21results about How to "Improve perceived quality" patented technology

The invention discloses a speech enhancement method and system based on a time-frequency graph convolutional network in the technical field of speech signal processing, and the method comprises the steps: obtaining to-be-enhanced noisy speech data, inputting the to-be-enhanced noisy speech data into an improved metric generative adversarial network (CMGAN), and outputting an enhanced speech time domain signal; the improvement method of the metric generative adversarial network CMGAN comprises the following steps: replacing an encoder module of the metric generative adversarial network CMGAN with a time-frequency graph convolution encoder module; in the time-frequency graph convolution encoder module, a time convolution network and a graph convolution network are used for respectively extracting a long-time dependency relationship and global frequency characteristic information of noise-containing voice data to be enhanced, and dense residual error connection is introduced to fuse multi-scale characteristic information at the same time. The speech enhancement performance is improved under the condition of less model parameter quantity and calculation quantity.

The invention provides an All-In-One image restoration method based on deep learning in the technical field of image restoration, and the method comprises the steps: S1, obtaining a large number of original images, carrying out the randomization processing of each original image, obtaining a corresponding noise image, and constructing a data set based on each original image and the noise image; s2, creating an image restoration model based on the prompt network, the encoder and the decoder, and setting a loss function of the image restoration model; the prompt network, the encoder and the decoder are connected in sequence; the encoder is provided with a plurality of data stream architecture blocks; the decoder is provided with a plurality of prompt blocks; s3, training the image restoration model through the data set, and continuously optimizing hyper-parameters of the image restoration model in the training process; and S4, deploying the trained image restoration model, and performing image restoration by using the deployed image restoration model. The method has the advantages that the generalization ability and the restoration quality of image restoration are greatly improved.

The invention discloses an electric pedal and an optimization design method and device thereof, and relates to the technical field of optimization design of electric pedals. According to the technical scheme, when design optimization is conducted on the electric pedal driven by the four-connecting-rod contraction and expansion mechanism to turn over, whether the pedal body of the electric pedal has inward horizontal displacement or not in the outward turning-over process of the electric pedal is judged during optimization, and when the pedal body has inward horizontal displacement, the electric pedal is driven to turn over by the four-connecting-rod contraction and expansion mechanism; if the outward displacement of the pedal in the horizontal direction cannot be accurately controlled on the basis of the structure of the current four-connecting-rod contraction and expansion mechanism, the position of the hinge point of the four-connecting-rod contraction and expansion mechanism is readjusted until the pedal is smoothly unfolded outwards in the overturning process. According to the technical scheme, the four-connecting-rod contraction and expansion mechanism of the electric pedal can be optimized, and the accuracy of horizontal displacement control during overturning of the electric pedal is improved. Meanwhile, according to the optimization design method, the overturning process of the electric pedal can be smoother, and the sensing quality of the electric pedal is improved.

The application provides an audio processing, speech recognition and model training method, device and program product, the method relates to the field of artificial intelligence, the multi-frame speech features of audio are obtained by encoding the audio through a speech mark model, the spectrum and time characteristics of the input audio can be captured; the multi-frame speech features are converted to the frequency domain to obtain multi-frame frequency domain features, each frame of frequency domain features is decomposed into frequency domain feature components on multiple frequency bands, and the frequency domain feature components on the multiple frequency bands are quantized to generate speech marks corresponding to each frame of frequency domain features, the speech mark sequence of the audio is determined according to the speech marks corresponding to the multi-frame frequency domain features, the speech mark sequence of the audio is generated by quantizing the multiple frequency bands in the latent feature space of the speech features, the information in different frequency bands is quantized to optimize the bit rate allocation, the quantization efficiency and the perceptual quality can be improved, and the quality and efficiency of audio discretization are improved.

The present invention provides a vehicle side enclosure structure design method and related equipment, side enclosure structure, and vehicle. The design method includes: obtaining a vehicle three-dimensional model; establishing a simulated observation point according to the vehicle three-dimensional model; determining the sight shielding line of the side enclosure structure of the vehicle three-dimensional model according to the simulated observation point; adjusting the contour parameters of the side enclosure structure and / or the layout position parameters of the vehicle's bottom arrangement parts according to the sight shielding line to avoid the bottom arrangement parts from passing through the sight shielding line. The design method provided by the present invention can avoid the bottom arrangement parts from passing through the sight shielding line by adjusting the contour parameters of the side enclosure structure and / or adjusting the layout position parameters of the bottom arrangement parts, thereby reducing the possibility of the bottom arrangement parts being exposed to the side enclosure structure when the designed vehicle is observed from the side, providing a reliable guarantee for improving the overall image of the vehicle, improving the overall perceived quality of the vehicle, and facilitating improving the user experience of the vehicle.

