178results about How to "Improve conversion performance" patented technology

The invention discloses an image domain conversion network based on generative adversarial networks (GAN) and a conversion method. The image domain conversion network includes a U-shaped generative network, an authenticity authentication network and a pairing authentication network. An image domain conversion process mainly includes the following steps: 1) training the U-shaped generative network,and establishing a network model of the U-shaped generative network; and 2) normalizing a to-be-converted image, and then input the image into the network model established through the step 1) to complete image domain conversion of the to-be-converted image. According to the image domain conversion network, an image domain conversion task of a local region in the image can be achieved, image local-domain conversion quality is high, judgment ability of the network is high, stability of image conversion is high, and authenticity of a generated image is greatly improved.

A bidirectional DC-DC converter applies to a charging and discharging system equipped with a battery and especially to power fluctuations arising from the charging and discharging of the battery. The bidirectional DC-DC converter has a first full-bridge switching unit, transformer, resonance unit, second full-bridge switching unit, and frequency change control module. The first full-bridge switching unit connects with a first DC power source. The transformer has a primary side connected to the first switching unit to receive a power from the first DC power source and has a secondary side connected to the second full-bridge switching unit. The resonance unit connects with the transformer's secondary-side winding and receives power to produce resonance. The second full-bridge switching unit connects with a second DC power source. The frequency change control module instructs the switching units to perform bidirectional buck-boost switching and changes the operating frequency to adjust voltage gain.

The invention relates to a training method and device of a neural network model for image processing. The method comprises that video frames adjacent in time are obtained; the neural network model output intermediate images corresponding to the video frames respectively; change of the video frames in relatively earlier time and optical flow information of the video frames in relatively later timeare obtained; and image formed by changing the intermediate image, corresponding to the video frame in relatively earlier time, according to the optical flow information is obtained; time loss betweenthe intermediate image corresponding to the video frame in relatively later time and an image after change is obtained; characteristic loss between the intermediate image and a target characteristicimage is obtained; and the neural network model is adjusted according to the time loss and the characteristic loss, the step of obtaining the video frames adjacent in time is returned to, and trainingis not stopped until the neural network model satisfies a training ending condition. Thus, the neural network model obtained by training has a better video characteristic conversion effect.

The invention discloses a video frame rate up conversion method based on a sparse redundant representation model. The method comprises the steps that (1) a forward optical flow vector vt-1,t+1(i, j) at any pixel location (i, j) between front and back frames ft-1 and ft+1 is estimated; (2) according to the assumption of movement symmetry, based on the forward optical flow field vt-1,t+1 of the front frame ft-1, bidirectional three-dimensional recursive searching is carried out pixel by pixel, and the optimal optical flow vector vt, t+1(i, j) at the current pixel position (i, j) is calculated (3) an optical flow vector in the optical flow field vt,t+1 is smoothened pixel by pixel to ensure the optical flow consistency in a local area and prevent a false optical flow vector; (4) front and back frames ft-1 and ft+1 and the optical flow field vt,t+1 are used to construct a sparse reconstruction model of the current frame ft; (5) a regularization parameter and a sparse representation dictionary in the sparse reconstruction model are estimated; and (6) a nonlinear optimization method is used to solve the sparse reconstruction model to acquire the current frame estimation. Compared with the prior art, the method provided by the invention can effectively improve the performance of video frame rate up conversion.

The invention discloses a multi-to-multi speaker conversion method based on STARGAN and ResNet, which comprises a training stage and a conversion stage, wherein the STARGAN and the ResNet are combinedto achieve a voice conversion system; a ResNet network is utilized to solve the problem of network degradation existing in the STARGAN; the learning capability of semantics and the synthesis capability of a voice spectrum of a model can be better improved; therefore, the individuality similarity and the voice quality of converted voice are better improved; meanwhile, the data is standardized by using the Instant norm; the noise generated in the voice conversion process can be well filtered; the problem of poor voice similarity and natural degree of the converted voice in the STARGAN is solved. A voice conversion method with high quality can be achieved. The method can achieve voice conversion under the condition of non-parallel text, and does not need any alignment process in the trainingprocess, a conversion system with multiple source-target speaker pairs is integrated in a conversion model, namely, the multi-to-multi speaker conversion is achieved.

