124results about How to "Improve the recognition rate" patented technology

The invention provides a human body movement recognition method based on convolutional neural network feature coding and mainly aims to solve the problems of complicated calculation and low accuracy in the prior art. According to the implementation scheme, TV-L1 is utilized to obtain a video light steam graph; convolutional neural network coding, local feature accumulation coding, dimension-reducing whitening processing and VLAD vector processing are sequentially performed in a video space direction and a light stream movement direction, and space direction VLAD vectors and movement direction VLAD vectors are acquired; and information in the video space direction and information in the light steam movement direction are merged to obtain human body movement classification data, and then classification processing is performed. According to the method, convolutional features are subjected to local feature accumulation coding, so that the recognition rate is increased when complicated background data is processed, and the calculated amount is reduced; the features acquired by fusing video VLAD vectors and light stream VLAD vectors has higher robustness to environmental changes, and the method can be used for performing detection and recognition on human body movement in a monitoring video in areas such as a community, a shopping mall and a privacy occasion.

The invention relates to a checking method and a device of automobile deviation, wherein the method is mainly an image identification and processing procedure, wherein the image identification and processing procedure comprises the following steps: an image acquisition step which comprises utilizing an image pick-up device to extract image picture data, and a lane line identification step which comprises analyzing image picture data, distinguishing a lane line position, utilizing a quadratic curve to fit an equation, connecting a segmental lane line, and making a lane line predicating model, wherein the lane line predicating module which is got through utilizing the above steps is used to check automobile deviation, and is supplied for warning. The checking method of automobile deviation mainly comprises an image pick-up device which is used to extract road image information, a calculating platform which is used to analyze the road image information extracted by the image pick-up unit, and judges the corresponding relationship of a current automobile and a road, and a signal outlet unit which utilizes light, visual variation or sound to reach the behavior of warning. The invention can give off warning to prompt a driver to pay attention and reach the behavior of warning when an automobile does not normally run.

The invention discloses a light guide module of an optical fingerprint identification system. The light guide module guides the reflected light to an image sampling device and comprises a press plate for light transmittance and at least one light diffusion plate. The press plate is provided with a plate surface for a finger to press. The light diffusion plate is arranged between the image sampling device and the press plate and provided with an opening and a plurality of micro structures formed on a surface and guiding the incident lights to disperse on the surface of the press plate, wherein the opening is opposite to the press plate and the image sampling device and allows the reflected light to pass through. Therefore, the micro structure ensures that the incident lights are distributed on the press plate by adopting the micro structures, thereby increasing the light brightness focused on the finger, and further effectively improving the recognition rate.

The invention discloses a real-time online tool wear monitoring method based on wavelet analysis and neutral network. The real-time online tool wear monitoring method comprises steps as follows: step one, detection and transmission of three-way cutting force; step two, wavelet analysis processing; step three, normalization processing; step four, determination of input of the neural network; step five, neural network processing. The method has simple steps and is convenient to implement, the tool wear state can be rapidly and efficiently detected, the processing quality can be guaranteed, and the production efficiency is improved.

A sound control method and device using single push-button to control the speech instruction is disclosed. A speech input unit is used to input the speech instruction information from a user, and convert it to digital information. A speech instruction module can receive the speech instruction information of the speech input unit and compare it with the prestored speech instruction information to generate corresponding speech instruction code and information which are received by a speech control host to control the devices to be controlled. The speech information is converted to the analog signal to generate sound for pilotting the user.

The invention belongs to the technical field of network security, and particularly relates to an Android malicious software detection method and device based on a capsule network, and the method comprises the steps: collecting an Android software file sample, decompressing a to-be-processed file, converting the to-be-processed file into an RGB three-channel color image, and enabling the RGB three-channel color image to serve as training sample data; constructing a capsule network, and training the capsule network by using the sample data to obtain a trained network model containing a graph structure and network parameters, the capsule network realizing transmission between feature vectors in a capsule layer through an iterative dynamic routing algorithm; and inputting the to-be-detected target file into the trained capsule network model for testing, and judging whether the to-be-detected target file is a malicious software file or not through an output result. The method and device canefficiently run on the Android operation platform, occupy few resources, are high in efficiency and accuracy, can realize high-accuracy classification detection tasks even under the condition of small-scale training samples, and achieve the purpose of protecting the Android intelligent mobile terminal.

