8results about How to "Drift suppression" patented technology

The invention discloses a dynamic modeling analysis system for combustion atmosphere data of a biomass oxygen-enriched combustion boiler, and relates to the technical field of dynamic modeling analysis. The system comprises a working condition self-adaptive feature extraction module, a time sequence maintaining model training module, an online self-adaptive correction module and an incremental model updating module. Through a working condition adaptive feature extraction module, a time sequence maintenance model training module, an online adaptive correction module and an incremental model updating module, adaptive feature vectors are generated through feature extraction, a basic model is obtained through training, and online correction and incremental updating are performed by combining feature vector verification threshold judgment. And finally, iterative upgrading is completed through an incremental learning fine tuning model, the reliability of dynamic modeling analysis of the combustion atmosphere data of the biomass oxygen-enriched combustion boiler is improved, and the problem that in the prior art, the reliability of dynamic modeling analysis of the combustion atmosphere data of the biomass oxygen-enriched combustion boiler is low is solved.

This invention discloses a control method, system, device, and storage medium based on inertial sensor attitude calculation, including: acquiring acceleration and angular velocity; calculating motion attitude data based on acceleration and angular velocity; calculating the relative attitude angle change from the motion attitude data; mapping the relative attitude angle change to relative directional displacement to obtain relative directional displacement data and sending it to the host. This application utilizes the feature of existing smart wearable devices equipped with inertial sensing units to convert motion attitude into relative attitude angle changes, providing an attitude calculation method. This application can use BLE wireless transmission, balancing mobility and portability, making it suitable for wearable devices such as smart glasses. Furthermore, this application introduces complementary filtering, dead zone, and output filtering steps into the attitude calculation method, improving the accuracy of attitude calculation. It also adds an automatic calibration step, which can suppress drift and jitter, reducing the false touch rate.

This invention discloses a method for the collaborative restoration of structure and texture in Dunhuang murals based on structural uncertainty, belonging to the field of image restoration technology. The method includes: a structure prediction step, which generates a structure reconstruction result through an encoder-decoder architecture based on codebook quantization and simultaneously outputs pixel-level structural uncertainty parameters; a texture reconstruction step, which models the texture completion process as texture flow propagation on a graph structure and adaptively adjusts the propagation range and constraint strength according to the structural uncertainty parameters; and a fusion step, which fuses structural features and texture features at multiple scales and dynamically adjusts the fusion strength according to the structural uncertainty parameters to generate the final restoration result. This invention achieves effective linkage between structure prediction and texture reconstruction through structural uncertainty parameters, suppressing the transmission of structural errors to the texture generation process when there is large-area damage or severe information loss, while maintaining overall consistency and natural detail, making it suitable for the digital restoration of Dunhuang murals.

The present disclosure provides a model federated learning method, device, electronic equipment and storage medium, relating to the technical field of deep learning. The method comprises: the server issuing a residual weight to be updated and a global low-rank factor to be updated; each client performs a round of iteration of a model to be trained; the server aggregates the low-rank first factor and the low-rank second factor updated by each client respectively, and a global low-rank factor; the server aggregates and fuses the low-rank first factor and the low-rank second factor to obtain a dense fusion result; the low-rank first factor and the low-rank second factor are fused and aggregated to obtain a target dense aggregation result; the difference between the target dense aggregation result and the dense fusion result is determined to obtain a residual weight; and the residual weight and the global low-rank factor are issued. The present disclosure can inhibit client drift and catastrophic forgetting in a non-independent and identically distributed environment.

The utility model relates to a split charging device, specifically disclose a powder ration split charging device, including conveying device, and be located the blanking device above conveying device, the blanking device includes the blanking pipe of verticality and located the blanking pipe directly above the second conveying belt, and the feeding device of with blanking pipe upper end side wall intercommunication, the material blocking device of being located blanking pipe lower extreme, be located extruding device in blanking pipe, and the annular pressing plate of being located blanking pipe lower end portion, the powder ration split charging device of the utility model can avoid the powder of dispersing in the split charging process under the premise of guaranteeing split charging precision as far as possible.

This application discloses a multi-axis coordinated motion control method and apparatus for a dynamic hovering arc radiotherapy device, relating to the field of equipment control, and particularly to the field of radiotherapy equipment control. The method includes: acquiring the actual position information of multiple execution axes; determining the target control state of the multiple execution axes at the current moment based on the preset target motion trajectory and the actual position information; when the target control state is a linked motion state, controlling the multiple execution axes to perform coordinated motion according to the target motion trajectory; when the target control state is a coordinated locking state, controlling at least one first execution axis among the multiple execution axes to enter a position locking mode, and controlling at least one second execution axis among the multiple execution axes to enter a position maintenance mode. This application solves the technical problem of low multi-axis synchronization accuracy in existing multi-axis coordinated devices (such as dynamic arc radiotherapy devices) during dynamic hovering.

This invention relates to the field of electrocardiogram (ECG) signal denoising technology, and discloses an ECG signal denoising method based on multi-scale deep learning, including a training phase and an inference phase. The training phase includes the following steps: Step S101, acquiring approximately noise-free ECG data as a reference signal; Step S102, preprocessing the dataset; Step S103, constructing a multi-scale one-dimensional convolutional reconstruction network; Step S104, defining a morphology-aware composite loss function; Step S105, training the network using the training set. Addressing the problem that traditional frequency domain filtering methods struggle to balance signal fidelity and noise suppression due to overlap between noise and ECG signal frequency bands, this invention constructs a multi-scale feature extraction network and combines it with multi-scale time-frequency consistency constraints. This effectively suppresses various complex interferences, including baseline drift and power frequency interference, significantly reducing the risk of introducing time-domain artifacts while suppressing noise, achieving a better balance between noise suppression and fidelity.

The invention provides a hybrid expert model negative feedback iteration method and device, a storage medium and a program product, and the method comprises the steps: carrying out the cross-modal scene semantic analysis of vehicle end feedback data collected from a vehicle end, roadside sensing data of a roadside device, and macroscopic traffic data of a cloud system, and generating corresponding scene label information; based on the scene label information and the macroscopic traffic data, determining an activation weight of each expert sub-model in the hybrid expert model; negative feedback data in the vehicle running process are collected, and in combination with the influence degree of the target expert sub-model in the negative feedback scene, a responsibility expert sub-model needing to be optimized is determined; and performing directional incremental training on the responsible expert sub-model based on the negative feedback data, performing comprehensive evaluation on the expert sub-model after incremental training, and executing parameter replacement or version updating when an updating condition is met. According to the method, rapid and accurate local optimization is realized while the overall reliability of the model is ensured, and global parameter drift is effectively inhibited.