The present invention relates to the field of image processing, and in particular to an image compression system, device and method, which are used to solve the problem that existing image compression methods are often difficult to achieve a good balance between compression efficiency and image quality, especially when processing high-resolution images, the compressed image may have obvious quality loss; the system uses intelligent segmentation technology and then compresses separately, thereby achieving efficient image compression while ensuring the quality of the compressed image and improving the perceived quality of the image; quality monitoring is also performed after compression, which can accurately reflect the quality of the compressed image and provide a reliable basis for adjusting compression parameters. The system achieves efficient image compression while ensuring image quality, achieves a good balance between compression efficiency and image quality, is suitable for different types of images and compression requirements, and has strong versatility and expansibility.

The application discloses a sliding door lock system and relates to the technical field of sliding door lock arrangement. The sliding door lock system comprises an outer opening assembly, an inner opening assembly and a lock body. The outer opening assembly comprises a base, an outer opening handle arranged on one side of the base, a pin shaft arranged in the base, a first torsion spring arranged on the pin shaft and an outer opening rotating arm arranged on one side of the base. The inner opening assembly comprises an inner opening bottom plate, an inner opening rotating arm arranged on one side of the inner opening bottom plate, a rotating shaft arranged in the inner opening bottom plate and an inner opening handle arranged on one side of the rotating shaft. The application has the beneficial effect of omitting the conversion link of the inner opening device and directly controlling the lock body by the inner and outer opening handles. In order to realize the function, the outer opening handle assembly is improved to connect two cables to synchronously control the unlocking functions of the two lock bodies. A newly developed inner opening handle assembly is composed of a handle, a rotating shaft, a bottom plate, a rotating arm and other components and connects two cables to control the unlocking functions of the two lock bodies.

This invention discloses a multi-agent cooperative control method for cross-domain unmanned swarms across land, sea, air, and space, comprising the following steps: S1: Constructing a decoupling model to separate the complexity of cooperative control along three dimensions: hierarchical, frequency-based, and domain-based; S2: Based on the decoupling model, designing a unified cooperative strategy, which generates personalized control objectives adapted to the dynamics and roles of agents in different dimensions according to a dynamically updated cross-domain cooperative situation map; S3: Introducing an event-triggered distributed reinforcement learning mechanism for each agent, ensuring that the agent triggers only when preset triggering conditions are met. This invention, through a hierarchical-frequency-domain triple decoupling model, separates the extreme complexity of cross-domain cooperation along three orthogonal dimensions: task level, control frequency, and physical space, and defines a standard interface to encapsulate the dynamics of heterogeneous platforms, reducing the complexity of system design and analysis.

The utility model discloses a wall-hung adjustable gradienter placing support, which belongs to the technical field of building construction and comprises a vertical keel, positioning components for positioning the use position of the vertical keel are arranged at the upper end and the lower end of the vertical keel, and a vertically movable shell is slidably mounted on the surface of the vertical keel. A mounting assembly for mounting a gradienter is arranged on one side of the up-and-down movable shell, and a sliding assembly is arranged between the up-and-down movable shell and the vertical keel; according to the utility model, through the arrangement of the positioning assembly, the vertical keel can be mounted on a decorative wall surface, through the arrangement of the mounting assembly, the gradienter can be mounted on the up-down movable shell, and a user can adjust the horizontal height of the gradienter by moving the use position of the up-down movable shell; the gradienter can ascend along with the decoration face, operators do not need to manually ascend and descend the elevation, accumulated errors are eliminated, construction efficiency is improved, rework is reduced, and quality impression is improved.

This invention belongs to the field of high-voltage connecting aluminum busbar technology, and specifically relates to a dual-conductor, dual-insulation high-voltage connecting aluminum busbar structure for electric vehicles. It includes: two sets of flat aluminum conductors arranged parallel to each other vertically; a high-temperature resistant insulating tape wrapped around the flat aluminum conductors, allowing the two sets of flat aluminum conductors to form independent positive and negative connecting wires respectively; and an outer insulating layer wrapped around the entire outer side of the positive and negative connecting wires. The manufacturing process of the high-voltage connecting aluminum busbar structure includes the following steps: aluminum busbar extrusion, insulating tape winding, outer insulating layer extrusion, cutting, 3D bending, peeling, conductor punching, labeling, and packaging. The flattened design of this invention significantly reduces vertical space occupation, making it suitable for wiring in narrow areas such as the top and side walls of the battery pack, improving the overall vehicle space utilization. Furthermore, the production process is simplified from 17 steps to 9 steps, reducing carbon emissions and aligning with green manufacturing trends.

