30results about How to "The output is stable" patented technology

The invention provides an electroencephalogram sleep staging method based on a deep convolutional neural network. The electroencephalogram sleep staging method comprises the following steps that S1, sleep signals of a subject are collected, and multi-lead electroencephalogram signals in the sleep signals are extracted; S2, performing data preprocessing on the electroencephalogram signals; S3, constructing and training an end-to-end deep convolutional neural network classifier; and S4, performing electroencephalogram sleep staging by using the deep convolutional neural network classifier. Compared with the conventional CNN electroencephalogram sleep staging method, the electroencephalogram sleep staging method provided by the invention has the advantages that under the condition of the sameiteration times and learning rate, each batch of the model adopts higher data, and the obtained output result is more stable. In terms of accuracy and F score, the method provided by the invention has better classification performance.

An image processing circuit and method thereof for enhancing text displaying of an image are provided. The image processing circuit performs the method. Firstly, at least a first illumination area and at least a second illumination, which is located adjacent to the first illumination area in the image are defined according to luminance values of a plurality of pixels of the image. Next, a luminance regulation value of at least one of the pixels in the first illumination area is calculated, and the luminance value of the corresponding pixel in the first illumination area is adjusted according to the luminance regulation value. During processing of the image, the luminance values of all of the pixels in the second illumination area are restricted from any adjustment.

The invention discloses a method and system for detecting a weak vital sign signal of an airborne radar. The method comprises steps controlling the airborne radar to scan a target region, and obtaining a multichannel echo signal of the target region; preprocessing the echo signal to obtain multi-channel coherently superimposed image domain data; making image domain data be subjected to processingof any one or more of motion compensation, iterative side lobe suppression and strong reflection detection to obtain a corrected target; and detecting and tracking the positions of targets, and outputting the number and real-time positions of the targets. The method is advantaged in that an unmanned aerial vehicle platform and the radar are integrated, a micro human body target in the target areacan be accurately detected and positioned, static false alarms caused by the wall and the ground can be filtered out, the radar detection speed is high, and radar detection efficiency is high.

The invention provides a 4-10 decoder design method based on strand displacement, and relates to the field of DNA strand displacement. According to the method, firstly concentration of DNA signal chains is set according to the input logic value of a DNA 4-10 decoder, and each DNA signal is divided into 10 different DNA signal chains having the same logic value through a fan-out gate module; then hybridization reaction with an AND gate module and an OR gate module is performed based on the Watson-Crick (C-G, T-A) complementary matching principle according to difference of domains attached on the outputted 80 DNS signal chains; and finally 20 DNS signal chains are outputted, and the 20 DNS signal chains represent different logic values according to the size of concentration so that the function of the 4-10 decoder can be realized. The beneficial effects of the 4-10 decoder design method based on strand displacement are that the AND gate module, the OR gate module and the fan-out gate module are combined together based on double tracks firstly so that the DNA 4-10 decoder is realized; and the output result is enabled to be more stable by regulating the concentration coefficient of the threshold gate in the fan-out gate module, the AND gate module and the OR gate module.

The invention discloses a transformer multi-parameter fault monitoring system and a detecting method. Partial discharge signals and vibration signals are respectively collected; the partial dischargesignals are combined with the vibration signals to obtain one combined signal group; amplitude limiting, filtering, and amplification are carried out on the combined signals and then over-voltage andover-current protection are carried out to obtain analog signal output results; and with one group of analog signal output results, two kinds of fault signals can be detected simultaneously. The partial discharge signals are combined with the vibration signals to obtain one combined signal group by using the signal combination technique; the partial discharge fault of the transformer and the internal vibration fault of the winding can be understood simultaneously by observing one group of signals; and the final analog signal output change can be caused no matter which signal group has a problem, so that a transformer fault result is obtained. Therefore, the detection efficiency is improved and the detection accuracy is enhanced stably. The many references are provided for the workers; andthe fault detection and maintenance pressure of the transformer are reduced substantially.

The invention provides a generation method and system of an online shopping user large index model. The generation method and system generate an online shopping consumption index, an online shopping satisfaction index, an online shopping rights protection index, and an online shopping false purchasing index. The online shopping consumption index is calculated through online shopping platform transaction data. The online shopping satisfaction index is calculated through commodity comment text recognition. The online shopping rights protection index is calculated through commodity comment rightsprotection text recognition. The online shopping false purchasing index is calculated through false purchasing text recognition. On the basis of the model calculation results of the steps, an onlineshopping user large index is obtained through model integration calculation, and an online shopping user large index prediction model is obtained through an ARMA model. The calculation data of the online user large index model generated by the method and system can be easily obtained, each output result of the model is unique and stable, and the overall development state of the online consumptionlevel, satisfaction degree, rights protection behavior and false purchasing comments of specific commodities and specific consumer groups in a certain period of time can be dynamically monitored and predicted.

