16results about How to "Improve decoupling ability" patented technology

The invention relates to a load prediction method TSR-MAEN based on decomposition representation and an attention encoder network. Firstly, the algorithm provides a trend-seasonal decomposition representation method based on comparative learning so as to enhance the decoupling performance of a model on a trend component and a seasonal component of a load sequence and the representation capability of decomposition characteristics. Secondly, in order to deeply mine and decompose time sequence information contained in a feature hidden space and dependency between features, the algorithm proposes a mixed attention encoder network; through segmented embedding, a trend-season dynamic projection mixed self-attention mechanism and a multi-head mixed output mechanism, trend and season multi-dimensional features after load sequence decomposition are effectively utilized to express and enhance the long-range dependence information extraction capability of the model. Experiments show that the TSR-MAEN algorithm provided by the invention has excellent performance in an electrical load prediction task of oil and gas development, and the prediction error is reduced by 8.51% compared with that of an optimal baseline algorithm.

The invention discloses an iteratively optimized multi-component voice conversion method based on enhanced decoupling, and the method comprises the steps: respectively extracting rhythm, content and tone information in voice and pitch information in a pitch contour through employing a multi-channel encoder group with an information bottleneck, and generating the voice through employing a decoder in combination with the encoder group; random resampling and pitch enhancement are adopted to selectively eliminate rhythm or pitch information flowing into an encoder, and then the speech de-entanglement capability of the model is enhanced through a multi-task training method of cooperative pitch contour reconstruction, adversarial speaker classifier group, content feature reconstruction and content feature-rhythm feature mutual information minimization. Besides, the training process is simplified through a two-stage method, the convergence efficiency of the model is improved, and the method aims at achieving high-quality multi-component voice conversion so as to solve the problem that a traditional voice conversion method can only decouple timbre and other voice components are still entangled together.

The invention discloses an automatic testing method and system, and relates to the technical field of software testing. The test method comprises the following steps: establishing communication connection between a script end and a client; the script end compiles an SDK interface automatic test case, packages the SDK interface automatic test case and related information into request information in a Json format, and sends the request information to the client; the client side receives the request information transmitted by the script side, analyzes the request information and determines calling information; the analyzed request information is converted based on the calling information, so that the analyzed request information conforms to the native language type corresponding to the client side; performing SDK interface testing based on the converted request information and calling information, packaging test result data into a Json format, and returning the test result data to the script end; and the script end receives the test result data which is returned by the client and is packaged into the Json format, and generates a corresponding test report according to the test result data.

The invention discloses a fast reflector hybrid control method based on a multivariable extended state observer, and relates to the technical field of precise motion control. The control method comprises the following steps: obtaining transfer functions of X and Y axes of a voice coil type fast reflector system by adopting a system identification method, and obtaining a fast reflector model; designing a model auxiliary multivariable expansion state observer according to the obtained fast reflector model; and designing a mixed state feedback controller on the basis of the model-assisted multivariable extended state observer. According to the method, the extended state observer is combined with hybrid control, known model information is added into one observer to design a model-assisted multivariable extended state observer, and X and Y axes are observed at the same time, so that the estimation error of disturbance of each channel is reduced, and the decoupling capability and robust performance of the extended state observer are improved.

The present invention discloses a method and system for enhancing the time-frequency aggregation of non-stationary signals, and relates to the technical field of signal time-frequency aggregation enhancement, including collecting non-stationary signals for preprocessing; performing time-frequency analysis on non-stationary signals, and defining weight operators; optimizing the time-frequency analysis and weight operator objective functions, and constructing an amplitude modulation and frequency modulation composite signal model to optimize the time-frequency aggregation of non-stationary signals; evaluating the time-frequency aggregation of the optimized non-stationary signals and outputting them, and storing the data. The present invention proposes a high-order weighted time-frequency sparse representation, defines a weight operator, and uses a fast iterative shrinkage threshold algorithm for optimization to establish an amplitude modulation and frequency modulation composite signal model. The present invention has higher time-frequency aggregation and robustness for fast-changing signals, and the time-frequency results are more accurate than the theoretical instantaneous frequency. Compared with other time-frequency methods, the time-frequency aggregation and decoupling performance are improved.

