion model. Comprising the following steps: acquiring a hash value of a color plaintext image, and obtaining a plurality of random sequences according to the hash value and three-dimensional memristor chaotic mapping; decomposing the color plaintext image into three channel matrixes, and performing scrambling and diffusion operation on the channel matrixes based on a random sequence to generate a color ciphertext image; adding noise to the carrier image to obtain a preprocessed carrier image; obtaining a plurality of wavelet coefficient matrixes according to the preprocessed carrier image; generating a ciphertext-containing wavelet coefficient matrix according to the color ciphertext image and the wavelet coefficient matrix; performing inverse discrete wavelet transform on the confidential wavelet coefficient matrix to obtain a noisy and confidential carrier image; and de-noising the noisy and confidential carrier image by using the steganography diffusion model to generate a final confidential carrier image. According to the method, the visual quality of the carrier image can be ensured, steganography analysis is effectively resisted, and the image steganography safety is improved.

The application discloses a laser communication link correction method, system, device, medium and product based on wavelet interpolation, relates to the field of optical communication, and comprises the following steps: performing stationary wavelet transform and discrete wavelet transform on a light spot image respectively to obtain stationary subband coefficients and discrete subband coefficients; correcting the discrete high-frequency subband coefficients after interpolation processing by using the stationary high-frequency subband coefficients to obtain corrected discrete high-frequency subband coefficients; replacing the discrete low-frequency subband coefficients with the coefficients of the expanded light spot image to obtain replaced discrete low-frequency subband coefficients; through inverse discrete wavelet transform, the centroid method is used again to calculate the centroid coordinates of the light spot image after wavelet interpolation processing and the deflection angle of the beacon light, the deflection angle of the receiving end is adjusted according to the deflection angle of the beacon light, and the correction of the laser communication link is completed; and the stability of the laser communication link can be improved by improving the subdivision accuracy of the light spot.

The invention relates to the technical field of smart power grids, and provides a subsynchronous oscillation suppression method and system for a mine mining area power supply system, and the method comprises the steps: collecting different types of electric data of each electric shovel in the mine mining area power supply system, and obtaining an electric data sequence of each electric shovel; marking a target electric shovel, performing multilayer wavelet decomposition on an electric data sequence of the target electric shovel, obtaining all detail coefficients of each layer, and determining an adaptive adjustment threshold value of the detail coefficients of the same layer; determining a consistency characteristic value of each type of electric data of the target electric shovel, obtaining an adjustment value of a detail coefficient of each layer of the same type of electric data of the electric shovel, and obtaining a de-noised electric data sequence of the electric shovel according to the detail coefficient and inverse discrete wavelet transform; and separating subsynchronous oscillation fault data according to the de-noised electric data sequence to realize subsynchronous oscillation suppression of the mine field mining area power supply system. The subsynchronous oscillation suppression effect of the power supply system can be improved.

The application belongs to the field of three-dimensional reconstruction, and particularly relates to a micro-morphology analysis method based on wavelet multi-layer frequency band decomposition. The method comprises the following steps: firstly, a multi-focus image sequence of a micro scene to be measured is collected, and multi-scale spatial features are obtained through feature coding and projection mapping; then, a two-dimensional discrete wavelet transform method is introduced, the spatial features are decoupled into a low-frequency approximate subband representing global geometry and a high-frequency detail subband representing fine texture, and structure feature correction and directional texture enhancement are respectively implemented; on this basis, inverse discrete wavelet transform and residual fusion are used to construct time-frequency enhanced features, a deep regression network is used to analyze pixel-level focus probability distribution to generate a continuous depth map; finally, a cascading mechanism driven by three-dimensional geometric reconstruction for scene morphology perception is established, and depth geometric edge constraints are used to reversely guide the high-fidelity generation of a full-focus image. Through the wavelet multi-layer frequency band decomposition strategy, the application effectively solves the problem of confusion between texture details and out-of-focus noise in micro imaging, and realizes the synchronous robust analysis of high-precision geometric structure and full-depth clear texture of micro morphology.

