15 results about "Convolutional decoding" patented technology

The invention discloses an ultra-high-definition video stream adaptive coding method based on deep learning visual saliency, and the method comprises the steps: carrying out the five-scale Gaussian filtering processing and image pyramid construction of a video frame, and combining Sobel gradient, Laplacian edge and local binary pattern feature extraction to generate a multi-scale feature map; a pre-training saliency detection network is adopted, and a smooth saliency thermodynamic diagram is generated through processing of a feature adaptation layer, a residual encoder, a self-attention mechanism and a transposed convolution decoder; dividing the video frame into a high region, a middle region and a low region according to the saliency thermodynamic diagram, and establishing a regionalization coding parameter table; performing differentiated prediction modes, motion estimation and quantization strategies on different salient regions; and organizing coded data according to an H.265 / HEVC standard, and embedding the saliency thermodynamic diagram into supplementary enhancement information for transmission. According to the method, the important region concerned by the user can be intelligently identified, a differentiated coding strategy based on content semantics is realized, and the coding efficiency is remarkably improved.

This invention discloses a filter physical perception optimization method based on residual attention and transposed convolution, comprising the following steps: constructing a filter characteristic prediction surrogate model based on residual self-attention mechanism and transposed convolution decoding to establish a nonlinear mapping relationship between filter geometric parameters and electromagnetic response parameters; training the filter characteristic prediction surrogate model using an S-parameter weighted mean square error loss function to obtain a trained filter characteristic prediction surrogate model; and using a physical perception-improved differential evolution algorithm, employing the trained filter characteristic prediction surrogate model as a predictor to find the optimal combination of geometric parameters that satisfies the filter design specifications in the solution space. This method is a filter optimization design method that balances high-precision prediction and high-efficiency optimization. This method can predict the S-parameters of microwave filters more accurately than traditional neural networks, effectively replacing time-consuming full-wave electromagnetic simulation.

Method for transmitting digital data between a bottom hole assembly and a surface assembly of a drilling rig, wherein at least one of the bottom hole assembly and the surface assembly comprises a transmitting transducer actuatable to generate acoustic waves propagating in a drilling fluid and the other of the bottom hole assembly and the surface assembly comprises a receiving transducer configured to detect acoustic waves, comprising the following steps : processing in sequence the data by means of convolutional coding, an interleaving operation and at least one spread spectrum modulation technique, obtaining a processed digital signal; - converting the processed digital signal into an analog signal; - actuating the transmitting transducer in such a way as to generate, in the drilling fluid, acoustic waves that correspond to the analogue signal; receiving the analog signal through the receiving transducer; converting the analog signal into a received digital signal; - processing in sequence the digital signal received by means of at least one demodulation technique, one "de- interleaving" operation and one convolutional decoding, obtaining the received data.

Machine learning network topologies, and training systems and methods therefor, are described for implementing unified sound conversion (USC), such as for automated name detection. Such automated name detection can support automated attention handling in wearable audio components with active noise control (ANC) to suppress ambient sound. Embodiments of USC network topologies include a feature generator comprising a convolutional encoder, an axial self-attention (ASA) block, and a deconvolutional decoder. A MEL converter generates input filter bank energies (FBEs) from an audio sample of a spoken word. The feature generator is trained to estimate output FBEs from the input FBEs, such that the output FBEs represent the linguistic information of the spoken word absent any speaker-specific suprasegmental features.

The invention provides an equipment residual life prediction method based on a gating multi-scale U-Net framework. The method comprises the following steps: S1, preparing and preprocessing input data; s2, constructing an equipment residual life prediction model based on a gating multi-scale U-Net framework; s3, training an equipment residual life prediction model and hyper-parameter configuration by using the preprocessed training set and verification set data in the S1; and S4, performance evaluation and effect display. According to the invention, by introducing a gating residual error MLP encoder, a multi-scale channel attention U module and a double-branch time sequence convolution decoder, adaptive extraction and fusion of multi-scale degradation features are realized, and the perception ability of the model to key spatial-temporal features is enhanced, so that high-precision and high-robustness residual life prediction is realized in a complex industrial environment.

The present application relates to a kind of fusion lightweight backbone network and remote sensing change detection method of convolution decoding, including using backbone network feature extraction module, for extracting multi-scale features with strong representation ability from input dual-phase remote sensing image;CNN-Decode feature alignment module, using convolutional coding structure is spatially aligned to dual-phase feature, avoid the feature confusion problem of traditional attention mechanism;And change detection module, based on the feature after alignment realizes pixel-level change identification.The present application is optimized by double-path cooperation, while enhancing the feature expression ability of model, it enhances spatial alignment accuracy, effectively improves the integrity and boundary accuracy of change detection under complex scene, under the premise of keeping low computational complexity, significantly improves detection performance, can be widely applied in urban planning, disaster assessment and environmental monitoring and other fields.

Systems and methods for encoding and decoding ultra wide band (UWB) communication signals, where the encoding includes block interleaving at a bit level or symbol level after convolutional encoding, and where the decoding includes block de-interleaving UWB symbols or log-likelihood ratios (LLRs) prior to convolutional decoding. Decoding may additionally include detecting an anomalous UWB symbol or LLR that is determined to be affected by interference prior to block deinterleaving and making the anomalous UWB symbol or LLR null (or reduced).

The invention provides a voltage anomaly detection method based on an automatic encoder. The method comprises the following steps: acquiring and standardizing a voltage signal of an energy storage system; constructing a deep convolution automatic encoder containing a cavity causal convolution encoder and a transposed convolution decoder; normal data are sampled through a sliding window, and unsupervised training is carried out through a dynamic weighting loss function; and calculating a reconstruction error on line, and outputting an anomaly detection result based on a self-adaptive threshold mechanism. According to the invention, the precision and anti-interference capability of voltage anomaly detection of the energy storage system can be improved.

