246results about How to "Avoid low accuracy" patented technology

A method for realizing content filtering gateway based on network filter includes realizing network content analysis and real time monitor as well as realizing purpose of filtering application layer information at network layer based on Linux system platform, utilizing network filter Net-filter frame to carry out collection and filtering on data packet under kernel state for optimizing network path of data packet and for raising collection efficiency of data, furthermore using shared internal memory technique of user space and kernel space to submit session content from kernel state to content filtering module of user state.

The invention provides a counting device and a counting system for trapping injurious insects. The counting device comprises a box body, a trapping chamber (1), a high voltage power grid (2), an insect collecting hopper (3), an infrared sensor unit (4), an image acquisition unit (5), an injurious insect object table (6) and a main controller unit (8). In the counting device and the counting system for trapping the injurious insects, the injurious insects are counted with the combination of the infrared counting technology and the image counting technology at the same time, the problem of a low accuracy rate caused by single counting is solved, meanwhile because the image acquisition unit collects the image information under triggering of the infrared sensor unit, mutual verifying based on two counting results of infrared counting and image counting is achieved, and accuracy of counting the injurious insects is further improved. In addition, in the counting device and a counting system for trapping injurious insects, when the counting results of the two counting methods are different, retrace further can be conducted according to output results of the counting device.

The embodiment of the invention discloses a semantic recognition method and device, a storage medium and electronic equipment. The method comprises the steps that voice data and a scene mode are acquired; the voice data and the scene mode are inputted to a pre-constructed recognition model to perform semantic recognition; and the semantic recognition result outputted by the recognition model and corresponding to the scene mode is acquired, and the corresponding operation of the semantic recognition result is performed. With application of the technical scheme, the recognition model is enabledto have the capacity of accurately "understanding" the real meaning of the voice of the user under different scene modes, the problem of low accuracy rate of semantic recognition due to the fact thatthe corresponding meaning of the natural language cannot be accurately understood can be avoided, and the corresponding semantics of the voice data inputted by the user under different scenes can be accurately and rapidly determined.

Disclosed is a wafer-level fingerprint recognition chip packaging structure and method. The method includes: providing a substrate which comprises a plurality of induction chip areas and is provided with a first surface and a second surface opposite to the first surface, wherein a first surface of each induction chip area comprises an induction area; forming a covering layer on the first surface of the substrate; forming a plug structure in each induction chip area of the substrate, wherein one end of each plug structure is electrically connected with the corresponding induction area, and the other end of each plug structure is exposed out of the second surface of the substrate. By the forming method, the packaging process of a fingerprint recognition chip can be simplified, the requirement on sensitivity of the induction chip is lowered, and the packaging method is wider in application.

The invention relates to a method for detecting suspected online water armies for microblog, and belongs to the technical field of computer application. The method for detecting suspected online water armies for the microblog comprises the following six steps: correspondingly related microblog data collecting; data preprocessing; user characteristics extracting; training set building; water armies detection model training; and unlabeled data predicting and distinguishing. Compared with the prior art, the method for detecting suspected online water armies for the microblog has the advantages that the full utilization of the data is realized, the group discovery is conducted conveniently and quickly, a complicated classification detection model is not needed to be established, so that the complexity of an algorithm is reduced, the modularity of the algorithm is relatively high, the algorithm can be put into large-scale data for calculation, and the stability is relatively high; a single user can be conducted on online water armies detection, a group of users in a certain event can be identified, the modularity of the method is extremely high and the method can be stably applied to large-scale data calculation framework.

It is equipped with an injector body 4z which has formed therein high-pressure paths 6az, 6bz, and 6cz through which high-pressure fuel flows to a spray hole and stores therein a piezo-actuator 2z (i.e., an opening / closing mechanism) and a back-pressure control mechanism 3z (i.e., an opening / closing mechanism) which open or close the spray hole, a stem 51z (i.e., an elastic body) which is installed in said body 4z and elastically deformed when subjected to pressure of the high-pressure fluid, and a strain gauge 52z (i.e., a sensing device) which converts the degree of strain occurring at the stem 51z into an electric signal and outputs it as a measured-pressure signal. Specifically, the stain gauge 52z is installed on the stem 51z that is constructed as being separate from the body 4z.

The invention provides a method and a system for blind people to read Chinese characters, and relates to the technical field of natural language processing and the technical field of disabled-oriented human-computer interaction. The method comprises the following steps: obtaining a Chinese language text, carrying out a word segmentation operation on the Chinese language text to generate a Chinese character string, converting each word in the Chinese character string into corresponding Pinyin by referring to part-of-speech tagging obtained by word segmentation through a pronouncing dictionary, a polyphone dictionary and word frequency information, and connecting the Pinyin into a Pinyin string; looking up a Pinyin and blind character contrast dictionary, converting the Pinyin string into a blind character string, carrying out braille word segmentation on the blind character string through a word segmentation model to generate initial braille segmentation words, fusing the Chinese character string with the initial braille segmentation words to generate new braille segmentation words, and regulating the new braille segmentation words according to a braille segmentation word ligature rule; and carrying out braille tone marking on the new braille segmentation words regulated according to the braille segmentation word ligature rule to generate final braille segmentation words, and displaying the final braille segmentation words.