The application discloses a beacon light centroid positioning method and device based on improved wavelet threshold in optical communication, and relates to the technical field of optical communication. The method comprises the following steps: acquiring multiple beacon light images; performing median filtering processing and discrete wavelet transform on each image, acquiring multiple wavelet decomposition multi-layer coefficient matrices, and obtaining a denoising matrix by using a VisuShrink method; performing inverse discrete wavelet transform on the denoising matrix to obtain multiple denoising light spot images; calculating the centroid coordinates of light spots in each denoising light spot image by using a gray centroid method; and performing summation and average processing on the centroid coordinates of the multiple images to obtain beacon light centroid positioning coordinates. By improving the wavelet threshold processing, the application effectively removes noise interference, improves the accuracy and stability of beacon light centroid positioning, and provides reliable technical support for accurate positioning in optical communication.

The invention relates to the technical field of sequence recommendation and AI intelligence, in particular to a sequence recommendation method based on a wavelet hybrid expert framework. The method comprises the following steps: acquiring a historical interaction sequence of a user, and converting the historical interaction sequence into embedded representation; performing discrete wavelet transform on the embedded representation, and decomposing the embedded representation into a low-frequency approximation coefficient and a high-frequency detail coefficient; directly reserving the low-frequency approximation coefficient as a trend expert for representing the long-term preference of the user; for the high-frequency detail coefficient, generating a time refining feature through a learnable weight matrix multiplied by the high-frequency detail coefficient element by element; local experts representing instantaneous interests of the user are obtained through a context-aware router; inverse discrete wavelet transform is carried out, and enhanced embedding representation is reconstructed; and adopting a multi-task learning strategy, and carrying out joint training to generate a recommendation model. According to the method, the accuracy and robustness of the recommendation result are improved.

The invention discloses a meta-optimization adversarial sample generation algorithm combined with frequency domain feature fusion. The meta-optimization adversarial sample generation algorithm comprises the following steps: S1, frequency domain feature decomposition; s2, low-frequency reconstruction: reconstructing a low-frequency reconstructed image of the input image by adopting inverse discrete wavelet transform, and calculating a corresponding high-frequency image according to an original image and a low-frequency reconstructed difference value; s3, feature storage: respectively inputting the low-frequency image and the high-frequency image of the clean sample into a white box model, and extracting and storing features corresponding to each layer; s4, performing feature fusion; s5, inputting the fusion features obtained in the step S4 into a subsequent layer of the white box model to obtain model output; and S6, executing a meta-optimization strategy, calculating and combining internal and external gradient information to update the adversarial sample, and outputting a finally generated adversarial sample. According to the method, the image is decomposed and reconstructed in the frequency domain, and the multi-frequency features and the meta-optimization updating strategy are extracted and fused respectively, so that conflicts among different frequency domain features are effectively relieved, and the attack mobility of the generated confrontation sample to a conventional model and a defense model is improved.

The invention discloses a Beidou positioning power frequency harmonic interference suppression method and system and a medium, and the method comprises the steps: receiving a mixed radio frequency signal, and carrying out the down-conversion of the mixed radio frequency signal to obtain an analog intermediate frequency signal; converting the analog intermediate frequency signal into a digital intermediate frequency signal sequence; performing multi-layer discrete wavelet transform on the digital intermediate frequency signal sequence to generate an approximate coefficient sub-band and a detail coefficient sub-band; calculating the energy of the wavelet coefficients in all the sub-bands, comparing the energy with an adaptive energy threshold, and identifying an interference dominant sub-band; a self-adaptive threshold function is used for wavelet coefficients in the interference dominant sub-band so as to suppress interference components in the wavelet coefficients, and a processed wavelet coefficient set is obtained; performing inverse discrete wavelet transform on the processed wavelet coefficient set to obtain a reconstructed digital intermediate frequency signal sequence; and performing baseband processing on the reconstructed digital intermediate frequency signal sequence to obtain a Beidou positioning radio frequency signal after interference suppression. According to the invention, continuity and accuracy of positioning in a complex electromagnetic environment can be guaranteed.

The invention provides a dosing control method, system and equipment for a water treatment process and a medium, and the method comprises the steps: carrying out the discrete wavelet multi-scale decomposition of a standardized input sequence, and obtaining a low-frequency trend component and a high-frequency detail component; inputting the standardized input sequence into a large language sub-model of a pre-trained prediction model to obtain a high-dimensional time sequence feature; inputting the low-frequency trend component, the high-frequency detail component and the high-dimensional time sequence feature into the prediction model to generate a prediction coefficient of the fusion feature; and performing discrete wavelet inverse transformation recombination processing on the prediction coefficient of the fusion feature to obtain an expected output sequence of the water quality parameter data at the future moment, extracting the instantaneous dosing flow in the expected output sequence, and performing dosing control on the water treatment process according to the extracted instantaneous dosing flow. According to the invention, the expected accurate control of the dosing treatment at the future moment can be improved.

