6108results about How to "Improve recognition efficiency" patented technology

An audio fingerprint is extracted from an audio sample, where the fingerprint contains information that is characteristic of the content in the sample. The fingerprint may be generated by computing an energy spectrum for the audio sample, resampling the energy spectrum logarithmically in the time dimension, transforming the resampled energy spectrum to produce a series of feature vectors, and computing the fingerprint using differential coding of the feature vectors. The generated fingerprint can be compared to a set of reference fingerprints in a database to identify the original audio content.

A method and system for endpoint automatic detection of audio record is provided. The method comprises the following steps: acquiring a audio record text and affirming the text endpoint acoustic model for the audio record text; starting acquiring the audio record data of each frame in turn from the audio record start frame in the audio record data; affirming the characteristics acoustic model of the decoding optimal path for the acquired current frame of the audio record data; comparing the characteristics acoustic model of the decoding optimal path acquired from the current frame of the audio record data with the endpoint acoustic model to determine if they are the same; if yes, updating a mute duration threshold with a second time threshold, wherein the second time threshold is less than a first time threshold. This method can improve the recognizing efficiency of the audio record endpoint.

An audio fingerprint is extracted from an audio sample, where the fingerprint contains information that is characteristic of the content in the sample. The fingerprint may be generated by computing an energy spectrum for the audio sample, resampling the energy spectrum logarithmically in the time dimension, transforming the resampled energy spectrum to produce a series of feature vectors, and computing the fingerprint using differential coding of the feature vectors. The generated fingerprint can be compared to a set of reference fingerprints in a database to identify the original audio content.

The invention relates to the field of image recognition, and especially relates to a complex character recognition method based on deep learning. Through the analysis of character complexity, a training sample, which contains a to-be-recognized image noise model and a distortion characteristic model, generated by a random sample generator is employed for the training of a deep neural network. The training sample comprises complex noise and distortion, and can meet the demands of the recognition of various types of complex characters. A few of manually annotated first training sample sets and a large amount of randomly generated second training sample sets are mixed and then inputted to the deep neural network, thereby solving a problem that a large number of manually annotated training samples are needed for character recognition through the deep neural network. Moreover, the most advanced deep neural network is employed for automatic learning under the condition that the noise and distortion of a to-be-recognized image are retained, thereby avoiding information loss caused by noise reduction in a conventional OCR method, and improving the recognition accuracy.

The invention discloses a method and device for programidentification based on machine learning. The method comprises: analyzing an inputted unknown program, and extracting a feature of the unknown program; coarsely classifying the unknown program according to the extracted feature; judging by inputting the unknown program into a corresponding decision-making machine generated by training according to a result of the coarse classification; and outputting an identification result of the unknown program, wherein the identification result is a malicious program or a non-malicious program. The embodiments of the invention adopt the machine learning technology, achieve the decision-making machine for identifying a malicious program by analyzing a large number of program samples, and can save a lot of manpower and improve the identification efficiency for a malicious program by using the decision-making machine; and furthermore, can find an inherent law of programs based on data mining for massive programs, prevent a malicious program that has not happened and make it difficult for a malicious program to avoid killing.

An audio fingerprint is extracted from an audio sample, where the fingerprint contains information that is characteristic of the content in the sample. The fingerprint may be generated by computing an energy spectrum for the audio sample, resampling the energy spectrum, transforming the resampled energy spectrum to produce a series of feature vectors, and computing the fingerprint using differential coding of the feature vectors. The generated fingerprint can be compared to a set of reference fingerprints in a database to identify the original audio content.

The invention discloses a video-based detection and recognition system and a method of vehicles. The system comprises a target characteristic database module, a moving target detection module and a vehicle recognition module. A vehicle sample target characteristic database is built to store data by virtue of the target characteristic database module. After a moving target zone is detected by the moving target detection module, the vehicle recognition module builds a search window and recognizes vehicles. The system and the method resolve the technical problems of poor vehicle distinguishing capability and low distinguishing accuracy of the video-based vehicle detection under the condition of obvious background movement, vehicle adhesion, vehicle occlusion and the like in the prior art. The video-based detection and recognition system and the method of the vehicles have the advantages of excellent vehicle distinguishing capability and high accuracy.