The invention relates to a double speed axial angle digital converter that uses DSP as central process unit. The process includes the following steps: the excitation signal of rotating transformer accessing to DSP through synchronizing signal generating circuit, and the rotating transformer signal of raw and fined channels accessing to DSP through signal adjusting circuit, the DSP first gaining compounding axial angle location and speed, and realizing digital quantity output through interface circuit after taking angle error correction and combining, the random dither signal generated by DSP would be accessed to signal adjusting circuit through filtering circuit and rotating transformer. The speed ratio and converting accuracy could be adjusted. The whole converting process could be realized by software arithmetic. Thus, it could not only reduce hardware circuit, lower cost, but also could adjust diameter by software and improve adaptability.

Disclosed is a catalytic material for purifying an exhaust gas component. The catalytic material comprises a composite oxide which contains, as essential components, zirconium (Zr) and neodymium (Nd), and further contains a rare-earth metal R other than cerium (Ce) and neodymium (Nd), wherein each of the zirconium, neodymium and rare-earth metal R constituting the composite oxide is contained, in the form of oxide, in such a manner that a ratio of Nd2O3 / (ZrO2+Nd2O3+RO) is 3 mol % or more, and a ratio of (Nd2O3+RO) / (ZrO2+Nd2O3+RO) is 33 mol % or less. The catalytic material of the present invention can oxidize / burn PM in a short period of time, while suppressing CO emission during the burning of the PM, and can achieve further enhanced NOx conversion performance.

A catalyst for purifying exhaust gases includes a substrate, and a catalytic layer. The catalytic layer includes a lower catalytic layer, a first upper catalytic layer, and a second upper catalytic layer. The lower catalytic layer being loaded with Pd and / or Pt is formed on the substrate. The first upper catalytic layer being loaded with Pd covers an upstream side of the lower catalytic layer, and exhibits a concentration of loaded Pd that falls in a range of from 4.5 to 12% by mass when the entirety of the first upper catalytic layer is taken as 100% by mass. The second upper catalytic layer being loaded with Rh covers a downstream side of the lower catalytic layer.

A ferrous-metal-alkaline-earth-metal mixed silicate based phosphor is used in form of a single component or a mixture as a light converter for a primarily visible and / or ultraviolet light emitting device. The phosphor has a rare earth element as an activator. The rare earth element is europium (Eu). Alternatively, the phosphor may have a coactivator formed of a rare earth element and at least one of Mn, Bi, Sn, and Sb.

The invention relates to a method for manufacturing a front side grid line electrode of a solar battery. The technical scheme of the method is as follows: the method for manufacturing the front side grid line electrode of the solar battery is characterized in sequentially comprising the following steps of: 1. acting a rectangular tooth press mould with proper aspect ratio on the front side of the thin polymer film-coated solar battery to form into a rectangular tooth pattern corresponding to the press mould; 2. removing a polymer film at a groove and reducing a reflection coating by means of chemical corrosion; and 3. performing a selective metal chemical plating technology and an electroplating technology to manufacture the grid line electrode. With the scheme, the method overcomes the disadvantage of the prior art, provides a new scientific and reasonable method for manufacturing the front side grid line electrode of the solar battery, prevents a battery substrate from being damaged, leading the press mould to be hardly deformed, being suitable for recycle in industrialization, and reducing the cost.

The invention provides an integration and conversion system for network security data. The system comprises a network security detection tool, an agent module, a configuration module, a data acquisition device and a data sorting and merging processing module, wherein the number of the network security detection tool is at least one, each network security detection tool comprises at least one agent module and at least one configuration module; the configuration module comprises a configuration document, and the configuration document comprises a data format conversion rule described by a regular expression; the network security detection tool obtains the network security data of the located network; the agent module obtains the network security data from the network security detection tool according to the configuration document, extracts security event data in the network security data according to the regular expression and converts the extracted security event data into a uniform data format corresponding to the security event; and the data acquisition device collects the security event data formed by conversion of each agent module, and sends the security event data to the data sorting and merging processing module, so that the data sorting and merging processing module can sort and merge the data.