Systems and methods for encoding and decoding ultra-wide band (UWB) communication signals includes encoding by block interleaving at either bit-level or symbol-level following convolutional encoding, and decoding by block de-interleaving UWB symbols or log-likelihood ratios (LLRs) prior to convolutional decoding. Decoding may further include detecting and nulling (or lowering) outlier UWB symbols or LLRs determined to be affected by interference, prior to block de-interleaving.

The invention discloses a method for preventing tampering of electric power internet of things sensing data of cascaded codes, and belongs to the field of data processing. According to the method, the characteristic advantages of cascade connection of the RS codes and the convolutional codes are fused, the grading strategy of triggering the RS codes on demand is adopted, the optimized convolutional decoding algorithm is matched, and on the premise that calculation and transmission expenses are not remarkably increased, the requirement of an electric power internet-of-things scene for real-time transmission and the core target of data reliability are considered. Meanwhile, the improved Viterbi algorithm is used, values of other branch metrics are rapidly calculated through symmetry and a logic relation, and therefore the convolutional code decoding efficiency is improved. Besides, an optimized search algorithm is used, and based on path measurement characteristics of Viterbi decoding, error positions are subjected to probability sorting and a search sequence is optimized, so that the efficiency of RS code decoding error correction is improved.

The invention relates to the technical field of time sequence knowledge graph reasoning, and provides a knowledge graph representation learning model based on multi-view and long-short term hierarchical relationship, which comprises a multi-view construction module, an embedded coding module and a fusion decoding module. The multi-view construction module extracts a complete view, a query-related view and a query-unrelated view from the historical sub-graph sequence, and connects the same entities with different timestamps to construct a global graph. The embedded coding module comprises a multi-view cyclic encoder and a hierarchical relation graph neural network, the multi-view cyclic encoder processes three views through a graph convolutional network and a time gate mechanism to learn short-term dynamic conditions, and the hierarchical relation graph neural network processes a global graph through a time coding function and an attention mechanism to learn long-term features. The fusion decoding module carries out weighted fusion on long and short term historical embedding through a gating integration unit, and then a convolution decoder carries out convolution and similarity matching on the fused embedding so as to calculate probability distribution of each candidate entity.

The application discloses a U-shaped image segmentation method based on a convolution enhanced cross self-attention transformer, which comprises a convolution embedding module, a convolution merging module and a convolution enhanced cross self-attention transformer module as an encoder, and a convolution upsampling module and a convolution decoding module as a decoder. The image block is divided into horizontal and vertical strips by the convolution enhanced cross self-attention transformer module, and feature merging is carried out in the channel dimension, so that the perception range of the self-attention is greatly enhanced, the local coding module is fused into the deformation to enhance the local information processing capacity of the model, the feature representation of the overall style and contour of the object can be provided, the model based on a convolution neural network (CNN) has more stable recognition performance and more accurate segmentation effect for the object with a local change.

This application relates to the field of battery state estimation technology, and provides a battery state estimation method, apparatus, terminal device, and storage medium. The method includes: acquiring battery operating data information for estimating the current state of the battery; inputting the battery operating data information into a pre-trained SOC estimation model to obtain the battery's SOC estimate and SOC estimation range. The SOC estimation model includes a convolutional coding module and a convolutional decoding module. The convolutional coding module includes a first CNN layer, a first positional coding layer, a coding layer, and a distillation layer connected in sequence. The convolutional decoding module includes a second CNN layer, a second positional coding layer, a decoding layer, and an output layer connected in sequence, with the output of the distillation layer connected to the input of the decoding layer. This application can reduce the impact of outliers on the estimation results and greatly improve the accuracy of SOC estimation.

The invention provides an electromagnetic signal enhancement method and device, electronic equipment and a storage medium, and the method comprises the steps: carrying out the feature processing of a to-be-processed electromagnetic signal based on a deep convolution encoder and a deep convolution decoder, and determining the electromagnetic features of the to-be-processed electromagnetic signal; carrying out first enhancement processing on the electromagnetic features on the geometric transformation dimension based on a dynamic hybrid enhancement strategy, and determining the electromagnetic features after the first enhancement; and based on a pre-trained generative adversarial network, performing second enhancement processing on the electromagnetic features after the first enhancement in a spatial dimension, and outputting an enhanced electromagnetic signal. Through an inter-stage feature complementation mechanism, the problem of feature confusion easily generated by an end-to-end model in a complex electromagnetic environment is overcome, and the physical rationality of an enhanced signal is remarkably improved.

This invention discloses a method and apparatus for table anomaly detection to improve the security of network platforms. The method includes: feeding the original feature matrix into a Transformer branch, capturing global cross-column dependencies using multi-head self-attention, and injecting relative position information using rotational position encoding to output global features; using the global features output from the Transformer branch as input to a depthwise separable convolutional encoder, extracting local neighborhood features through channel-wise and point-wise convolution to obtain global-local fused features, which are then used as input to initiate parallel dual-branch operations; using the anomaly score obtained from the reconstruction error of the depthwise separable convolutional decoder through a learnable scaling layer as a soft label, calculating the mean square error with the anomaly probability obtained from the classification branch through softmax, forming a consistency regularization term; weighting and summing the reconstruction loss, classification loss, and consistency loss according to learnable weights to obtain the overall loss, and training the entire network end-to-end through gradient backpropagation; and automatically optimizing the network structure hyperparameters using Bayesian optimization to finally obtain the table data anomaly detection results.