The embodiment of the invention discloses a face recognition method and a device. The method comprises the steps of extracting the Haar feature of a current to-be-recognized face image, and detecting the human face area of the to-be-recognized face image by adopting an ADaBoost classifier so as to obtain a face region image; performing the multi-scale feature extraction on the face region image by utilizing a convolution neural network model, and obtaining a feature vector of the face region image; inputting the feature vector, a pre-built legal face database and a preset user similarity threshold value into a multi-task learning model pre-constructed according to a Softmax loss function and a Triplet loss function, and judging whether the to-be-recognized face image is a legal user or not according to the output value of the multi-task learning model. The extracted feature is good in robustness and good in generalization ability. Therefore, not only the face recognition rate improved, but also the accuracy of face recognition is improved. The safety of identity authentication is improved.

The invention provides a method and a device for face recognition. The method includes an M frame image acquired through an image acquisition device is obtained; when information of N faces of N people is contained in the M frame image, information of each face in the N faces is subjected to face recognition to obtain N first recognition records; when information of N faces at second time of the N people is contained in a number i+j frame image of the M frame, information of each face in the N faces is subjected to face recognition to obtain N second recognition records; and recognition results of the N faces of the N people are determined at least based on the N first recognition records and the N second recognition records. According to the method, the technical problem that a plurality of faces are inaccurately recognized or detection omissions occur in recognition processes in the prior art for face recognition cannot be accurately performed in prior face recognition in the condition of face increases or decreases in videos is solved.

The invention relates to an obstacle identification method for smart vehicles. The obstacle identification method includes the following steps that: 1) data points of original scanning point cloud three-dimensional laser radar of surrounding environment of the vehicles under a spherical coordinate system are obtained, and obstacle points are screened out from all the data points; 2) the obstacle points are grouped according to the horizontal azimuth angles of the obstacle points and the radial distances of the obstacle points relative to a three-dimensional laser radar sensor; and 3) each group of obstacle points correspond to one obstacle, and the categories of the obstacles can be obtained according to the relative position relationships of the obstacle points in each group. Compared with the prior art, and according to the obstacle identification method for the smart vehicles of the invention, the intrinsic unity of the measurement principles of the a three-dimensional laser radar and a point cloud data spherical coordinate representation method is utilized; it is point cloud data that are analyzed based on spherical coordinates, and the Cartesian coordinates of the point cloud data are not analyzed, and therefore, high efficiency can be realized; and at the same time, the original data of point cloud are directly analyzed, and grid division is not needed to perform on the point cloud, and therefore, processing efficiency can be improved.

The invention provides a display screen. The display screen comprises a display layer and a blocking layer, wherein the display layer is provided with a displaying face facing a user, the blocking layer is arranged on the displaying face in a stacked mode, the blocking layer comprises a fingerprint recognition region, the fingerprint recognition region comprises at least one first through hole, and the first through holes are used for transmitting induction signals transmitted and received by a fingerprint module below a display screen. The invention provides a display device, the display device comprises a display screen and an optical fingerprint module, and the optical fingerprint module is arranged at the side, away from the blocking layer, of the display layer and is located at the position corresponding to the optical fingerprint recognition region. The fingerprint module comprises an optical transmitting device and an optical inductor, and light signals transmitted by the optical transmitting device are conveyed to the fingerprint lines through the first through holes and are received by the optical inductor through the first through holes after being reflected by the fingerprint lines. The invention further provides a mobile terminal, and the recognition efficiency of the optical fingerprint module is improved.

The invention discloses a method and a system for recognizing images, which relate to a method and a system for recognizing static target images by adopting the image recognition technology. The invention solves the problem that the recognition speed is relatively slower in the prior image recognition technology. The method and the system for recognizing the images are characterized in that image information in the identification area and characteristic information in the identification area are stored as template information, then the area to be recognized in the images to be recognized is confirmed by adopting the identification area, then the characteristic information in the identification area is compared with that in the area to be recognized, when the similarity of both characteristic information in the identification area and that in the area to be recognized is in the error range, users consider that the images to be recognized and the target images are mutually matched, so as to realize image recognition, in the recognition process, the characteristic information in the area to be recognized is just needed to compare, the data quantity is less, and the logical reasoning and the mathematical operation of the characteristic information are not required, thereby quickening the speed of image recognition. The invention is mainly used for searching matched images, for example, bill recognition, seal recognition and the like.

The embodiment of the invention discloses a laser-point-cloud-based urban road identification method and apparatus. The method comprises: a corresponding road edge model is constructed according to laser point clouds collected by a laser sensor; a height of a mobile carrier with the laser sensor is determined and a corresponding road surface model is constructed based on the height and the laser point; and according to the road edge model and the road surface model, a road surface point cloud and a road edge point cloud in the laser point cloud are eliminated, the rest of laser point clouds are segmented by using a point cloud segmentation algorithm, and an object corresponding to the a segmentation result is identified. The height of the mobile carrier is estimated based on the laser point cloud and the road surface model corresponding to the laser point cloud is constructed by using the height, so that the construction efficiency and accuracy of the road surface model are improved. Therefore, the identification efficiency and accuracy of the corresponding object are improved.