The oxidation of nitrogen oxide (NO) in an oxygen-containing exhaust gas flow from a diesel or other lean-burn engine may be catalyzed using particles of co-precipitated and calcined manganese (Mn), cerium (Ce) and zirconium (Zr) mixed oxides. In preferred embodiments, the molar ratios of Mn, Ce and Zr to the total amount of base metals in the ternary mixed oxide catalyst are in the range of 0.25-0.35, 0.40-0.50 and 0.20-0.25, respectively. Further, this ternary mixed oxide catalyst is less susceptible to sulfur poisoning than previously-disclosed binary mixed oxide catalysts. The ternary mixed oxide catalyst may also be regenerated—and the inhibiting effect of SO2 reversed—by briefly exposing the catalyst to a reducing exhaust gas environment.

The invention relates to the TTS (Text-To-Speech) technical field, and discloses a method and system of transforming speech. The method comprises: obtaining speech signals of a source speaker; extracting spectrum envelope characteristics and fundamental frequency characteristics of the speech signals; transforming the spectrum envelope characteristics according to a preset spectrum envelope transformation model to obtain transformed spectrum envelope characteristics; and generating speech signals of a target speaker according to the transformed spectrum envelope characteristics and the fundamental frequency characteristics. The method and system can effectively improve the timbre of transformed speech.

The invention discloses a method and a system for converting a Word format file into an EPUB format file. The method is for a Word file in a. Docx format. By identifying and processing the catalog ofthe Word source file, a catalog structure of the source Word file can be identified. An EPUB electronic book can be automatically generated, and the method comprises the steps of Word file analysis, XML file analysis, Word file splitting, HTML file generation and EPUB file generation. According to an EPUB e-book automatic generation method capable of identifying the source Word file directory, theproblems that in the prior art such as poor conversion effect tedious operation of manually adding title directories and low efficiency are solved .The integrity of document content is guaranteed. The conversion effect of documents is improved and working efficiency is improved.

The invention discloses an intelligent-document-platform-based multi-channel information acquisition and exchange method. The intelligent document platform comprises a server and a database, wherein various application services are deployed on the server; an application program is installed on a variety of terminal equipment which serve as information acquisition channels; the application services communicate with the application program which is distributed on the terminal equipment by calling a unified interface; and the interface is an interface of an application program interface (API) function, which is irrelevant to the terminal equipment. By the multi-channel information acquisition and exchange method, a public data processing center is established for different acquisition channels according to different terminal equipment and acquisition environments; therefore, the information processing and converting capacity is improved. The information is acquired through the method, the information acquisition and exchange time can be greatly shortened, the information flow is accelerated, and the information acquiring and processing period is reduced.

The invention relates to the technical field of direct-current sampling circuits, in particular to a direct-current high-voltage isolation sampling circuit which comprises buses to be sampled, an A / D (analog / digital) conversion circuit, an optical coupled isolation circuit and a signal processing unit. A voltage division sampling circuit is serially connected among the buses including positive buses and negative buses, an output end of the voltage division sampling circuit is connected with an input end of the A / D conversion circuit, an output end of the A / D conversion circuit is connected with an input end of an optical coupled isolator, and an output end of the optical coupled isolator is connected with the signal processing unit. Compared with the prior art, the direct-current high-voltage isolation sampling circuit has the advantages that signal processing links are few, and the sampling precision is improved; the A / D conversion precision is improved owing to the independent A / D conversion circuit; the direct-current high-voltage isolation sampling circuit is simplified, and the quantity of devices is lowered; and as signals are isolated, the requirement on an optical coupler is greatly lowered, and the cost of the direct-current high-voltage isolation sampling circuit is saved.