The application provides a wavelet transform-based atmospheric data fusion method, and relates to the field of satellite data processing. The method takes FY-4B data as a unified space-time base plate, performs two-dimensional discrete wavelet transform on the spatially overlapped regions of FY-4B and FY-3 series data to obtain wavelet coefficients of each scale of each satellite data; performs time enhancement processing on the wavelet coefficients of each scale of the FY-4B data, and performs inverse discrete wavelet transform on the wavelet coefficients of each scale of each satellite data to reconstruct each satellite image signal; calculates the comprehensive scores of each satellite data to determine the initial fusion weight, and corrects the final fusion weight through an intelligent credibility switching mechanism; and performs weighted summation on each satellite image signal according to the final fusion weight to generate a surface reflectivity product. The application can fuse the complementary advantages of geostationary satellites and polar-orbiting satellites, and improve the spatio-temporal resolution and detail fidelity of multi-source satellite data fusion.

An image processing apparatus and an image processing method thereof are disclosed. The image processing apparatus includes a discrete wavelet transform (DWT) device that performs DWT and down-sampling on a first source image to split the first source image into a low-frequency sub-image including low-frequency components in horizontal and vertical directions and a plurality of high-frequency sub-images including high-frequency components in at least one of the horizontal or vertical directions; a frame memory that stores a low-frequency sub-image of a second source image; a first noise reduction device that reduces noise in the low-frequency sub-image of the first source image using the low-frequency sub-image of the second source image; and an inverse discrete wavelet transform (IDWT) device that applies IDWT to the low-frequency sub-image of the first sub-image and the high-frequency sub-images of the first sub-image, which are noise-reduced by the first noise reduction device, to restore the first source image.

A depth prediction model for predicting a depth map from an input image is disclosed. The depth prediction model leverages wavelet decomposition to minimize computations. The depth prediction model comprises a plurality of encoding layers, a coarse prediction layer, a plurality of decoding layers, and a plurality of inverse discrete wavelet transforms (IDWTs). The encoding layers are configured to input the image and to downsample the image into feature maps including a coarse feature map. The coarse depth prediction layer is configured to input the coarse feature map and to output a coarse depth map. The decoding layers are configured to input the feature maps and to predict wavelet coefficients based on the feature maps. The IDWTs are configured to upsample the coarse depth map based on the predicted wavelet coefficients to the final depth map at the same resolution as the input image.

The present application discloses a method, device, storage medium, and electronic device for adaptively enhancing low-light images, relating to the field of image processing technology. The method comprises: decomposing a low-light image to be enhanced using a discrete wavelet transform to obtain sub-band images, wherein the sub-band images include a first image representing low-frequency color features and multiple second images representing high-frequency detail features; enhancing the first image using an automatic color scale enhancement method to obtain a first enhanced image; enhancing each second image using a fast Fourier transform method to obtain a second enhanced image corresponding to each second image; reconstructing the first enhanced image and each second enhanced image using an inverse discrete wavelet transform reconstruction method to obtain an initial low-light enhanced image; and enhancing the initial low-light enhanced image using a structured weighted guided filtering method to obtain a target low-light enhanced image. The method of the present application can improve the quality of low-light image enhancement.

The invention discloses a dual-channel-based digital watermark processing method and device and electronic equipment, and relates to the technical field of digital watermarking. A dual-channel-based digital watermark processing method comprises a watermark embedding process: performing discrete wavelet transform decomposition on a first audio signal to obtain a low-frequency approximate component and a high-frequency detail component; respectively performing singular value decomposition on the low-frequency approximate component and the high-frequency detail component to obtain a high-frequency singular value and a low-frequency singular value; constructing a second channel according to a ratio feature formed by mapping the high-frequency singular value and the low-frequency singular value, and embedding watermark information; performing inversion calculation on a low-frequency singular value of the embedded watermark based on each ratio characteristic component in the second channel of the embedded watermark information, and reconstructing the low-frequency singular value into a low-frequency approximate component of the embedded watermark; and carrying out inverse discrete wavelet transform processing on the watermark-embedded low-frequency approximate component and the watermark-embedded high-frequency detail component to obtain a watermark-embedded second audio signal.

