13results about How to "Filter noise" patented technology

This application discloses a circuit board integrated within the rear cover of a motor, applicable to the field of electronic circuits. This circuit board integrates an electromagnetic compatibility (EMC) suppression circuit and a Hall sensor signal feedback circuit with memory function. The EMC suppression circuit, used to suppress electromagnetic interference from the motor, includes an EMC filter, a transient voltage suppression diode, a first filter capacitor, a second filter capacitor, and the motor itself. The Hall sensor signal feedback circuit includes a Hall sensor, a third filter capacitor, and a pull-up resistor. The Hall sensor is a bipolar latching type, used to output speed and position feedback signals based on the magnetic field of a bipolar magnetic ring mounted on the motor rotor. Thus, through the coordinated operation of the EMC filter, the transient voltage suppression diode, and multiple filter capacitors, the EMC of the motor is improved.

This invention discloses a semantically guided method and system for separating auditory evoked brainwave latent space, belonging to the field of brain-computer interface technology. The method includes: acquiring multi-channel EEG signals evoked by mixed audio from multiple sound sources and preprocessing them to obtain mixed spatiotemporal features; compressing the features into a low-dimensional latent space to generate an EEG latent code sequence; fusing semantic text prompts and time-span prompts to generate a semantic conditional embedding sequence; using conditional embedding as a guide, generating a separation latent code specific to the target sound source through diffusion in the latent space; and decoding the latent code to output the corresponding auditory evoked EEG components. The system adopts a five-layer architecture consisting of a hardware layer, a preprocessing layer, an interaction layer, a core model layer, and an output layer, and is equipped with a multimodal joint separation model to achieve end-to-end processing. This invention requires no original audio input and can complete source-level neural signal separation based on natural language, exhibiting high computational efficiency and excellent separation accuracy, making it suitable for auditory brain-computer interface and cognitive neuroscience research.

The present application relates to the technical field of medical film, and particularly relates to a medical film quality detection system based on multi-parameter fusion, which comprises an acquisition module, a discrete determination module, an identification module, a risk determination module, an adjustment module and an early warning module. The present application takes a dynamic discrete coefficient as a trigger condition, analyzes the contribution of each process parameter to the current thickness fluctuation to identify the primary root cause, forms a quantitative process risk index by comprehensively considering the out-of-control degree of the root cause parameter and the thickness deviation, and according to the process risk index, an adjustment instruction with clear direction and strength suggestion can be sent, and the adjustment efficiency is evaluated after adjustment, early warning is given when the control loop fails or the problem escalates, and the problems of lagging quality defect discovery and blind and lagging process adjustment caused by the failure to quickly and accurately locate the process abnormal root cause in the complex multi-process coupling production process are effectively solved.

An integrated circuit device includes: a substrate; and a metal oxide semiconductor capacitor (MOSCAP) on the substrate. The MOSCAP includes: a lower semiconductor device on a substrate, the lower semiconductor device including a pair of lower source / drain regions and a lower gate structure between the pair of lower source / drain regions; and an upper semiconductor device on the lower semiconductor device, the upper semiconductor device including a pair of upper source / drain regions and an upper gate structure between the pair of upper source / drain regions. A lower gate structure is electrically connected to both of the pair of upper source / drain regions.

This invention discloses a robot grasping pose estimation method and system based on multi-frame fusion and depth calibration. The method includes: acquiring multiple frames of RGB-D data using a depth camera, extracting a target mask using a lightweight instance segmentation network; using a random sample consensus algorithm to detect ground planes based on the depth map and calculate the relative height of the camera; extracting point clouds based on the mask and depth data, and calculating the initial 6D pose using principal component analysis; performing nonlinear calibration of the pose based on radial basis functions, combined with camera height and observation distance; and robustly fusing the multi-frame poses using a Z-score-based outlier removal algorithm to output the optimal target pose and convert it into a robot arm base coordinate system command.

The invention provides a multi-frequency-hopping signal sorting method, system and device used in a complex electromagnetic environment and a storage medium, and aims to realize multi-frequency-hopping signal sorting in the complex electromagnetic environment without any prior information. The method comprises the steps that multi-frequency-hopping signal data in a complex electromagnetic environment are acquired, and the multi-frequency-hopping signal data comprise effective signal data and noise signal data; performing time-frequency analysis processing on the multi-frequency-hopping signal data to obtain time-frequency data corresponding to the multi-frequency-hopping signal data; predicting an optimal radius parameter in a DBSCAN clustering algorithm by using a neural network learning method; based on the optimal radius parameter, performing DBSCAN clustering algorithm processing on the time-frequency data to obtain a cluster of effective signals; and extracting characteristic parameters of the effective signals based on the clusters of the effective signals.

The utility model discloses a reinforcing steel bar detector with a probe for identifying distribution of a reinforcing steel bar mesh, which belongs to the technical field of reinforcing steel bar detection and comprises a sensor module and a host controller connected with the sensor module through a signal line. The sensor module comprises an electromagnetic induction module, an ultra-wideband radar module, an ultrasonic transducer, a frequency-mixing medium-frequency amplifying circuit, an orthogonal frequency-mixing filter circuit, an analog signal conditioning circuit, an A / D converter, a measurement module and a laser module; the electromagnetic induction module comprises a signal generation circuit, an induction coil and a signal acquisition circuit; the host controller comprises a microcontroller module, a data transmission module, an interface module, a clock module, a memory module, a USB port module and a power supply module. According to the utility model, the detection of the steel bar diameter and the steel bar burial depth in the building component is effectively realized, and the requirements in actual engineering detection are met.