The invention belongs to the technical field of overvoltage protection, and discloses a high-voltage detection circuit for overvoltage protection. The high-voltage detection circuit is characterized by comprising a decoder, an adjustable voltage reduction module and a comparison circuit, wherein the decoder is a 3-8 decoder and comprises three input ports and seven output ports; the adjustable voltage reduction module is used for reducing input high voltage to low voltage and comprises seven sections of voltage reduction circuits, and the seven sections of voltage reduction circuits are respectively matched with the seven output ports; the comparison circuit comprises a positive feedback loop, the positive feedback loop comprises a phase inverter and a lvlshift module, a comparison threshold value of the comparison circuit is set based on the lvlshift module, and comparison between the voltage reduced by the adjustable voltage reduction module and the comparison threshold value is achieved. In conclusion, voltage reduction of different degrees can be effectively formed on the basis of the seven sections of voltage reduction circuits, the area of the voltage detection circuit is reduced, the output result is more stable, and the effect of improving the circuit efficiency is further achieved.

The embodiment of the invention provides a lane line identification method and device, equipment and a storage medium, and the method comprises the steps: obtaining a video, and determining the attribute parameters of the lane line according to the video frame information in the video, and enabling the attribute parameters of the lane line to comprise the type of the lane line, the stability of the lane line and the position of the lane line; determining attribute parameters of historical frame lane lines and attribute parameters of future frame lane lines according to the attribute parametersof the lane lines; and determining the attribute parameter of the current frame lane line according to the attribute parameter of the historical frame lane line and the attribute parameter of the future frame lane line. The invention relates to the technical field of image processing, and the attribute of the lane line of the current frame is reasoned by using the time sequence attribute of the front and rear frames of the video and the output of the segmentation model so that the judgment of the type of the lane line is more accurate, the lane line output result is more stable, and the sudden change of the type is reduced.

The invention discloses a prediction type constant-speed box system and a constant-speed control method for a wind generating set. The system comprises a planetary gear train, a control gear, a power source, an anemometer and a controller, wherein the outer circumference of a gear ring of the planetary gear train is arranged to be a gear face meshing with the control gear, a planet carrier and a sun gear of the planetary gear train are fixedly connected with a wind wheel shaft and a generator main shaft of the wind generating set respectively, the power source is in transmission connection with the control gear, the controller is in control connection with the power source, and the anemometer is arranged in a working environment of the wind generating set and is in signal transmission connection with the controller. According to the method, the controller predicts a rotating speed of the wind wheel shaft according to anemometer data, a rotating speed of the control gear is controlled through the power source, and a rotating speed of the generator main shaft is kept constant to be a set target value. When external conditions change, a more stable output result can be obtained through active control, and wind energy is better utilized for power generation.

The invention discloses a method for detecting phase alternation line (PAL) type video signal subcarrier phase conversion, which comprises the following steps of: acquiring initial identification signals of a PAL row and a national television system committee (NTSC) row; acquiring local identification signals of the PAL row and the NTSC row; and comparing the local identification signals with theinitial identification signals, and outputting identification signals of the PAL row and the NTSC row according to the comparison result. A device for detecting the PAL type video signal subcarrier phase conversion comprises an initial identification module for acquiring the initial identification signals of the PAL row and the NTSC row, a local signal generating circuit, a statistical comparisonmodule and a logic control module. The method and the device for detecting the PAL type video signal subcarrier phase conversion finally output the local generated signals and have stable output results.

The invention relates to an on-orbit detection method for a steel rail in a running state. The detection method comprises the steps of information acquisition, image acquisition, image preprocessing, judgment and classification of a neural network model, regression analysis of a multiple linear regression model, relative error calculation and information transmission. Real-time detection of the damage condition of the steel rail in the running state is achieved, interpretation is conducted through the neural network algorithm with the deep learning capacity, the confidence coefficient and stability of interpretation are improved, regression analysis is conducted on the interpretation result through the multiple linear regression model, the influence of train running parameters and meteorological parameters on the interpretation result is effectively avoided, and the detection result is closer to the real situation; and meanwhile, the invention further provides detection equipment based on the steel rail on-orbit detection method, and compatible sharing, intercommunication and interconnection with a train control system are effectively achieved through equipment carrying, wire harness integration and program implantation.

The invention provides a signal sensor with magnetic field bias. The signal sensor with magnetic field bias includes a current pulse-generating circuit, a signal sensor, a bias magnetic field apparatus and a waveguide wire, wherein the current pulse-generating circuit is connected with the waveguide wire to form a loop; the waveguide wire passes through the center of the signal sensor; and the bias magnetic field apparatus is arranged at the outside of one end of the signal sensor. The signal sensor with magnetic field bias has the following advantages: by increasing one bias magnetic field apparatus, the signal sensor enables the bias magnetic field generated from a natural magnetic field where the waveguide wire exists to realize fixed polarization so that both the magnetic pole and the field intensity of the magnetic field can be fixed. Therefore, the external magnetic field interference or other cannot influence the magnetic field of the waveguide wire; signal output is stabilized; and the output result is stabilized.