The invention belongs to the technical field of image encryption, and particularly provides a frequency domain image encryption method and system based on SVD and double random phase coding, a hash function is applied to a plaintext image P to generate a hash value K, two chaotic sequences X and Y are further generated based on the hash value K, and six chaotic sequences X, Y, W, Z, J and L are finally obtained after preprocessing; performing wavelet transform operation on the plaintext image P to obtain a sub-image, and performing hybrid operation on the sub-image and the six chaotic sequences to obtain a reconstructed sub-image; and performing diffusion processing on the reconstructed sub-image to obtain a new sub-image, and performing inverse discrete wavelet transform to obtain a final encrypted image E. According to the method, the singular value matrix is dynamically recombined, and the key sensitivity of the chaotic system is introduced, so that the ciphertext histograms tend to be uniformly distributed, the attack based on the pixel statistical law is effectively resisted, meanwhile, the frequency domain structure correlation is destroyed, and the encryption robustness is improved.

The application relates to the fields of artificial intelligence and financial technology, and discloses an audio watermark embedding method and device, computer equipment and a storage medium, which are characterized in that the method decomposes host audio to be embedded with a watermark into a plurality of subbands of different frequency bands through discrete wavelet transform; an encoder in a U-Net network model is used to perform layer-by-layer down sampling on each of the subbands, and hidden space features of the host audio are extracted; watermark information to be embedded is determined, and the watermark information is encoded to obtain a watermark vector; the watermark vector is embedded in the hidden space features to obtain fusion features; a decoder in the U-Net network model is used to perform layer-by-layer up sampling on the fusion features, and audio information is reconstructed; inverse discrete wavelet transform is performed on the audio information to generate target audio embedded with the watermark information; and therefore, the application can effectively improve the flexibility, adaptability and robustness of audio watermark embedding.

The invention discloses a medical image registration method based on wavelet guidance and cross-scale frequency domain correction, and the method comprises the steps: obtaining a moving image and a fixed image in a single-channel gray 3D volume data form, defining a dense displacement field (DVF), learning a mapping relation through a network, and carrying out the standardization preprocessing of the images; and respectively inputting the moving image and the fixed image into two weight sharing paths Pm and Pf of a wavelet pyramid encoder in the medical image registration model, decomposing and extracting multi-scale wavelet features layer by layer through discrete wavelet transform, and respectively outputting a moving image feature pyramid and a fixed image feature pyramid. The decoder part guides Swin Transform to reconstruct a deformation field layer by layer through inverse discrete wavelet transform, and meanwhile, a cross-scale frequency domain self-correction module is introduced to optimize the deformation field layer by layer and suppress error accumulation. Compared with a spatial domain registration method in the prior art, the method breaks through the limitation of spatial domain registration, realizes space-frequency domain collaborative optimization through full-stage wavelet guidance, and remarkably improves the registration precision.

The present application relates to the technical field of smart grid, and proposes a method and system for suppressing subsynchronous oscillation of a mining area power supply system, comprising: collecting different types of electric data of each power shovel in the mining area power supply system, and obtaining an electric data sequence of each power shovel; marking a target power shovel, performing multi-layer wavelet decomposition on the electric data sequence of the target power shovel, obtaining all detail coefficients of each layer, and determining an adaptive adjustment threshold of the detail coefficients of the same layer; determining a consistency eigenvalue of each type of electric data of the target power shovel, obtaining an adjustment value of the detail coefficients of each layer of the same type of electric data of the power shovel, and obtaining a denoised electric data sequence of the power shovel according to the detail coefficients and inverse discrete wavelet transform; and separating subsynchronous oscillation fault data according to the denoised electric data sequence, so as to realize suppression of subsynchronous oscillation of the mining area power supply system. The present application can improve the effect of suppressing subsynchronous oscillation of the power supply system.