Disclosed a conversion device for a quick charger, comprising an input end, a voltage regulator module, a recognition module, a control module and an output end, wherein, the voltage regulator module receives a power voltage through the input end to stabilize a voltage, and outputs the power voltage after voltage stabilization to the recognition module and the control module; the control module reads a conversion voltage of the output end, and produces a conversion signal group according to the conversion voltage to output to the recognition module; and the recognition module produces a handshake signal and a voltage recognition signal according to the conversion signal, and outputs through the input end. Through the device and method, equipment that are not quickly charged may also be charged by the quick charger, so that an application scope of the quick charger is expanded.

The invention discloses a model training method, a video fuzzy frame conversion method, equipment and a storage medium, and relates to the field of artificial intelligence. The method comprises the steps of obtaining a sample video; extracting a training data pair from the sample video, wherein the training data pair comprises a sample clear frame and a sample fuzzy frame, and the sample clear frame and the sample fuzzy frame belong to the same video frame interval; inputting the training data pair into a generator of an image generation model to obtain a sample generation frame and sample optical flow information output by the generator; inputting the sample generation frame into a discriminator of the image generation model to obtain a discrimination result output by the discriminator; and training an image generation model according to the training data pair, the sample generation frame, the sample optical flow information and the discrimination result. When the image generation model obtained by training in the embodiment of the invention is used for converting the blurred video frames in the video, the time sequence consistency between the video frames can be kept, the conversion effect of converting the blurred video frames into the clear video frames is improved, and the quality of the converted video is further improved.

The invention relates to a household electrical appliance intelligent voice interaction control method and apparatus. The method comprises the following steps: an active interaction step, i.e., a control terminal acquiring environment parameters and transmitting the environment parameters to a server, the server, according to the environment parameters, bringing forward an adjustment scheme and issuing the adjustment scheme to a control terminal, and after a command allowing adjustment is received, the control terminal executing the adjustment scheme; and a passive interaction step, i.e., after a voice control function of the control terminal is started, the control terminal monitoring voice information, identifying the name of the control terminal and also transmitting voice information fragments after the name of the control terminal to the server for identifying a control instruction, and the server then issuing the control instruction to the control terminal. According to the invention, active adjustment and voice command receiving adjustment are combined, the voice information is identified through two voice identification links, the speed and the conversion effect are improved, household electrical appliances are controlled through voice, voice interaction can be provided, and the household electrical appliance voice control effect and user experience are improved.

The invention provides a speech conversion method based on asymmetric speech database conditions of the speaker model alignment. The method includes firstly training spectrum characteristics of an original speaker and target speaker respectively to acquire speaker modules, utilizing speaker module parameters to find the conversion function between the original speaker feature vector and auxiliary vector and the conversion function between the auxiliary vector and the target speaker feature vector, and finally utilizing the two conversion functions to figure out the conversion functions between the original speaker and the target speaker. During the speech conversion, the method of speaker model alignment is adopted, and the speech conversion effect is further improved by combining the speaker model alignment and a Gauss mixed model. According to the experimental result, the effects of spectrum distortion and relevance and speech conversion quality and similarity are better as compared with those of a traditional speech conversion method based on INCA.

The invention relates to a novel winding method for high-frequency transformer coils, which comprises: firstly, dividing the winding number of a primary coil and a secondary coil into a plurality of equal parts; and secondly, winding a first equal part of the primary coil and winding the winding number of first equal parts of the secondary coil after the winding number of the first equal parts of the primary coil is wound, so that each equal part of the primary coil and each equal part of the secondary coil are wound in staggered form until all the equal parts of the primary coil and the secondary coil are wound, wherein the secondary coil adopts shunted and parallel winding of two or more strands of coil; and the parallel winding strand number of the secondary coil is calculated according to the current for the secondary coil to pass through. The transformer coils wound by the novel winding method can greatly improve the conversion efficiency; when the transformer coils are applied to switching power adapters, the energy and the electricity can be greatly saved and the environment-friendly requirement can be completely met; simultaneously the energy efficiency loss of transformers wound by the novel winding method is low, the transformers cannot generate high heat, and the safety requirement of electrical appliance components in the adapters does not need to be too high, thereby the loss and the volume can be reduced and the production cost can be saved.