The invention provides a window adjusting interface and an adjusting method thereof. through a technical means that the display size of an operating window is adjusted by vision parameters, the problem that in the prior art, the display of the operating window in a fixed size mode generates difficult identification of the operating window due to the differences of operating distance and vision degree of users can be solved, thereby being capable of achieving the technical efficacy of improving the identification rate of the operating window.

The invention relates to a visual drive method, control method and control system of a brain wave human-computer interface. In the visual drive method, a plurality of flickering light sources are displayed for user observation, and different phase codes are made for the flash time sequence of the flickering light sources by utilizing time delay; and the visual evoked potential of a user is triggered by the flash time sequence for subsequent control, and the purpose of increasing system use comfort is achieved.

Disclosed are an identity recognition system and method based on hybrid biometric which capture an image of a subject that is projected by light of different wavelengths, extract various biometric information from the image, analyze and compare for each of the biometric information to generate a matching score, and determine an identity for the subject according to all of the matching scores. The system and method have higher recognition accuracy, lower false acceptance rate, lower false rejection rate, and higher flexibility.

The invention relates to a system and a method for charging billing of an electric vehicle during network anomalies. During the network anomalies, a user holds a terminal by hand to connect a centralized control through a WiFi network and submits a charging request of the user to the centralized control, after the centralized control compares a VIN (vehicle identification code) with a post blacklist stored in the centralized control after acquiring the VIN, and if the VIN is not in the blacklist, the centralized control starts a charger to perform charging billing on the electric vehicle through a charging module, stores charging information and arrearage data, then uploads the charging information and the arrearage data to a cloud server when the network is normal, and requests a newest post blacklist from the cloud server; the cloud server sends arrearage information to the user after receiving the charging information and the arrearage data which are uploaded by the centralized control; the cloud server sorts a new post blacklist and sends the new post blacklist to centralized controls of various charging stations; the cloud server receives payment information of the user and updates the blacklist. According to the method, the charging request is finished through access of a local area network; the vehicle is positioned uniquely by acquiring the VIN of the vehicle, the identification rate is high, and the feasibility is good.

The invention specifically provides an electronic tag test system and a test method thereof, and belongs to the technical field of radio frequency recognition. The test system comprises an antenna, a radio frequency front end, an analog front end, a data collecting module, a configurable digital baseband and control module, a power supply, an auxiliary module, a data storage device and an interface thereof. The data collecting module collects the output data of the analog front end, the output data is transmitted to the digital baseband and control module, and temporarily stored in the data storage device, and the interface is connected to a PC after the testing process is completed. The interface controller of the digital baseband and control module controls the interface, and the data temporarily stored in the data storage device is transmitted to the terminal of the PC. The electronic tag test system and test method thereof can real-time record the electromagnetic characteristics of the electronic tag in application environment and variation condition thereof via a system level testing tag, perform theoretical analysis by guiding into a computer via a standard interface, search the reason that the electronic tag is not recognized, provide basis for selecting appropriate optimistic method, and improve the recognition rate of an RFID system.

The invention provides a classification method based on a semi-supervised extreme learning machine with a deep structure. Firstly automatic feature learning is performed on original data by utilizing a deep structure to obtain advanced expression of data features; and then similarity measurement between marked and unmarked sample features is calculated, and new features are classified by utilizing the semi-supervised extreme learning machine, so that the classification accuracy is improved. On one hand, the problem that a cascaded extreme learning machine performs learning only by utilizing marked samples and ignores useful information in unmarked samples is solved; and on the other hand, the problems of insufficient learning of sample features and the like due to the fact that the semi-supervised extreme learning machine is limited by a single-layer network structure are solved and finally the purpose of improving the network generalization performance and the classification accuracy is achieved.

It provides an input device and its method. It comprises at least an input interface, a processing unit and an execution unit. The input interface is similar to a touch screen or a touch panel to allow the user to input by a finger or touch pen to show the hand writing. Processing unit can analyze the handwriting. Last, the execution unit will execute a preset function based on the comparison of the bend trend.