The invention provides a finished bridge linear smoothness judgment method and system based on a main beam splicing process, and relates to the technical field of linear smoothness judgment, and the method comprises the following steps: obtaining a design curve of a main beam finished bridge target elevation, an elevation curve of an actual main beam splicing section and a to-be-judged splicing section length; and if the actual elevation curve exceeds the preset threshold, determining the elevation curve of the first bridge section according to the design curve. And fitting the actual elevation curve with the elevation curve of the first bridge section by adopting a preset spline curve fitting method to generate an elevation curve of the second bridge section. And combining the elevation curves of the actual splicing section, the first bridge section and the second bridge section into a virtual bridge forming line shape, performing high-pass filtering on the line shape, extracting an irregularity maximum absolute amplitude, comparing the irregularity maximum absolute amplitude with a preset irregularity allowable limit value, and judging whether the smoothness of the splicing section meets the requirement or not. According to the method, the smoothness judgment precision of the finished bridge line shape can be effectively improved.

This invention relates to the technical field of behavioral analysis of children with autism, and provides a method and system for collecting daily behavioral data of children with autism. The method includes: real-time acquisition of triaxial acceleration and triaxial angular velocity from an inertial measurement unit as raw data; extraction of feature data from the raw data; judgment of the physical consistency of the raw data based on the feature data, distinguishing between real limb movement signals and sensor relative motion signals; decomposition of the raw data into a first information stream and a second information stream; identification of behavioral patterns of limb movements in children with autism based on the first information stream; extraction of features from the second information stream to obtain second information features; the second information features characterizing the intensity of sensor relative motion corresponding to the sensor relative motion signal; calculation of a behavior intensity index based on the second information features; and output of the behavior pattern and behavior intensity index. This invention improves the accuracy of collecting daily behavioral data of children with autism.

The embodiment of the invention discloses an image background similarity determination method and device, equipment and a storage medium, and relates to the technical field of image processing, and the method comprises the steps: segmenting video frames through employing a preset panoramic segmentation network, and obtaining a semantic region node set corresponding to each video frame in a one-to-one manner; constructing an initial space-time diagram according to all semantic region node sets, and updating the initial space-time diagram by using a preset graph neural network to obtain a target space-time diagram; determining a global background descriptor according to a target background region node corresponding to the video frame; according to the global background descriptor corresponding to the first video frame and the global background descriptor corresponding to the second video frame in the to-be-processed video frame sequence, the image background similarity between the first video frame and the second video frame is accurately determined, so that the background matching accuracy can be remarkably improved; and the reliability of medical environment safety assessment or financial safety event analysis can be enhanced.

The invention relates to the technical field of video processing, and discloses an edge computing and AI fused real-time video stream processing optimization method. The edge calculation and AI fused real-time video stream processing optimization method comprises the following steps: preprocessing a real-time video; performing video optimization processing; obtaining an optimization result; according to the invention, the internal noise of the video is filtered by using a block and low-rank tensor recovery video denoising algorithm, so that the video definition is improved, and the video edge information is more significant; the edge of a denoised video is detected through a PSVM, the video edge detection effect is improved, the noise of an original video image can be effectively removed through an algorithm, the definition of the video image is improved, edge information is made to be more remarkable, when the noise combination level is higher, the entropy value, the AG value and the PSNR value of the two video sequences which are expanded and denoised through the method are all in a descending trend, and the noise combination level is higher. Therefore, the method provided by the invention has a better denoising effect when the video sequence with a simple and complex structure is processed.

The application relates to the technical field of voltage quality monitoring, in particular to a power distribution line voltage quality real-time monitoring method based on a low-voltage transformer area topology, which comprises the following steps: based on the historical output power of each distributed photovoltaic in the power grid topology structure of the low-voltage transformer area, a probability density distribution model is established, the distributed photovoltaic is classified, and the weight coefficients of each distributed photovoltaic are quantified, and the typical probability distribution model corresponding to each category is extracted; the fluctuation range of the output power is divided into multiple focal element intervals, and a basic trust degree is allocated to the focal element intervals to establish a focal element model of photovoltaic output; a probability interval power flow model is constructed; the probability interval power flow model is solved to obtain an interval solution of the node voltage amplitude, an accumulated trust function curve is constructed through evidence theory, and the quality confidence of the node voltage is obtained, so that the voltage quality of the power distribution line is evaluated. The application effectively improves the identification accuracy and early warning response capability of the voltage out-of-limit risk of the power distribution line.

The invention discloses a large language model integrity verification method and system based on output feature distribution statistical test, and belongs to the field of artificial intelligence safety. In order to solve the problem of unreliable verification caused by output randomness of a large language model, the method comprises the following steps of: obtaining model finger with expected output and actual output of a to-be-verified model; performing multi-level feature extraction on expected and actual outputs to obtain two groups of feature vector sets; and performing double-sample statistical test on the two groups of feature vector sets to judge whether the feature distribution has significant difference or not. Preferably, the check is a maximum mean difference (MMD) check in conjunction with a permutation check to obtain reliable p values in small samples. And finally, judging the integrity of the model according to a comparison result of the p value and a preset significance level. According to the method, random interference can be overcome, model tampering can be detected with high sensitivity by using a small amount of API calling, and the robustness and reliability of verification are remarkably improved.