A backlight module, a method for identifying a fingerprint under the screen, a device and an electronic device are provided. The fingerprint recognition device is applied to an electronic device having a backlight module, and the fingerprint recognition device includes: a fingerprint recognition module. The fingerprint recognition module is disposed under the backlight module, and the fingerprintrecognition module is configured to receive the infrared light signal emitted by the infrared light source and illuminate the human body finger and pass through the backlight module. The infrared light signal is used to detect the fingerprint information of the finger. The haze of the infrared light signal when passing through the backlight module is smaller than the haze when the visible light for displaying an image passes through the backlight module. An screen fingerprint recognition device of the embodiment of the present application can effectively implement the screen fingerprint recognition of the passive light-emitting display.

The invention provides a multi-task deep learning network-based training method, a multi-task deep learning network-based training system, a multi-task deep learning network-based identification method and a multi-task deep learning network-based identification system. The training method includes the following steps that: the face region of a face image in a training set is obtained; key point detection is performed on the face region, so that key feature point positions are obtained; affine transformation is performed on the face image according to the key feature positions, so that an aligned face image can be obtained; and the aligned face image is inputted into a multi-task deep learning network, so that training can be carried out, and therefore, a multi-task deep learning network model can be obtained. The identification method includes the following steps that: affine transformation is performed on a face image to be identified according to the key feature positions of the face image to be identified, so that an aligned face image can be obtained; the aligned face image is inputted into a trained multi-task deep learning network model, so that feature extraction can be carried out, and feature information can be obtained; and the feature information of the face image to be identified is matched with feature information corresponding to each face image in a registration set, so that identification results can be obtained. With the methods and systems adopted, the training and identification efficiency of the multi-task deep learning network can be improved.

Provided is a neural network optimization method based on a floating number operation inline function library, wherein the model of neural unit is Y=1 / (1+exp(-Sigma wi*xi)). The value range of i is 1 to n, and n is the number of neural units. The floating-point operations inline function is structured in a dikaryon chip, namely, function library IQ math Library. In the neural network optimization method based on a floating number operation inline function library, except that the step _IQ(x[i]) is carried out in circulation, the rest steps are all carried out outside the circulation body, and the efficiency of the execution of all the steps as a whole is greatly improved compared with a floating point arithmetic. The neural network optimization method based on a floating number operation inline function library optimizes the transplant of back propagation (BP) neural network on transcranial magnetic stimulation (TMS) 3206464T, and the accuracy of a result is decided through a beginning decimal calibration. On the premise of guaranteeing the accuracy of the result does not influence the recognition rate, BP network awareness efficiency is greatly improved.

The invention relates to a recognition method of digital music emotion, belonging to the field of computer pattern recognition; the recognition method solves the problem that the existing recognition method of digital music emotion can not recognize sampling-based digital music format, the sorting technology based on a multi-class support vector machine is adopted, acoustic characteristic parameters and music theory characteristic parameters are combined, so as to carry out emotion recognition of digital music; the recognition method comprises the following steps: (1) pretreatment; (2) characteristic extraction; (3) training the multi-class support vector machine; (4) recognition. The music emotion is classified into happiness, impassion, sadness and relaxation, the emotion recognition is carried out based on a sampling-based digital music format file, the common acoustic characteristics in the speech recognition field are not only extracted, and a series of music theory characteristics are extracted according to the theory characteristics of music; meanwhile, the sorting method based on the support vector machine is adopted, the leaning speed is rapid, the sorting precision ratio is high and the recognition efficiency is improved.

The invention provides a subway construction drawing and engineering parameter automatic identification method. A user interface module, a basic primitive identification rule module, a building element identification rule module, an engineering parameter extraction rule module, a basic primitive identification module, a building element identification module and an engineering parameter extraction module are adopted; and the subway construction drawing and engineering parameter automatic identification method is specially used for implementing a technical scheme for identifying engineering parameters related to the risk management in an engineering drawing on the aspect of subway engineering construction and the data basis is provided for a further risk management technical scheme.