The invention relates to a voice emotion recognition method, which comprises the following steps of obtaining a plurality of voice signals from a voice database; combining a plurality of voice features, a plurality of nonlinear features and a plurality of macroscopical features into a plurality of groups of feature combinations; and extracting the voice features, the nonlinear features and the macroscopical features of each first voice signal in the voice signals so as to obtain a plurality of calculated features. The method also comprises the following steps of selecting a plurality of target feature quantities from the calculated features; using the target feature quantities to train one voice emotion classification model so as to obtain a plurality of recognition rates corresponding to the feature combinations; and selecting a plurality of target feature combinations from the feature combinations according to the recognition rate. The method also comprises the step of executing the voice emotion recognition operation according to the target feature combinations and the voice emotion classification model.

The invention provides a flip-chip welding spot defect detection method through temperature observation and measurement, and relates to a flip-chip welding spot defect detection method. The method includes the steps of: arranging a thermal imaging system at one side of a flip chip, arranging an infrared laser at one side of a substrate, aligning an infrared laser beam to a to-be-detected welding disc of the substrate of the flip chip, adjusting power and pulse width parameters and performing thermal excitation on the power and pulse width, wherein the thermal imaging system detects the temperature rising process of a chip ball-welding zone connected with the welding disc in real time, observes and shoots a thermal image of a highest temperature rising point, and judges a flip-chip welding spot defect according to the temperature rising curve or the thermal image. The flip-chip welding spot defect detection method for empty welding adopts point-by-point detection and has the characteristics of no damage, high defect recognition rate and simplicity and intuition in judgment; and besides, the method has a wide application range, can be suitable for the detect detection in chip side ball placement and the detect detection in substrate side ball placement during three-dimensional assembly.

The invention discloses a method and a device for extracting music characteristics. The method comprises the following steps: segmenting a received audio signal to generate at least two segmented audio signals; carrying out Fourier transform on each segmented audio signal to obtain a frequency domain signal of each segmented audio signal; calculating a frequency mass center of each segmented audio signal according to the frequency domain signal of each segmented audio signal and a corresponding frequency of the frequency domain signal, and taking the frequency mass center as a music characteristic; calculating bandwidth of each segmented audio signal according to the frequency mass center and the frequency domain signal and the corresponding frequency of the frequency domain signal of each segmented audio signal, and taking the bandwidth as a music characteristic. Through the method and the device, the received audio signal is segmented; the mass center and the bandwidth of each segmented audio signal are calculated according to all the frequency information meeting the condition in each segmented audio signal; and the mass center and the bandwidth serve as the music characteristics, thereby reducing the influence of the environment on the audio signal characteristics and improving the recognition rate of a system.

The invention discloses a multi-feature weight adaptive shadow elimination method. The method is characterized by, to begin with, obtaining a background image of a video frame image to be processed and an initial foreground region; then, obtaining foreground masks corresponding to chrominance, spectrum direction and texture features; starting from different angles, setting a quality evaluation function of a foreground segmentation image; carrying out quality evaluation on the foreground masks under different features; setting fusion coefficient of the three foreground masks according to the evaluation results and carrying out weight fusion to obtain corresponding time domain foreground probability spectrum; and carrying out constraint on the current detection result through time domain relativity to realize the multi-feature weight adaptive vehicle shadow elimination method. The method prevents the problem that features cannot make up disadvantages of each other in a single-feature or multi-feature cascade method; and while extracting all shadows as far as possible, shadow false detection is reduced, and vehicle shadows can be eliminated more completely and accurately.

A light source device which is uniform in the intensity of distribution of an emitted light and high in the optical utilization efficiency, a display device which is high in recognability and is energy-saving, and an information terminal are provided. A reflecting portion is formed on a surface opposed to an emitting surface of a linear light guide. The reflecting portion is periodically formed with a total reflecting surface guiding a light incident on a linear light guide from a point light source, a light extracting surface allowing the light to reflect in such a way as to emit the light guiding the linear light guide from the emitting surface, and a re-incidence plane which takes the light transmitted but not reflected on the light extracting surface into the linear light guide again. By changing the size of the light extracting surface and the re-incidence plane, the intensity of distribution of the emitting light from the linear light guide can be made uniform.