The invention discloses a front structure of a vehicle and the vehicle. The front structure includes a hood and a headlight, the hood being disposed over and covering the headlight. The bonnet has a front edge. The headlight includes a front wall, and the front wall has an upper edge adjacent a front edge of the bonnet. A first gap configured to face inwards in the longitudinal direction is formed between the upper edge and the front edge, and at least part of the first gap is shielded by the engine cover. Thus, even if the first gap is enlarged due to downward deviation of the headlight, a user is still not easy to see or notice the first gap, thereby improving aesthetics and quality perception of the vehicle.

The invention discloses a passenger car threshold decorative plate side visibility checking method, and belongs to the technical field of automobile design, and the method comprises the steps: obtaining a ground plane S1 based on the circle centers of front and rear wheels of a vehicle; making a straight line perpendicular to the ground plane S1 based on two points on the outermost edge of the doorsill decorative plate, and making a vehicle profile through the straight line; determining a position point P6 where a simulated observer is located and an eye point position point P7 of the simulated observer based on the vehicle profile, and determining a tangent line L5 tangent to the threshold decorative plate based on the point P7; and comprehensively judging whether the lateral visibility is passed or not based on the lowest point of the battery position. According to the method, the side visibility is checked and evaluated in the design stage of the left and right doorsill decorative plates, the bottom battery or the side wall metal plate flanging of the whole vehicle is in an invisible state from the perspective of a user, and the side user perception quality is improved.

The invention relates to an underwater image compression method and device under an extremely low bit rate and a storage medium. According to the method, after an original underwater image is acquired, a foreground image and a semantic graph are extracted from the original underwater image, a decoded foreground image is acquired by using a preset foreground image high-pixel fidelity compressor based on the foreground image, and the decoded foreground image is acquired based on the semantic graph and the decoded foreground image. Obtaining a reconstructed underwater image by using a preset background image high-perception fidelity compressor; wherein the foreground image high-pixel fidelity compressor comprises a foreground target encoder and a foreground target decoder which are connected in sequence, the foreground target decoder comprises an underwater prior guidance module, and the background image high-perception fidelity compressor comprises a semantic graph encoder-decoder and a foreground-assisted background generation network which are connected in sequence. Compared with the prior art, the method has the advantages that the high-perception fidelity quality of the whole underwater image and the high-pixel fidelity quality of the foreground target are ensured, and the compression efficiency of the underwater image is improved.

This invention discloses a method, apparatus, and device for selecting User Plane Function Entities (UPFs). The method, applied to network devices, includes: receiving a User Plane Function Entity (UPF) service report sent by a Session Management Function (SMF); sending a latency measurement message to a base station via an Access and Mobility Management Function (AMF) based on the UPF service report; obtaining an N3 path latency table and a recommended UPF from the latency measurement message; and selecting a target UPF by the SMF based on enhanced UPF selection parameters. The enhanced UPF selection parameters indicate that the UPF is selected based on the transmission latency of the N3 interface. Through this method, the solution of this invention can select the UPF with the lowest transmission latency for the terminal, improving the user experience.

The invention discloses an image super-resolution reconstruction method based on a SAUGAN network, and relates to the field of image and machine vision technology of artificial intelligence. The invention integrates a GAN network, a self-attention mechanism, an RRDB+ module, and a U‑Net architecture. A multi-scale feature extraction module is designed in the shallow feature extraction part of the generator, which can retain more fine details of the original image; an RRDB+ module is designed in the deep feature extraction part of the generator, and combined with a self-attention mechanism module, the characteristics of the residual network are fully utilized, the global information of the image is paid attention to, and the global feature information is better utilized for super-resolution image reconstruction, so as to obtain an image with better perceived quality; the discriminator adopts a U‑Net architecture to provide global and local feedback to the generator, improve the local details of the image, and maintain the global and local sense of reality. And spectral normalization regularization is used to stabilize the training process, and finally a good effect of local enhancement and suppression of artifacts is achieved.

The application discloses a kind of vehicle-road cooperative perception method and device based on vehicle-mounted point cloud, belong to the field of automatic driving, the method includes: using each vehicle in current time slot in driving environment complexity and communication link data rate of cooperation perception, construct the optimization problem of maximum total perception quality, solve it to obtain the point cloud processing task priority and optimal configuration corresponding to each vehicle;Each vehicle uses corresponding optimal configuration to sample compression and back to roadside server, so that roadside server uses point cloud processing task priority to carry out target detection and issue to each vehicle to carry out environmental perception.The optimization problem in the application considers the problem of communication resources, processing priority and the cooperative perception accuracy of multiple vehicles and roadside server, thereby solving the technical problems of poor perception quality caused by large delay, low accuracy and limited resources of existing automatic driving vehicles.