The invention discloses a method and a device for realizing malicious domain name identification. The method comprises the following steps: extracting a dynamic characteristic set of a domain name system (DNS) domain name, and making malicious domain name high credibility judgment on the dynamic characteristics of the dynamic characteristic set through a malicious domain name credibility judgment model of the dynamic characteristics; and determining whether the DNS domain name is a malicious domain name according to the malicious domain name high credibility judgment result of the dynamic characteristic set, and storing the result about whether the DNS domain name is a malicious domain name in a corresponding black or white list, wherein the dynamic characteristic set at least includes IP-related characteristics and / or authoritative DNS server main domain name consistence rate. By adopting the technical scheme, a malicious domain name can be determined according to the dynamic characteristic set, and the efficiency of malicious domain name identification is improved through static characteristic high credibility judgment and dynamic characteristic high credibility judgment.

The invention provides an identification verifying system for a living human body for electronic payment, so as to overcome defects existing in the prior art. Through various voice-shape-image biometric feature recognition of face recognition, voice recognition, voiceprint recognition and lip language recognition, whether a user is an operator or not and whether the user is an intelligent living body or not are verified, through a client of the electronic payment and the interactive communication of a server, the feature information of a face, a voice, a voiceprint and a lip language is extracted out from a video and an audio obtained by the client of the electronic payment through the client of the electronic payment, and less data flow is used and transmitted into a server terminal through the Internet so as to perform the feature information of the face, the voice, the voiceprint and the lip language. Direct contact between the identification verifying system and the operator is not needed, whether the user is the operator or not and whether the user is the intelligent living body or not are verified, the identification verifying system is convenient to use, easy to accept by the operator and good in customer experience, and besides, an additional hardware is not needed.

The invention relates to a method for identifying multi-class facial expressions at a high precision based on Haar-like features, which belongs to the technical field of computer science and graphic image process. Firstly, the high-accuracy face detection is achieved by using the Haar-like features and a series-wound face detection classifier; further, the feature selection is carried out on the high-dimension Haar-like feature by using the Ada Boost. MH algorithm; and finally, the expression classifier training is carried out by using the Random Forest algorithm to complete the expression identification. Compared with the prior art, the method can reduce training and identifying time while increasing the multi-class expression identification rate, and can implement the parallelization conveniently to increase the identification rate and meet the requirement of real-time processing and mobile computing. The method can identify the static image and the dynamic image at a high precision, is not only applicable to the desktop computer but also to the mobile computing platforms, such as cellphone, tablet personal computer and the like.

The invention discloses an abnormal behavior detection method, which comprises the following steps of: extracting dynamic human skeleton joints in a video by using a neural network human skeleton extraction model to form a skeleton data set; obtaining a higher-level behavior feature map corresponding to the bone, namely surface behavior features, by an ST-GCN network; and inputting the behavior feature map into the abnormal behavior classifier model, and performing matching to identify the behavior type. The invention also discloses an abnormal behavior detection system which comprises a videomonitoring module and a network model integration module. According to the method, various human body behaviors and a large amount of human body skeleton data can be accurately and efficiently processed, and abnormal behaviors appearing in video monitoring can be automatically identified.

The invention belongs to the field of computer vision and machine learning, and particularly relates to a cross-visual angle gait recognition method based on a multitask generation confrontation network. The objective of the invention is to solve a problem of reduced generalization performance of a model under big visual angle change of the gait recognition. The method comprises steps of firstly carrying out pretreatment on each frame of image for an original pedestrian video frame sequence, and extracting gait template features; carrying out gait hidden expression through neural network coding and carrying out angle transformation in a hidden space; generating a confrontation network through multiple tasks and constructing gait template features of other visual angles; and finally using the gait hidden expression to carry out recognition. Compared with methods based on classification or reconstruction, the method has quite strong interpretability and recognition performance can be improved.

The invention provides a 3D human face quick identity authentication method and apparatus. The method comprises the following steps of S1, obtaining a depth image and a two-dimensional image comprising a current user face; S2, obtaining the depth image and the two-dimensional image of the current user face; S3, performing identity authentication primary screening on the current user face to obtaincandidate reference data sets; S4, extracting feature information of the current user face, and determining a pose orientation vector of the current user face; and S5, performing identity authentication. The functions of input, detection, identification and the like of a face identity are realized by utilizing the depth image and the two-dimensional image, and the primary screening of the reference data sets is performed in combination with the depth image and the two-dimensional image, so that the identification efficiency is improved, the influence of an external complex environment is avoided, and absolute front face of a user is not required.