The utility model discloses a rear suspension structure and a vehicle, the rear suspension structure comprises a suspension support, a first side of the suspension support is provided with a plurality of first suspension assemblies, a second side of the suspension support opposite to the first side is provided with a plurality of second suspension assemblies, the first suspension assemblies extend along a first direction, and the second suspension assemblies extend along a second direction. The second suspension assembly extends in a second direction. The first suspension assemblies are installed on the first side of the suspension support, the second suspension assemblies are installed on the second side of the suspension support, the supporting performance of the suspension support on each first suspension assembly and each second suspension assembly is better, and the stability of the first suspension assemblies and the stability of the second suspension assemblies are higher. And the decoupling capability of the rear suspension structure is improved, so that the overall rigidity of the rear suspension structure is improved.

This invention provides a force sensor comprising an annular housing, a top cover and a base respectively fixed to one end of the housing, a loading platform fixed to the top cover near the base and spaced apart from the housing and the base, at least three force-measuring beam assemblies spaced apart from each other around the periphery of the loading platform, and a first sensor device and a second sensor device disposed on each force-measuring beam assembly. Each force-measuring beam assembly includes a horizontal force-measuring beam fixed to the periphery of the loading platform and a vertical force-measuring beam fixed to the loading platform near the base. The horizontal and vertical force-measuring beams in each force-measuring beam assembly are perpendicular to each other and located in the same plane. All force-measuring beam assemblies are rotationally symmetrically distributed. The force sensor of this invention can improve its decoupling effect and simplify data processing during measurement.

The application provides a program sequence recommendation method and system based on group interest multi-level decoupling, and belongs to the technical field of personalized recommendation. The method comprises the following steps: a group sequence encoder is used to encode a group interaction sequence processed by an embedding layer to obtain a group representation; a group-level interest decoupling module is used to decouple the interest representations of different members from the group representation; a group variational autoencoder in a member-level interest decoupling module is used to further decouple the interest representations of different members decoupled by the group-level interest decoupling module to obtain decoupled representations of different members and a group consensus representation; the decoupled representations of different members and the group consensus representation are used for next item prediction to obtain candidate item scores of the group at the next moment, and the group is recommended. By using the application, the problem of poor IPTV program recommendation accuracy and robustness of the existing family group can be solved.

The invention belongs to the field of artificial intelligence and finance, and relates to an anonymous voice conversion method, which comprises the following steps: preprocessing a voice data set to obtain a voice enhancement data set; performing first-stage training on a pre-constructed initial speech synthesis model by using the speech enhancement data set to obtain a trained speech synthesis model; performing second-stage training on the pre-constructed initial anonymous identity generator to obtain a final anonymous identity generator; replacing a speaker encoder in the speech synthesis model with the final anonymous identity generator to obtain an anonymous speech conversion model; and performing voice conversion on the to-be-converted voice based on the anonymous voice conversion model to generate anonymous voice. The invention further provides an anonymous voice conversion device, computer equipment and a storage medium. In addition, the invention also relates to a block chain technology, and the to-be-converted voice can be stored in a block chain. According to the invention, the diversity of anonymous voice can be improved, and privacy security is protected.

The invention belongs to the field of event-driven architectures, and provides a workflow-based event management method and device, and the technical scheme is as follows: arranging the response process of an event by using a workflow mode, introducing a process-driven capability to define an event consumption process, and introducing the process-driven capability. Flexible cooperation configuration and an intelligent pushing mechanism are provided, when the state of a business object changes, an event producer engine sends an event, an event operation center is connected with cooperation between business objects in series, an event consumption process is started and operated, and an event consumer engine receives and feeds back an event consumption result; therefore, each service system only focuses on own service processing and does not need to pay attention to cooperative logic with other related service systems, so that full decoupling of the service processing logic between the service systems is achieved, and the decoupling performance, expansibility and response speed of the system are improved.

The invention discloses a voice conversion method, device and equipment and a computer readable storage medium, based on a conditional VAE framework, tone information and content information of voice are extracted by using an encoder, then the voice is synthesized through a generator, and meanwhile, in order to decouple the voice into the content information and the tone information, the voice is converted into the content information and the tone information. A pre-trained self-supervised speech representation learning model is introduced to improve the content decoupling capability, a feature vector xssl which is output by the self-supervised speech representation learning model and contains source content information C is used as decoupling of conditional constraint content information C through a stream model, the forward and reverse characteristics of the stream model are utilized, two KL divergence criteria are used for guiding model parameter continuation, and the self-supervised speech representation learning model is obtained. According to the method and the device, the content information is ensured to be lossless, the tone information is modeled as standard normal distribution, the problem of tone leakage is reduced, the decoupling capability of the content information is improved, and the technical problem of tone leakage is solved.