The invention discloses a railway key sub-problem proxy modeling method based on meta-learning, and belongs to the technical field of railway engineering, and the method comprises the steps: constructing a pre-trained graph convolution wavelet transform network based on meta-learning; obtaining a wavelet coefficient of an input physical variable corresponding to the key sub-problem by using a multi-scale discrete wavelet analysis and decomposition module; processing a wavelet coefficient of an input physical variable corresponding to the key sub-problem by using a graph convolution feature aggregation and prediction module to obtain a processed wavelet coefficient; reconstructing the processed wavelet coefficient into a time domain feature through an inverse discrete wavelet transform module; and mapping the time domain feature into a target physical quantity through a full-connection network. According to the method, a solution with good generalization ability is provided for solving each key sub-problem in the railway, and powerful support is provided for fault diagnosis and performance evaluation of a railway system.

The invention relates to the technical field of data enhancement, and provides a wavelet and Fourier dual-domain data enhancement method and device, electronic equipment and a computer program product. The method comprises the following steps: acquiring a labeled sample image and an unlabeled sample image; performing discrete wavelet transform processing on the labeled sample image to obtain a first low-frequency sub-band and a first high-frequency sub-band of the labeled sample image, and performing discrete wavelet transform processing on the unlabeled sample image to obtain a second low-frequency sub-band and a second high-frequency sub-band of the unlabeled sample image; replacing the low-frequency feature of the first low-frequency sub-band with the low-frequency feature of the second low-frequency sub-band in the frequency domain through Fourier transform to obtain a third low-frequency sub-band after feature replacement; and performing discrete wavelet inverse transform processing on the third low-frequency sub-band and the first high-frequency sub-band to obtain an enhanced sample image. By adopting the method, sample images with image style differences can be generated to increase the diversity of training data, so that the generalization ability of an image segmentation model is improved.

This invention discloses an anti-interference perceptual encryption method for UAV images, belonging to the field of image encryption technology. The method constructs a joint deep learning model, including a visual encryption module, a visual decryption module, and a differential attack module based on a bidirectional reversible network. During training, the UAV image is processed through discrete wavelet transform, the bidirectional reversible network, and inverse discrete wavelet transform to generate an encrypted image and key information. The differential attack module simulates interference to generate an interfered encrypted image; this image, along with the key, is then input into the decryption module to recover the original image. The encryption and decryption modules are simultaneously optimized using a composite loss function. This method achieves perceptual encryption and high-fidelity reversible recovery, possesses excellent anti-interference capabilities, is suitable for UAV transmission scenarios, supports real-time processing and high-security decryption, and can be deployed on airborne edge computing platforms or cloud systems, meeting privacy protection requirements.

The invention discloses a medical image segmentation method based on wavelet condition potential diffusion, and relates to the technical field of medical image segmentation, and the method comprises the following steps: obtaining a medical input image, executing a structure-maintained color enhancement operation in a training stage, and using an original input image in a reasoning stage; the preprocessed image is embedded by adopting a multi-scale wavelet encoder extraction condition, and the encoder is realized through multi-layer discrete wavelet transform, high-frequency residual enhancement, inverse discrete wavelet transform and down-sampling splicing; encoding a segmentation target corresponding to the input image through a pre-trained and parameter-frozen auto-encoder to obtain a potential mask; inputting the semantic priori condition and the potential mask into a diffusion UNet, embedding a frequency domain sensing gating block in jump connection of the diffusion UNet, and performing potential space denoising through a regularization iterative algorithm to obtain an optimized potential mask; and decoding the optimized potential mask through a decoding function of the auto-encoder, and outputting a final medical image segmentation result.

This invention discloses a sequence recommendation method based on multi-layer wavelet decoupling and causal convolution. The method first acquires the user's historical interaction behavior sequence and maps it to a dense vector representation through an embedding matrix. Then, it performs a J-level discrete wavelet transform on the sequence, decomposing it into low-frequency approximation coefficients and multi-scale high-frequency detail coefficients, thus decoupling long-term and short-term interests. Next, it employs a heterogeneous dual-branch architecture for differentiated modeling: the low-frequency branch captures the global dependencies of long-term interests through a low-frequency self-attention module, while the high-frequency branch adaptively filters short-term interest signals and suppresses noise through a causal detail gating module. Finally, it reconstructs the complete sequence representation through inverse discrete wavelet transform and predicts the next interactive item. This method effectively solves the problems of insufficient decoupling of long-term and short-term interests and the inability of a unified processing strategy to distinguish between informative signals and noise in existing methods, and achieves recommendation performance superior to existing baseline methods on multiple public datasets.