The invention discloses an elevator safety early warning method based on machine learning. Automatic motion matching, noise interference processing, the super resolution and other methods are adopted,effective information is subjected to fusion processing, and related features are extracted to conduct dynamic capturing and early warning. According to the elevator safety early warning method, thevideo information is dynamically captured to be correlated to an automatic early warning mechanism, so that a security check person is reminded of appearance of targets so as to take measures in time;and meanwhile, videos of related information are automatically extracted so that maintaining personnel can conduct series and incorporate case analysis and video comparison analysis treatment, and valuable clues and the like are found.

The present invention relates to a behavior identification system and an identification method combined with audio and video, wherein the system comprises a data analysis module, a database and a calculating module. The database is used for storing a plurality of different audio-video relation modules, wherein each audio-video relation module comprises a characteristic acquisition parameter and an audio-video relation parameter; the data analysis module obtains corresponding gesture images and voice data and introduces them into each characteristic acquisition parameter, so as to form image characteristic sequences and voice characteristic sequences corresponding to each audio-video relation module, and the data analysis module calculates a plurality of audio-video state parameters by using each audio-video relation parameter; and the calculating module calculates the identification probabilities corresponding to each audio-video relation parameter by using the audio-video state parameters, the image characteristic sequences and the voice characteristic sequences, in order to select the maximum of all the calculated probabilities as the object parameter.

The invention discloses a lighting device for a motorcycle. The motorcycle comprises a vehicle body, a vehicle head pivoted on the vehicle head, a front wheel arranged at the front end of the vehicle body, a rear wheel arranged at the rear end of the vehicle body opposite to the front wheel and power unit driving the front wheel and a rear wheel. The vehicle body is provided with a vehicle body front cover, and the vehicle head is provided with a vehicle head front cover. The lighting device comprises a headlight set, as least one position light set and two directional light sets. The headlight set is arranged on the vehicle head front cover, and the position light set and the directional light sets are arranged on the vehicle front cover. The light beams projected from the large light set, the position light set and the directional light sets do not interfere one another. Thus, the identification rate of the motorcycle is improved and the riding safety of the motorcycle is improved effectively.

The invention provides a structure for improving fingerprint identification. The structure comprises a first glass layer, a lower film layer, an optical film layer and a fingerprint identification module. the optical film layer is arranged between the first glass layer and the lower film layer, the lower film layer is provided with a through hole, and the fingerprint identification module is arranged in the through hole of the lower film layer. The invention belongs to the technical field of electronic products. The structure for improving fingerprint identification not only can improve the identification rate of products, but also can maintain the strength of a cover plate; and meanwhile, the structure for improving fingerprint identification can prevent water blocking or moisture entering.

The present invention relates to a multimedia control apparatus based on gesture recognition, which comprises a positioner box, a signal acquisition box, a gesture acquisition camera, a driving mechanism, a human body proximity sensor, a photosensitive sensor, an auxiliary power supply and a control system, wherein the signal acquisition box is articulated with the positioner box by the driving mechanism; the control system is mounted in the positioner box; the gesture acquisition camera, the human body proximity sensor, the photosensitive sensor and the auxiliary power supplyare all mounted in the signal acquisition box; the control system is respectively and electrically connected with the gesture acquisition camera, the driving mechanism, the human body proximity sensor, the photosensitive sensor and the auxiliary power supply in addition; and the control system comprises a data processing module, a communication module, a gesture recognition module, a data storage module and a driving module. The multimedia control apparatus based on gesture recognition has a simple and compact structure, is reasonable in layout and has a high gesture recognition rate, high recognition accuracy, a high operation automation degree and high site and environment adaptability; and meanwhile, the optimal gesture recognition angle also can be automatically regulated according to the position of a human body.

The invention belongs to the field of online flutter identification in the cutting process of a numerical control machining center, and particularly discloses an online flutter identification method for a numerical control machining center based on a vibration signal. The method comprises: collecting spindle vibration signals generated when a numerical control machining center conducts cutting machining, and preprocessing the vibration signal; constructing a multi-scale permutation entropy and a multi-scale power spectrum entropy of the preprocessed signal; splicing the multi-scale permutationentropy and the multi-scale power spectrum entropy to serve as feature vectors to be input into a constructed gradient boosting tree model; and performing iterative training on the vibration signal to obtain an optimal gradient boosting tree model, preprocessing the main shaft vibration signal in the machining process , and then inputting the preprocessed main shaft vibration signal into the optimal gradient boosting tree model, so that on-line identification of flutter of the numerical control machining center is achieved. According to the method, the chatter sensitivity characteristics of the machine tool can be extracted, whether the numerical control machining center chatters or not and the chatter severity can be identified, and the method has the advantages of being high in monitoring real-time performance, high in identification accuracy, good in generalization ability and the like.

