8results about How to "Eliminate noise interference" patented technology

The application relates to the technical field of artificial intelligence, and discloses a speech synthesis method and device, computer equipment and a storage medium, which comprise the following steps: obtaining reference speech data and speech synthesis text; performing emotion extraction on the reference speech data to obtain reference emotion embedding; fusing the reference speech data, the speech synthesis text and preset noise features to obtain original fusion features; determining a dynamic emotion injection coefficient based on a preset denoising time step parameter; performing denoising processing on the original fusion features based on the dynamic emotion injection coefficient and the reference emotion embedding to obtain target synthesis spectral features; and performing audio conversion based on the target synthesis spectral features to obtain target speech data. The application can be applied to the speech synthesis scene of financial technology and medical health, and improves the emotion expression accuracy of speech synthesis.

This disclosure provides an image detection method, apparatus, electronic device, and storage medium. The image detection method includes: acquiring a multiplex immunofluorescence image to be tested, wherein the multiplex immunofluorescence image is imaging data after multiplex immunofluorescence staining of a tissue sample to be tested; performing tissue segmentation based on the multiplex immunofluorescence image to obtain a tissue region image; and performing image sharpness detection on the tissue region image to obtain a detection result. The image detection method provided by this application can effectively eliminate background noise interference in multiplex immunofluorescence images caused by staining at different locations in the image, greatly reducing the proportion of background in the input image for subsequent image detection, and improving the accuracy and reliability of the image detection results.

This invention discloses an algorithm for removing high-power noise interference based on vortex waves, belonging to the field of signal processing technology. In this invention, the transmitted signal is a vortex wave signal. At the receiving end, an antenna array (number of array elements N≥2) receives the signal containing high-power noise interference. The received mixed signal is then subjected to vortex wave-based noise reduction processing. The vortex wave noise reduction principle is as follows: The average amplitude or the average of the maximum and minimum amplitude values ​​and the average phase of the electric field of N echo signals are taken. Then, the average amplitude and phase values ​​are combined to form a noise reduction electric field. Alternatively, one echo signal from the N echo signals can be arbitrarily selected, and the noise reduction electric field is subtracted from that echo signal. Compared with existing plane wave-based signal processing algorithms, this invention can extract vortex wave signals under low signal-to-noise ratio conditions, which is beneficial for improving the anti-interference capability in the field of signal processing and has significant practical application value.

The application provides an artificial intelligence-based call center data analysis method, and relates to the field of call center data processing.The method collects real-time call audio and pre-processes it, extracts mel-frequency spectrum features and text data with timestamps;uses an analysis model integrating acoustic layer, text layer, cross-modal fusion layer and multi-task analysis head layer to obtain multi-dimensional analysis results such as emotional state, service compliance, service quality score and customer intent.In model training, an adversarial gradient reversal layer and an acoustic environment discriminator are introduced to eliminate environmental noise interference, and a timing risk encoder and a contrast learning are used to enhance the risk evolution capture capability.Based on the analysis results, a comprehensive risk index is calculated, an intervention strategy is matched and real-time pushing is performed, and a comprehensive analysis report is generated after the call is completed.The application realizes cross-environment adaptive purification and multi-modal fusion analysis, effectively improving the service quality monitoring and risk early warning efficiency of the call center.

The application relates to the technical field of new energy automobile battery safety management, in particular to a new energy automobile battery monomer short circuit detection system and method, the system collects target battery signals and battery physical and chemical data through a battery data acquisition module, filters, normalizes and extracts features of the target battery signals through a battery signal processing module, dynamically adjusts a short circuit feature self-adaptive threshold according to real-time working condition quantitative values through a monomer short circuit monitoring module, predicts a short circuit development trend and provides an intervention window period through a short circuit trend prediction module, and monitors an electric-power coupling value and dynamically adjusts and regulates a regulation strategy through a risk window regulation module, so that early and accurate detection and graded early warning of battery monomer short circuits are realized, and the accuracy of short circuit detection and the timeliness of graded regulation are improved.

The application relates to the technical field of semiconductor equipment monitoring, in particular to a guide rail cleanliness monitoring system and method for semiconductor equipment; the system comprises a signal acquisition module, a signal processing module, a pollution identification module and an evaluation decision module; the signal processing module performs signal analysis on the collected electrostatic induction signals and acoustic emission signals, and performs event detection in combination with motion state information to obtain a characteristic event list and a pollution index vector; the pollution identification module performs space-time correlation analysis on the characteristic event list to obtain internal wear sources, and identifies external deposition sources in combination with the pollution index vector; an excitation signal is applied to the external deposition sources, and a silent pollution layer is identified according to a response result; the evaluation decision module generates a comprehensive evaluation result of the guide rail cleanliness state based on the internal wear sources, the external deposition sources and the silent pollution layer; the application realizes online monitoring of guide rail cleanliness and accurate identification of pollution sources, and improves the reliability of semiconductor equipment operation.

The invention provides a method for quickly identifying hot spots of a divertor of a fusion device under a cyclic thermal load. The method comprises the following steps of: recording surface temperature distribution and an average value once by using an infrared thermal imager every N / 5 cycles; calculating absorption heat flux density through coolant inlet and outlet temperature difference, and screening effective heat load data; monitoring inlet flow, temperature and pressure; setting a relative percentage threshold as a standard acceptance factor AC; the average temperature of each tungsten block is firstly compared integrally, then local highest temperature comparison is carried out on abnormal blocks, and finally global analysis is carried out. The relative temperature rise acceptance factor is introduced based on the minimum average temperature, and the judgment statistical significance is improved; dynamically calculating the heat flux density to ensure the loading effectiveness; the method is suitable for divertor delivery testing and fusion reactor in-situ monitoring, rapid and automatic hot spot identification is achieved, and the testing efficiency and accuracy are improved.

This application relates to an intelligent welding tracking method, system, device, and medium for alloy chain plates, belonging to the field of intelligent welding technology. The method includes: acquiring point cloud data and image frame data of the area to be repaired; converting the point cloud data to a base coordinate system to obtain clean point cloud data, and identifying all points below the unworn surface with a difference exceeding a set height threshold to form the point cloud contour of the worn area; performing edge detection on the image frame data to obtain edge pixels containing the wear boundary and crack lines, and back-projecting them to the base coordinate system to generate a projection point set; fusing the projection point set and the point cloud contour according to a fusion distance threshold, and applying B-spline curve fitting to generate the contour baseline of the area to be repaired; selecting suspected crack edge segments from the edge pixels and mapping them back to the contour baseline while recording crack parameters; generating a welding path based on the contour baseline of the area to be repaired and the crack parameters, thereby achieving automatic welding and improving the stability of the alloy chain plate repair quality.