The embodiment of the invention discloses a video voice conversion method, a video voice conversion device and a server, and relates to the technical field of multimedia processing. The video voice conversion method, the video voice conversion device and the server are used for reducing the translation cost of voice in the video, and improving translation efficiency and translation accuracy. According to the method, the voice signal of a source language in the video is extracted, the voice signal of the source language is segmented, and sub voice signals of at least one section of source language are obtained; for the sub voice signals of each segment of source language, the sub voice signals of the source voice are converted into sub voice signals of a target language according to a pre-built voice model; the obtained sub voice signals of each segment of target language are merged with the video to obtain a video including the voice signals of the target language. When the video voice conversion method, the video voice conversion device and the server provided by the invention are adopted, the voice translation cost in the video can be reduced, and the translation efficiency and the translation accuracy can be improved.

An embodiment of the invention provides a navigation route generating method and device. The navigation route generating method includes acquiring road condition influence parameters corresponding to each road section within an electronic map area range; generating navigation routes from starting points to finishing points according to the road condition influence parameters corresponding to each road section and basic navigation data corresponding to each road section. By the navigation route generating method and device, the navigation routes from the starting points and the finishing points are optimized, so that the acquired navigation routes are kept away from road sections influenced by road conditions; the road condition influence parameters can track dynamic changes of the road conditions of each road section included within the electronic map area range in real time, and frequency of the dynamic changes of the road conditions of each road section is higher than the frequency of batch updating of the basic data, so that the problem of low accuracy in static navigation routes generated simply according to the basic navigation data due to the fact that the basic navigation data cannot reflect actual road conditions of the road sections in real time is avoided, the acquired optimized navigation routes are kept away from the road sections influenced by the road conditions, and correct navigation routes can be provided for users in real time.

The expansion joint reliability fatigue life assessment method comprises the following specific steps that 1, determining weak links of the fatigue life of an expansion joint product, corresponding failure mechanisms and sensitive factors influencing the fatigue life; step 2, performing nonlinear simulation analysis and fatigue life analysis, and establishing a finite element model of the expansion joint; step 3, determining key parameters influencing the fatigue life of the expansion joint according to the sensitive factors determined in the step 1; carrying out statistical distribution fitting on each key parameter by applying an actual measurement value, and finishing multiple random sampling of the key parameters by adopting a Monte Carlo important sampling method; and 4, evaluating the reliability fatigue life of the expansion joint, inputting sample parameters obtained after multiple times of random sampling into the established expansion joint finite element model, carrying outstatistical distribution fitting on a fatigue life analysis result, and obtaining the reliability fatigue life of the expansion joint under a given confidence coefficient.

The invention relates to a voice synthesis method and device, a storage medium and electronic equipment, and the method comprises the steps: inputting a to-be-synthesized text into an information extraction model, and obtaining voice feature information corresponding to the to-be-synthesized text; inputting the voice feature information into a voice synthesis model to obtain acoustic features corresponding to the to-be-synthesized text, the voice synthesis model comprising a duration sub-model and an acoustics sub-model, and the duration sub-model and the acoustics sub-model being subjected tojoint training to obtain the voice synthesis model; and obtaining audio information corresponding to the to-be-synthesized text according to the acoustic features. Therefore, the acoustic features can be directly obtained through the voice synthesis model according to the voice feature information corresponding to the to-be-synthesized text without cooperation of multiple models, so that the voice synthesis efficiency can be improved, the error accumulation can be effectively reduced, and the accuracy of the voice synthesis method is improved.

The invention discloses a video clip positioning method and device, computer equipment and a storage medium, and belongs to the technical field of video processing. The method includes: obtaining a plurality of video frame features of a target video and text features of a target text through a video recognition model; and determining a candidate fragment which can be matched with the target text,finely dividing the candidate fragment based on the matching degree of the video frame in the candidate fragment and the target text to obtain a plurality of sub-fragments, and taking the sub-fragmentwhich has the highest matching degree with the target text as the target video fragment. Such feature matching by multiple times, the obtained candidate fragments are divided again, the mode of the target video clip is determined; therefore, the video recognition model does not need to learn the boundary characteristics of the target video clip, the sample video does not need to be accurately labeled during model training, the training period of the video recognition model is shortened, and the problem of low model output result accuracy caused by inaccurate sample video labeling is avoided.