Traveling-wave optoelectronic wavelength conversion is provided by a monolithic optoelectronic integrated circuit that includes an interconnected traveling-wave photodetector and traveling-wave optical modulator with a widely tunable laser source. Either parallel and series connections between the photodetector and modulator may be used. An input signal modulated onto a first optical wavelength develops a traveling wave voltage on transmission line electrodes of the traveling-wave photodetector, and this voltage is coupled via an interconnecting transmission line of the same characteristic impedance to transmission line electrodes of the traveling-wave optical modulator to modulate the signal onto a second optical wavelength derived from the tunable laser. The traveling wave voltage is terminated in a load resistor having the same characteristic impedance as the photodetector and modulator transmission lines. However, the interconnecting transmission lines and the load resistor may have different impedances than the photodetector and modulator.

The invention relates to a multi-energy microgrid distributed scheduling method based on a potential game, which can give consideration to the benefits of different energy operators, can improve the decision autonomy of the energy operators, and can model the scheduling problem of a multi-energy microgrid system with electric energy, gas energy, G2P, P2G, energy storage and other devices into a game problem among subnets. A multi-energy microgrid scheduling model based on a potential game is established, and a potential function is constructed; and according to the method, the autonomy and economy of each subnet in the multi-energy microgrid are effectively improved, the method better conforms to the actual situation, and the introduction of G2P and P2G in the model can effectively improvethe overall economy of the system. The invention aims to solve the problems of information complexity and asymmetry existing in multi-energy microgrid scheduling, realize the autonomy of decision making of each energy subnet operator, and provide a solution for a manager of the multi-energy microgrid to search for an optimal strategy of multi-energy distributed scheduling.

The invention relates to an image processing method and device, a model training method and device, a medium and equipment. The image processing method comprises: obtaining a to-be-processed image belonging to a first image category and a reference image belonging to a second image category; encoding the to-be-processed image to obtain a content feature map corresponding to the to-be-processed image; encoding the reference image to obtain a category feature parameter corresponding to the second image category; performing category conversion on the content feature map through the category feature parameters to obtain an intermediate image; and decoding the intermediate image to reconstruct a target image corresponding to the to-be-processed image and belonging to the second image category.According to the scheme provided by the invention, the image conversion effect is improved.

The invention provides a system for converting distributed PPT (Power Point) files into swf (small web format) files, which comprises clients, master servers and slave servers, wherein the clients are used for transmitting PPT files (to be converted into swf files) to the master servers to request for conversion; each master server is used for receiving a request of a corresponding client, dividing a PPT file into a plurality of single pages of PPT files, controlling the slave servers to carry out conversion, and outputting the obtained conversion results to the client; and each slave server is used for receiving the conversion notice of a corresponding master server, and converting a single page of PPT file into an swf file. The invention also discloses a method for converting PPT files into swf files. By using the system and method for converting PPT files into swf files disclosed by the invention, the problem that in the prior art, the conversion efficiency is difficult to be expanded is solved, and the slave servers are easy to be added for promoting the conversion performance, therefore, the horizontal scalability is good.

The invention discloses an internal heating type dryer for in-furnace coking coal dehumidification, which comprises a dryer outer frame, a drying area on the upper portion in the dryer outer frame, a fluidized bed drying heating area under the drying area and an air distribution area on the lower portion in the dryer outer frame. The drying area is provided with an exhaust space (10) and a heat exchange face, a coil heat exchanger (11) is arranged in the fluidized bed drying heating area, a fluidized air chamber (9), an air distribution plate (8) and a hood (12) are arranged in the air distribution area, and the air distribution plate (8) and the hood (12) form a module structure connected with the fluidized air chamber (9). The internal heating type dryer is simple in structure, good in exchangeability, convenient to overhaul, reliable in operation and reliable and flexible in dehumidification control adjustment, saves energy, is simple in system adjustment and responses quickly to heat adjustment of the heat exchanger. A large drying system can be formed by a plurality of dryers according to the production requirement, and only one backup module is needed to ensure normal production of a humidifying system, so that scale of spare equipment of the existing humidifying system is reduced.