A method for identifying character of hand-writing input carry out treatment on inputted stroke by utilizing corner cut of included angle between each adjacent stroke sections in character of hand-writing input for adjusting position and structure between stroke sections so as to pick up feature information of character inputted by hand-writing for carrying out identification.

The invention discloses a space-time multiple feature based vehicle shadow eliminating method. A background image and an initial foreground area of a video frame to be processed are obtained, a foreground mask layer of chroma, spectrum direction and texture features is obtained, weighted summation is carried out to obtain a spatial-domain multi-feature foreground probabilistic spectrum image, time-domain filtering of a time sliding window is carried out on the spatial-domain multi-feature foreground probabilistic spectrum image to obtain a corresponding time-domain foreground probabilistic spectrum, the time-domain foreground probabilistic spectrum is weighted to obtain a final foreground mask layer, and vehicle shadows are eliminated from a video frame image. Compared with a traditional shadow removing method based on a single feature and multi-feature cascading, the shadows are removed more effectively, and a foreground contour is more complete. Both higher shadow identification rate and higher shadow detection ratio are ensured.

The invention provides an image capturing device and an image capturing method. The image capturing device comprises an image capturing unit, a filtering module, a light sensing unit, an infrared light source and a processing unit. The filtering module is arranged near the image capturing unit. The filtering module comprises a first filter lens and a second filter lens. The first filter lens is used for filtering out infrared rays and allowing visible light to pass through. The second filter lens is used for filtering out the visible light and allowing the infrared rays to pass through. The light sensing unit is used for sensing the ambient light. When the processing unit determines that the ambient light intensity is higher than a threshold, the processing unit alternatively switches the first filter lens and the second filter lens to the front side of the image capturing unit, controls the image capturing unit to capture a visible light image through the first filter lens and controls the image capturing unit to capture an infrared image through the second filter lens. The image capturing device performs differential identification through front and back images and can also identify faces, objects, plate numbers or other specific information in a single image.

The invention discloses an X-ray transmission inspection method. The X-ray transmission inspection method comprises the steps of (1) starting up a detection device to check and clean foreign matters inside an inspection channel; (2) placing an article to be checked on a conveying belt of the detection device, starting up an X-ray generator to scan the article which is to be checked and enters the inspection channel, and transmitting obtained images to a main machine of the detection device in real time; (3) reflecting detailed image information of the article to be checked after the main machine of the detection device conducts analytic calculation and image processing on images; and (4) discriminating and storing the detailed image information of the article to be checked through the main machine of the detection device, and processing the article to be checked. According to the X-ray transmission inspection method, the X-ray generator is used for scanning the article which is to be checked and enters the inspection channel, image processing of different degrees is conduced after images of the article to be checked are obtained, the detailed image information of the article to be checked is reflected to conduct article safety identification, the whole process is simple and fast, the discrimination rate is high, and the inspection cost can be saved.

The invention discloses a mobile phone insurance method based on AN intelligent terminal, and the method comprises the following steps that a mobile phone data line is plugged, a terminal applicationprogram reads and stores relevant parameters and information such as a unique IMEI number, an MAC address, mobile phone screen information and mobile phone spare and accessory parts of a mobile phone,the terminal emits a hot spot, the mobile phone is connected with the hot spot, a terminal application program acquires mobile phone data and allocates a random code to a mobile phone MAC address togenerate a two-dimensional code, a color mark is selected to send an instruction, a terminal photographs and uploads the instruction to a cloud for comparison, identification and detection, and an insurance order is generated to store an insurance policy record for the mobile phone of which various indexes of the mobile phone meet insurance conditions. By utilizing the mobile phone insurance method, the mobile phone can be prevented from being dropped during insurance, the risk of insurance fraud during insurance of the mobile phone is avoided, the labor cost of an insurance company is saved,the experience and insurance rights and interests of a mobile phone user are facilitated, the insurance efficiency of the mobile phone is improved, and the market prospect is wide.