159results about How to "Improve the correct recognition rate" patented technology

The present invention discloses a logistics industry-oriented system and method for cascade protection of user privacy information. The method comprises: in a collection stage, using an electronic express form instead of a conventional express form, thereby ensuring that user information is transparent to a courier; in an express delivery transfer stage, applying an idea of secret division, so that each station can acquire only part of station information; in an express delivery notification stage, using a server proxy notification method, so that a courier can issue a notification without knowing contact information of a recipient; in a signing-for-accepting-the-express-delivery stage, verifying the identity of the recipient by using a verification code that is generated by the server in the previous stage; and at the same time, encrypting a content in network communication by combining symmetric encryption and asymmetric encryption.

The invention discloses a vibration event mode identification method. The method includes acquiring and obtaining the original vibration signals, including the vibration signal and the non-vibration signal, via a vibration sensor; separating the vibration signal in the original vibration signals; denoising the vibration signal; performing feature extraction on the denoised vibration signal and obtaining the characteristic vector including three characteristics of: A performing wavelet packet decomposition on the time-frequency domain and obtaining the energy characteristic vector, B performing the cepstrum analysis and extracting the cepstrum parameter characteristics, and C extracting the signal characteristic on the time domain; and establishing an identification model composed of two grades of classifiers, wherein the first grade of classifier is a classifier based on SVM and divides the vibration events into non-intrusive events and intrusion events by taking the characteristic vector extracted from the vibration signal as the input, and the second grade of classifier performs the identification based on the artificial nerve network for the intrusion events to obtain the classification result.

The invention discloses a method for test recognition based on verification and a device thereof, the steps of the method comprise receiving a verification request, splicing a generated identifying code segment and a to-be-recognized image segment comprising test characteristics into an identifying code image to transmit to a terminal, receiving user recognition information which is transmitted by the terminal and is input by users after the users recognize the identifying code image, and extracting text characteristics included by the image segment in the user recognition information to be the recognition result of the image segment. The scheme of the invention increases the accurate recognition rate of the recognition result, and resolves the limitation of the computer recognition.

A method for identifying iris of human eye includes detecting four boundary points of pupil and accurately positioning pupil centre, detecting four outer boundary point of its and confirming iris centre, expanding eye image in 360 degree around iris center to be rectangle for searching pupil boundary, carrying out topological conversion to realize iris image noralization.

The invention discloses a noise suppression method and device and a mobile device. The method comprises the following steps: determining real-time position of the mobile device in a working environment; obtaining a fixed position of a noise source in the working environment; determining a relative position of the noise source with respect to the mobile device according to the real-time position and the fixed position; and controlling the mobile device to carry out suppression on a noise signal of the noise source according to the relative position. Therefore, the mobile device can eliminate the noise signal of the fixed noise source very well.

The invention discloses a convolutional neural network model training method and device, an electronic device and a computer readable storage medium. The method comprises the following steps of dividing the convolutional neural network into a plurality of convolutional stages, wherein the convolution stage is composed of at least one convolution layer; determining the parameters of the convolutional neural network; inputting a positive training sample set into the convolutional neural network for training to obtain the feature images of a plurality of convolutional stages corresponding to each sample image; for each sample image, fusing the feature images of a plurality of corresponding convolution stages; obtaining a positive sample convolutional neural network model according to the fused feature image of each sample image, wherein the positive sample convolutional neural network model is used for identifying a target area. According to the embodiment of the invention, the feature images of the plurality of convolution stages of the convolutional neural network are fused during the training process of the positive sample convolutional neural network model, so that the correct recognition rate of the positive sample convolutional neural network model to the target area can be improved.