The invention discloses a text processing model training method, a text method and a device in the field of natural language processing in the field of artificial intelligence. The training method comprises the steps of obtaining a training text; respectively inputting the training text into a teacher model and a student model to obtain sample data output by the teacher model and prediction data output by the student model, the teacher model and the student model respectively comprising an input layer, one or more intermediate layers and an output layer, wherein the sample data comprises sample semantic features output by a middle layer of the teacher model and sample labels output by an output layer, and the prediction data comprises prediction semantic features output by the middle layerof the student model and prediction labels output by the output layer; and training model parameters of the student model based on the sample data and the prediction data to obtain a target student model. According to the technical scheme, knowledge migration is effectively carried out on the student model, so that the accuracy of the text processing result of the student model is improved.

The invention discloses a judging method and system for similar questions. The judging method comprises the following steps: analyzing heterogeneous data information in a target test question pair, wherein the target test question pair comprises a first test question and a second test question, and the heterogeneous data information comprises a text, pictures and knowledge points; transmitting theheterogeneous data information to a preset semantic representation calculation model, and calculating to obtain a first semantic representation and a second semantic representation; separately calculating the first semantic representation and the second semantic representation by adopting a preset similarity calculation method to obtain a similarity goal of the first test question according to the second test question; and judging whether the similarity goal is larger than the preset similarity goal threshold value or not, and if yes, judging that the first test question is similar to the second test question. By the adoption of the method, the similar test questions are judged by adopting the text, the pictures and the knowledge points, and the problem that the only text is used as the judgement condition, and the higher analyticity is lacked, so that the accuracy of the judgement result of the similar test questions is low is solved.

The invention discloses a method, a device and an electronic device for obtaining video heat degree. The method comprises the following steps: obtaining at least one video file and video index data ofthe video file on at least one preset video index; Obtaining an initial heat value of the video file and a time attenuation parameter of the video file based on the video index data; Using the time attenuation parameter, performing the attenuation calculation on the initial heat value of the video file to obtain the video heat value of the video file at the target time.

The invention discloses a risk assessment method, a device, a device and a medium based on micro expression. The method comprises the following steps of: acquiring first video data of an object to beevaluated; Using a microexpression recognition model to perform microexpression recognition on the first video data to obtain the first microexpression data of the object to be evaluated; Establishingfacial microexpression baseline data of an object to be evaluated according to the first microexpression data; Acquiring second video data of the object to be evaluated; Using a microexpression recognition model to perform microexpression recognition on the second video data to obtain the second microexpression data of the object to be evaluated; Using the normal numerical range of each index feature in the facial microexpression baseline data, the risk assessment is performed on the feature data to be identified of each index feature in the second microexpression data, and the risk assessment result of the object to be evaluated is obtained. The technical proposal of the invention realizes automatic risk assessment, and can improve the accuracy rate of credit risk assessment and reduce credit risk.

The invention discloses a keyword extraction method and device, electronic equipment and a medium. According to the invention, after a target text is obtained, the target text can be preprocessed based on a preset mode to obtain at least one candidate word, then a topic calculation model is utilized to determine at least one topic category corresponding to each candidate word, and extraction is performed from the at least one candidate word based on the at least one topic category to obtain at least one target keyword. According to the technical scheme, after irrelevant words of the target text are filtered, the topic calculation model and the target word embedding model are used for extracting the candidate keywords in the target text, and the target keywords interested by the user are obtained. Therefore, the problem of low accuracy caused by only depending on the occurrence frequency of the words to select the keywords in the prior art can be avoided.

The application belongs to the field of medical technology, and discloses a method, apparatus and device for processing a medical image and a medium. The method for processing the medical image disclosed in the application includes: acquiring a first image; the first image being any one of the image frames in a sequence cardiogram; obtaining, by a segmentation model, the region of interest in thefirst image and a first location of the myocardial spot in the region of interest; determining a second location of a corresponding myocardial spot in a second image based on the first location of themyocardial spot; the second image being an Nth image frame adjacent to the first image in the sequence cardiogram; extracting a first feature image of the first location in the first image, and a second feature image of the second location in the second image; and determining, by the first feature image and the second feature image, a motion trajectory of the myocardial spot.

The invention discloses an SOC correction method and device, a battery management system and a storage medium. The method comprises the following steps: collecting battery cell state data meeting a preset standing condition; if the length of the voltage data in the battery cell state data is greater than a first preset length and less than a second preset length, and the change value of the voltage data in the battery cell state data is greater than a preset change threshold, determining a first group of to-be-determined parameters of a preset steady-state battery model according to the battery cell state data; predicting a second group of to-be-determined parameters of a preset steady-state battery model according to the first group of to-be-determined parameters; and calculating to obtain a steady-state OCV estimated value according to the second group of undetermined parameters, and determining an SOC correction value corresponding to the steady-state OCV estimated value by utilizing a preset corresponding relationship between steady-state OCV and SOC so as to correct the current SOC. By adopting the embodiment of the invention, the steady-state OCV can be quickly estimated through the characteristics of the external circuit when the battery stands for a short time, and the applicability of an open-circuit voltage method is improved.

