5 results about "Statistical noise" patented technology

The application relates to the field of point cloud registration, and provides a point cloud registration method and system based on a multi-feature point point set. The method comprises the following steps: preprocessing point cloud by statistical noise reduction based on Gaussian distribution and voxel downsampling; extracting multi-feature points by adopting a directional bounding box algorithm, an improved internal morphological descriptor algorithm and a geometric centroid point algorithm; constructing a multi-feature point point set; and obtaining an initial transformation matrix by adopting a principal component analysis-based coarse registration algorithm; performing fine registration by adopting a G-ICP algorithm based on L2LOSS robust loss function optimization; and outputting a registration result. The application improves the accuracy and efficiency of point cloud matching and reduces the registration time.

The application discloses a kind of based on semantic drive's audio processing method and device;The method comprises: spatial decomposition is carried out to multi-channel audio signal, and forward target signal and backward environmental signal are extracted;Respectively carry out statistical noise reduction, obtain forward noise reduction gain and backward noise reduction gain;Respectively input acoustic semantic perception network, obtain forward and backward semantic probability vector;Based on semantic probability vector and semantic complementary weight matrix, the dynamic transmission weight of backward environmental signal is calculated;Weighted fusion is carried out using forward noise reduction gain, backward noise reduction gain and dynamic transmission weight, and enhanced audio signal is output;The present application transmits side rear important acoustic event on demand through semantic gating mechanism, retains environmental situation awareness capability while suppressing environmental noise, adopts statistical noise reduction and lightweight network heterogeneous deployment, meets the requirement of low delay and high fidelity of edge device, significantly improves the user experience of auditory enhancement device.

This invention provides a federated incremental modeling method based on dynamic probabilistic optimization in heterogeneous edge environments. The method includes the following steps: The class density vectors of devices in each factory, after statistical noise addition, are uploaded to a server; the server selects multiple candidate sets, calculates the distribution offset coefficient of each candidate device set, and selects the set with the smallest offset to participate in federated training; the selected device is trained and the optimal parameters are uploaded; the server aggregates and distributes global model parameters; the device calculates the newly added hidden layer output based on the global parameters and uploads it to the server; the server performs multi-index evaluation based on the device's performance, optimizes its selection probability using the Exp3 algorithm based on multi-index evaluation and fairness constraints, and selects the device to participate in training based on this probability; training ends when the number of hidden layer nodes exceeds the limit or the residual meets the expected tolerance, and the model is distributed to the device as a soft measurement model.

The method and device for repairing Internet of Things signals complete the whole process of loss statistics, noise modeling, weight generation and sample separation locally at each Internet of Things node, do not need to upload the original data of the nodes, fully fit the core privacy protection characteristics of federated learning, realize the fine quantification of noise through data-driven loss distribution statistics, break through the ideal data hypothesis limitations of traditional schemes on sample balance, data isomorphism and modal integrity, can strongly adapt to complex non-ideal working conditions in actual deployment of the Internet of Things, and the finally output weight parameters and sample separation results can suppress the interference of noise on federated collaborative training from the data source level, provide high-reliability sample support and weight basis for subsequent robust federated model training, and effectively solve the pain points that the existing technology cannot adapt to heterogeneous scenes and has great difficulty in cross-domain adaptation.

The application provides a trapped personnel position state analysis method and system, and relates to the field of data processing. In the method, a background entropy flow base state is formed by constructing the building structure distribution, conductive fluid flow channel and metal component distribution in the disaster target area, and a background entropy flow field is generated by statistically analyzing the noise strength probability distribution, channel joint fluctuation relationship and time stability index. In the search and rescue stage, the background entropy flow field is converted to a controlled mode by active disturbance, and the multi-physical channel noise response before and after the disturbance is collected by relying on a multi-field passive sensing network to construct a multi-dimensional local entropy response vector. The initial suspected area is identified and the entropy anomaly core area is extracted by combining the consistency, directionality and duration of multiple noise channels. Finally, the metabolic activity and posture restriction condition are inferred according to the local entropy response coordination relationship under different disturbance modes, and the trapped personnel position state result is output. The technical solution provided by the application is convenient for accurately identifying the position state of trapped personnel.