The invention provides a video classification method, a model training method and device thereof, and electronic equipment. The training method comprises the following steps: extracting initial features of a plurality of video frames through a convolutional neural network; extracting final features of the plurality of video frames from the initial features through a recurrent neural network; inputting the final feature into an output network, and outputting a prediction result of the multi-frame video frame; determining a loss value of the prediction result through a preset loss prediction function; and training the initial model according to the loss value until parameters in the initial model converge to obtain a video classification model. According to the method, the convolutional neural network and the recurrent neural network are combined, so that the operand can be greatly reduced, and the model training and recognition efficiency is improved; and meanwhile, the association information between the video frames can be considered in the feature extraction process, so that the extracted features can accurately represent the video types, and the accuracy of video classificationis improved.

The invention discloses an anti-collision method based on joint verification of binocular vision and laser radar in congested traffic. The method comprises steps as follows: a binocular vision system and a laser radar system are subjected to parameter joint calibration to obtain the corresponding and conversion relation among a camera coordinate system, a radar coordinate system and a vehicle coordination system; a left camera and a right camera collect information of the environment in front of a vehicle, and meanwhile, laser radar is used for performing multi-line scanning on a front area to obtain heterogeneous and asynchronous data of two different types of sensors for pre-processing; whether a barrier exists before the current vehicle is judged, and if the answer is positive, a joint robust verification method is adopted to obtain distance information of the current barrier relative to the current vehicle, and early warning is performed according to the distance information of the barrier. The barrier recognition efficiency and the robustness are greatly improved. The problem that the outline of the barrier obtained by the cameras is incomplete in the congested traffic environment is solved; meanwhile, more accurate and more reliable barrier parameter information can be obtained.

The present invention relates to an electric power device identification model construction method and system, and an identification method of an electric power device. The electric power device identification model construction method comprises: marking electric power device targets in an infrared image respectively corresponding to each type of an electric power device, and obtaining a sample training set; inputting the sample training set into a RPN convolutional neural network to allow a loss function to have a minimum value, and outputting a target candidate frame; inputting the target candidate frame into a Fast-RCNN convolutional neural network, calculating the conversion weight value of the target candidate frame to a corresponding type according to a fully connected layer and a regression function, and employing the frame regression to obtain the Fast-RCNN parameters of the position of the target candidate frame being offset to a corresponding tag position; and setting a sharing convolutional layer learning rage as 0, and performing initialization of the RPN convolutional neural network and the Fast-RCNN convolutional neural network, performing training of an input infrared image in the RPN convolutional neural network according to the Fast-RCNN parameters, and obtaining a RPN convolutional neural network model; and inputting the target candidate frame into the RPN convolutional neural network model, updating the Fast-RCNN convolutional neural network to form a uniform Fast-RCNN network, and outputting an electric power identification model.

The invention discloses an indoor inspection robot system for a substation and an inspection method for the indoor inspection robot system. The indoor inspection robot system comprises a remote monitoring center and multiple robot terminals communicating with the remote monitoring center. Each robot terminal comprises a control module for controlling a robot to move in a three-dimensional space. The control modules drive movement modules so as to drive the robot to move in the X-axis direction, the Y-axis direction and the Z-axis direction and walk to the target detection position. Detection modules monitor the environment of the position and transmit monitoring data to the remote monitoring center. In the running and monitoring processes of the robot, safety protection modules keep detecting barriers and prevent the robot from moving out of the track. The remote monitoring center dispatches the robot terminals. By the adoption of the indoor inspection robot system for the substation and the inspection method for the indoor inspection robot system, the working labor intensity is lowered effectively, the substation operation and maintenance cost is lowered, the intelligent level and the automated level of normal inspection work and management are increased, and a detection means and an all-around safety guarantee are provided for intelligent substations and unmanned substations.

The invention relates to a finger clicking character recognition method and a translation method based on artificial intelligence. The recognition method comprises the following steps of (1) respectively constructing and training each neural network; (2) using the acquisition device to acquire the current image of the pre-detection area in real time, and continuously inputting the image into the pre-trained finger positioning neural network to obtain the finger position information under the finger click state; 3) taking that position of the user's fin as the center, intercepting the image ofthe box area, inputting the angle recognition neural network, and outputting the rotation angle of the text in the image area; 4) rotating the rotation angle to intercept the frame area image with theposition of the user's fin as the center, and outputting the position information and the size information of the detected character area; (5) intercepting a corresponding image, inputting the OCR recognition neural network, and outputting the recognized text content. The invention not only improves the identification efficiency but also enables the identification artificial intelligence to be realized.