The invention discloses a novel recognition method of neural network patterns, which comprises a learning step and a recognition step. In the steps, a training sample set is set up, and training sample in the training sample sets are divided in to a plurality of blocks; a learning ahead masking SLAM model is set up, and RBF neurons in the model are utilized to obtain a training sample subset of apriority ordered neural network based on the connectivity nature of homologous congeners; the priority ordering for the training sample subset is carried out, and learning neurons are arranged according to the learning sequence; a PORBF network based on the connectivity nature of homologous congeners is set up, and a testing sample is input into the PORBF network; the output of each RBF neuron isdetected, and the activated output neuron with the smallest serial number is found; and the type of the neuron is taken as the decision output of the network. The invention has higher correct recognition rate, and is more suitable for solving the recognition problem of large-scale super-multi-class patterns.

The invention discloses a voice-control-based vehicle driving automation system and method. The system includes a voice input button which is used for triggering voice control, a voice input module which is used for collecting voice signals, a semantic recognition module which is used for identifying control command items expected by the voice signals, an identity recognition module which is usedfor identifying whether the voice signals come from a driver, a control module which determines whether to execute the control command items based on a semantic recognition result, an identity recognition result, the state of the voice input button, and the state of a driving automation unit to be controlled and then issues control commands and feedback information, a voice output module which isused for voice broadcast of the feedback information, a display module which is used for displaying the feedback information, and the driving automation unit which executes control commands issued bythe control module. The voice-control-based vehicle driving automation system and method can quickly and conveniently finish setting through voice control, driving burden is low, and driving safety and comfort are improved.

The invention discloses a double-layer license plate character segmentation method based on projection and recognition. The method comprises the following steps: S1, projection in a horizontal direction is carried out on the double-layer license plate, and the double-layer license plate is segmented into an upper part and a lower part; S2, a vertical projection method is used for carrying out character segmentation on the license plate of the lower layer; S3, after segmentation of characters of the lower layer is carried out, if the number of the characters is 5, a step 4 is directly carried out, if the number of the characters is larger than 5, a recognition module is used for acquiring reliability of character recognition, and through the highest reliability, the position information and the length and width information of the best five characters on the license plate of the lower layer are acquired; S4, according to the character segmentation information for the license plate of the lower layer, the width of the best character is obtained; and S5, according to the width of the best character, initial positions of the first character and the second character on the license plate of the upper layer are calculated, and segmentation is carried out.

The invention discloses a lung texture recognition method based on deep neural network extraction appearance and geometric features, which belongs to the field of medical image processing and computervision. 217 lung 3D computed tomography (CT) images are used as source data, a plurality of groups of data sets are obtained through preprocessing, and each group of data comprises a CT image small block, a corresponding geometric information image small block and a category label. A dual-channel residual network framework is built, the CT image small block and the corresponding geometric information small block are used as input of each channel respectively, the appearance information and the geometric information of the lung textures are learnt respectively through the dual-channel residualnetwork, the information is effectively fused, and the recognition rate is thus high. Besides, the provided network has a distinct structure, and building and realization are easy.

The invention discloses an efficient recognition method for the flight control signal of a low-altitude, slow-speed and small unmanned aerial vehicle. The method comprises the steps of step 1, carrying out primary detection on a flight control link signal; step 2, extracting key features, and constructing a target feature database; step 3, carrying out feature fusion and optimization; step 4, analyzing the signal according to the fused optimized features so as to obtain the type, the frame format and the keyword segment of the signal, and storing the obtained analysis result into the target feature database; step 5, according to a learning algorithm based on a neural network, realizing the accurate classification and recognition of the target, and obtaining a correct recognition result. According to the method, the recognition problem for the flight control signals of low-altitude, slow-speed and small unmanned aerial vehicles can be solved. Therefore, the recognition, management and control for a specific airspace can be realized. Meanwhile, the high-efficiency early warning and defense effect for various low-altitude, slow-speed and small unmanned aerial vehicles which have illegally intruded into secret-related and sensitive airspaces can be realized. The method is higher in correct recognition rate, lower in calculated amount, higher in processing efficiency and strong in applicability.