The vehicle body comprises a side wall assembly and a rear plate, the side wall assembly comprises a side wall inner plate and a side wall outer plate, and the side wall inner plate comprises an inner plate body and a first bent part extending in the width direction of the vehicle body, the side wall outer plate comprises an outer plate body and a second bent part extending in the width direction of the automobile body, the outer plate body is connected with the inner plate body, the second bent part is connected with the first bent part to define a first cavity, and the rear plate comprises a first connecting part extending in the width direction of the automobile body; the two sides of the first connecting part are connected with the first bent part and the second bent part respectively. The lap joint faces of the rear plate, the side wall inner plate and the side wall outer plate are changed from the Y direction to the X direction, so that force borne by welding spots on the welding face under the twisting working condition is changed from normal stripping force to tangential shearing force, the situation that the welding spots are prone to cracking is reduced, the visibility of the welding spots is low, the appearance perception quality of a vehicle is improved, and the service life of the vehicle is prolonged. And the overall strength and torsional rigidity of the vehicle are also improved.

The invention discloses a vehicle side wall assembly and a vehicle, the vehicle side wall assembly comprises a vehicle body side wall outer plate, the vehicle body side wall outer plate comprises a vehicle body side wall outer plate body and a first fin-shaped plate, and the vehicle body side wall outer plate body is provided with a first side surface and a second side surface which face oppositely and comprises a body boundary; the first fin-shaped plate extends out of the body boundary of the vehicle body side wall outer plate body; the side wall decorating part is attached to the first side surface of the vehicle body side wall outer plate body and is provided with an exceeding part extending across the boundary of the body; wherein the first fin plate extends from the body boundary to an edge of the overhang portion, the overhang portion being adapted to be attached to the first fin plate. According to the vehicle side wall assembly and the vehicle, the first fin-shaped plate is introduced to provide rigid support for the exceeding part, extending across the vehicle body side wall outer plate body, of the side wall decorating part, and therefore the perceived quality of the side wall decorating part on the vehicle is improved.

The application provides a frame-level alignment boundary adversarial attack method, system, device and equipment for a sequence recognition model and a medium, and belongs to the fields of computer vision, speech processing and adversarial machine learning. The method comprises the following steps: S1: inputting a sample into a target sequence recognition model in a test stage to obtain an initial reference alignment label sequence; S2: based on the initial reference alignment label sequence, constructing an alignment boundary margin between a reference alignment label and a competitive label; S3: based on the alignment boundary margin, generating a dynamic continuous gating weight by using a smoothing mapping, constructing a margin optimization target with gating weighting and iteratively updating to obtain a candidate adversarial sample; and S4: performing total variation (TV) smoothing under a success maintaining constraint and amplitude scaling search on the candidate adversarial sample to generate a high-fidelity adversarial sample. The application can improve the adversarial attack efficiency and success rate and can be used in sequence recognition model robustness evaluation and privacy protection, copyright protection and the like.

The utility model belongs to the technical field of vehicles, and provides a steering column outer cover structure and a vehicle. The steering column outer cover comprises an outer cover body and a decorative cover detachably connected to the outer cover body, the outer cover body is arranged on a steering column in a sleeving mode, and a first receding hole is formed in the outer cover body; the decoration cover covers the first receding hole and is provided with a second receding hole corresponding to the first receding hole so that an adjusting handle can stretch out of the inner side of the outer cover body. And when the vehicle is subjected to head-on collision, the decorative cover can fall off from the outer cover body. The decorative cover is arranged to shield the first receding hole, the structure of the steering column is prevented from being exposed, the interior appearance of a vehicle can be improved, and the perception quality of a user can be improved; and when a vehicle has a head-on collision, the decorative cover can fall off from the outer cover body, so that the crumple size of the steering column is ensured, the safety of the vehicle is ensured, and a living space is provided for a driver.

The invention provides a learning type image compression method based on stream matching residual reconstruction, and relates to the technical field of image processing and computer vision. The method comprises the following steps: constructing an image compression network model based on a flow matching latent variable refining module; wherein the flow matching latent variable refining module performs residual error reconstruction on quantitative latent variables by using a time continuous microfluidical field, and compensates for fine details lost in a quantization process; constructing a joint optimization objective function based on a rate distortion loss function and a stream matching loss function, and performing end-to-end training on the whole image compression network model by using an input image; and performing image compression on a to-be-compressed image by using the trained image compression network model. By adopting the method, fine details lost in the quantization process can be effectively compensated under the condition that reasoning delay is not remarkably increased, and the structural fidelity and perception quality of image reconstruction are remarkably improved.