This invention discloses an indoor data acquisition and positioning method, positioning system, and storage medium based on a neural network algorithm. It employs a multi-terminal collaborative architecture to achieve a closed-loop positioning process for small-scale indoor scenes. The system uploads indoor planar images via a front-end device and performs scale calibration to establish a mapping relationship between pixel coordinates and physical space. In acquisition mode, it acquires multi-source signals in real time and expands the effective sample. After back-end data processing and storage, the algorithm extracts signal features through a neural network, performs adaptive weight allocation and feature fusion, and trains to generate a fingerprint database model. In positioning mode, it performs signal matching and coordinate calculation based on this model, and finally, the positioning results are visualized and displayed on the front end. This invention requires no dedicated acquisition equipment or base station deployment, and has the advantages of high positioning accuracy, low deployment cost, convenient maintenance, and user-friendly operation, making it suitable for various small-scale indoor positioning needs.

A linear anti-disturbance control method for suppressing coupling interference of macro-micro linear motion platforms belongs to the field of motor control. The present invention solves the problem that the control law design process of the existing coupling interference suppression method is cumbersome and the decoupling effect is poor. The main technical scheme adopted by the present invention is: step one, establish the mechanical equation of the macro-micro linear motion platform; step two, perform Laplace transform on the established mechanical equation to obtain the equivalent coupling model of the macro-micro linear motion platform; step three, establish the spatial equation of the linear motor and the voice coil motor, estimate the state variables and coupling errors of the macro-motion platform based on the spatial equation of the linear motor and the extended state observer, and estimate the state variables and coupling errors of the micro-motion platform based on the spatial equation of the voice coil motor and the extended state observer; and design a linear state error feedback control law based on the estimated state variables and coupling errors. The method of the present invention can be used to suppress the coupling interference of macro-micro linear motion platforms.

The invention discloses a micro-seismic monitoring event classification method, and belongs to the technical field of micro-seismic monitoring. According to the method, features in channels are concerned through a single-channel classification model based on deep learning, a microseismic event classification model based on machine learning is used, and obtained single-channel information prediction classification results are integrated. The method is good in interpretability and decoupling performance, and a single-channel classification model only pays attention to the features of a single channel, so that during migration deployment, compared with an end-to-end classification model directly taking multiple channel data as input, the training time and hardware cost are remarkably reduced, and the migration cost is low.

The invention relates to a vehicle which comprises a vehicle frame, a power assembly and a suspension device, the suspension device comprises a bracket and a suspension assembly, and the bracket comprises a supporting plate; the two connecting plates are connected to the two ends of the supporting plate respectively, extend upwards and are connected with a frame respectively; the lower end of the suspension assembly is installed on the upper side of the supporting plate. The power assembly comprises a first installation point and two second installation points, the two second installation points are located on the left side and the right side of the power assembly, the first installation point is located on the bottom wall of the power assembly, the front portion of the power assembly is fixed to the vehicle frame through the two second installation points, and the rear portion of the power assembly is fixed to the upper end of the suspension assembly through the first installation point. The suspension assembly is connected to the first mounting point, so that the elastic center of the suspension assembly is closer to the torque shaft, and the decoupling effect is improved. The bracket can solve the problem that no frame used for being connected with the suspension assembly exists below the bottom wall of the power assembly, the suspension assembly is installed on the bracket, and the problem can be solved by connecting the bracket to the frame.

The present invention discloses a real-time synchronization system and method for massive data, wherein the method comprises: using Canal to obtain the incremental log of the source database, generating incremental data according to the incremental log, and pushing the incremental data as a message record to the message queue; using the incremental synchronization system to obtain the message record in the message queue, distributing the message record to the corresponding message subqueue for concurrent consumption according to the preset distribution rule; using the incremental synchronization system to process the message record in the message subqueue and then storing it in the target database. The present invention has high decoupling and can ensure the real-time and consistency of massive data synchronization.