The invention discloses a quick near-infrared brain-computer interface method. The method is mainly characterized in that a multi-channel near-infrared optical brain signal collecting system is used for obtaining multi-channel near-infrared optical brain signals generated by a user in the brain-computer interface task executing process in real time, after the signals are pre-processed, feature space transformation is carried out to extract feature vectors, and mode recognition is carried out to judge the type of a brain-computer interface task executed by the user. The method has the advantages that the recognition rate and recognition speed of brain-computer interface tasks are obviously improved compared with a traditional near-infrared brain-computer interface method which carries out recognition with amplitude as feature vectors.

The invention discloses a particle filter-based improved RAIM (Receiver Autonomous Integrity Monitoring) anti-deception jamming method. The invention relates to an improved RAIM anti-deception jamming method. For solving the problems of measurement failure of a single satellite, positioning misleading of a receiver by a control resolving flow, and neglecting of dependency and similarity among remaining vectors by RAIM, the invention provides a particle filter-based improved RAIM anti-deception jamming method. The particle filter-based improved RAIM anti-deception jamming method comprises the following steps: 1, forming a formula described in the specification; 2, obtaining a Rho n and three-dimensional coordinates of the satellite; 3, calculating a parameter described in the specification; 4, calculating y (m); 5, calculating w (m); 6, selecting wmax; 7, calculating a parameter described in the specification; 8, calculating a maximal visible satellite SLmax; 9, calculating a judgment threshold gamma; 10, judging the existence of a deception satellite; 11, estimating a sequence number of the deception satellite; and 12, if the mark F of the deception satellite is equal to 1, removing jamming, carrying out positioning resolving, and if F is not equal to 1, carrying out positioning resolving. The particle filter-based improved RAIM anti-deception jamming method is applied to the field of improved RAIM anti-deception jamming.

The invention discloses a vehicle license plate recognition method and system. The vehicle license plate recognition method comprises the steps that A, when the situation that a vehicle passes on the site is detected, a video image of a license plate of the vehicle is acquired; B, the acquired video image is processed, whether a license plate number can be automatically recognized through the processed video image or not is judged, if not, the step C is executed, and if yes, the step E is executed; C, the acquired video image is sent to an operating terminal; D, the received video image is displayed by the operating terminal, and a license plate number input by an operator according to the displayed video image is received; E, the license plate number is recorded and stored. The invention further provides the vehicle license plate recognition system. By the implementation of the technical scheme, the vehicle license plate recognition method and system increase the correct recognition rate of the license plate numbers and are more reasonable in practical application.

The invention provides an underwater acoustic target recognition method based on a weighted support vector machine. The method comprises the steps of performing framing preprocessing, performing feature extraction, building a weighted support vector machine recognition model, searching an optimal kernel function parameter and a penalty factor by utilizing a grid search method, training the weighted support vector machine recognition model, reflecting a recognition result of a classifier on an underwater acoustic target through a confusion matrix, and counting the recognition accuracy of the classifier. According to the method, an appropriate feature extraction method is selected for the characteristics of the underwater acoustic target, the method has the capability of autonomously selecting model parameters, the correct recognition rate of the underwater acoustic target is 80% or above, and the stability of a classifier is higher than that of an existing classification method.

The invention discloses a method for identifying the subcarrier modulation constellation diagram of wireless optical communication under atmospheric turbulence. The method comprises the following steps that: an atmospheric turbulence channel mathematical model is established; the constellation diagram of subcarrier modulation signals which have passes through an atmospheric turbulence channel is obtained according to the atmospheric turbulence channel mathematical model, fuzzy C-means clustering is performed, and the hard trend mean feature parameter of cluster centers is extracted; and an improved three-layer BP neural network classifier is designed, the feature parameter is adopted to train the improved BP network, and the trained network is adopted to identify the constellation diagram.The method for identifying the subcarrier modulation constellation diagram of the wireless optical communication under the atmospheric turbulence of the present invention has the advantages of simpletechnical routes, economical performance, high practicability, high feasibility and easiness in implementation. With the method adopted, the correct recognition rate of subcarrier high-order modulation under severe light intensity fluctuation caused by atmospheric turbulence can be improved, and the influence of the atmospheric turbulence on the detection of the constellation recognition can be well suppressed.