A sound source direction detector comprises FFT analysis sections (103(1) to 103(3)) for generating a frequency spectrum in at least one frequency band of acoustic signals for each of the acoustic signals collected by two or more microphones arranged apart from one another, detection sound identifying sections (104(1) to 104(3)) for identifying a time portion of the frequency spectrum of a detection sound which obtains a sound source direction from the frequency spectrum in the frequency band, and a direction detecting section (105) for obtaining the difference between the times at which the detection sound reaches the microphones, obtaining the sound source direction from the time difference, the distance between the microphones, and the sound velocity, and outputting it depending on the degree of coincidence between the microphones of the frequency spectrum in the time portion identified by the detection sound identifying sections (104(1) to 104(3)) in a time interval which is the time unit to detect the sound source direction.

The invention discloses a disease classification code recognition method and device, computer equipment and a storage medium, and the disease classification code recognition method is used for the computer equipment, and comprises the steps: obtaining diagnosis data; mapping the diagnosis data to an expression space where a disease classification code is located to obtain an intermediate vector corresponding to the diagnosis data; performing similarity measurement on the intermediate vector corresponding to the diagnosis data based on a vector matrix constructed by the intermediate vector corresponding to the disease classification code to obtain the similarity between the intermediate vector in the vector matrix and the intermediate vector corresponding to the diagnosis data; and determining a disease classification code corresponding to the diagnosis data according to the similarity between the intermediate vector in the vector matrix and the intermediate vector corresponding to thediagnosis data. By adopting the disease classification code identification method and device, the computer equipment and the storage medium provided by the invention, the problem of low accuracy of disease classification code identification in the prior art is solved.

The invention discloses a curling test method and device. The curling test method comprises steps that when a curling test signal is received, a movable mechanism is controlled for moving in a set direction and a curling shaft is controlled to rotate around the axis so as to stretch or curl a to-be-tested flexible board, and the movable mechanism is used for fixing a first end of the to-be-testedflexible board, the curling shaft is used for fixing a second end, opposite to the first end, of the to-be-tested flexible board, and the set direction is perpendicular to the axis direction of the curling shaft; in the process of curling or stretching the to-be-tested flexible board, the moving speed of the movable mechanism and / or the angular speed of the curling shaft are / is adjusted so that the stretching and curling speed of the to-be-tested flexible board located on an outermost ring of the curling shaft is equal to the moving speed of the movable mechanism. The curling test method is advantaged in that problems that accuracy of the curling test is reduced, and a flexible display device and the curling test device are easy to damage are solved.

The invention discloses a BPM detection method and a BPM detection device, belonging to the technical field of signal processing. The method comprises the following steps: carrying out signal processing on an audio signal segment, thus obtaining a peak point sequence; determining a first peak point and a second peak point from the peak point sequence; determining the rhythmic pattern according to sequence positions of the first peak point and the second peak point; determining a first candidate point and a second candidate point according to the sequence position and value of the first peak point, the sequence position and value of the second peak point, and the rhythmic pattern; and calculating BPM according to the first candidate point, the second candidate point and the rhythmic pattern. With the adoption of the BPM detection method and the BPM detection device, the problem that after music compression, the position with energy amplitude lower than a preset threshold value is taken as a hold point, and the accuracy rate in calculating BPM according to the periodic value of hold point is low, is solved, and the effect of improving the accuracy rate of BPM detection is improved.

The embodiment of the invention discloses a vehicle remaining mileage determining method and device and vehicle. The method comprises the steps of obtaining the pre-stored traffic flow average speed corresponding to all subsection road conditions, the average acceleration of the vehicle at the current collection moment, the current environment temperature and the consumed power of accessories; determining a current power consumption value of each subsection road condition according to the traffic flow average speed, the average acceleration and the current environment temperature correspondingto each subsection road condition; according to the consumed power and the average speed of each traffic flow, determining the total power consumption of accessories of each subsection road condition; and determining the remaining mileage according to the accessory total power consumption and the current power consumption value of each subsection road condition. The problem that the actual remaining driving range of the vehicle cannot be accurately predicted in the vehicle driving process is solved, and the method is simple and easy to implement; the problem of low accuracy caused by predicting the remaining mileage through historical power consumption information under the condition that the driving working condition changes greatly is avoided, and the accuracy of remaining mileage prediction is improved.

The invention discloses a water area shoreline construction method and system based on an unmanned ship. The method comprises: acquiring the position of an unmanned ship; downloading a satellite map of an area where the position is located according to the position; extracting a water area boundary of the satellite map to obtain an initial boundary; controlling the unmanned ship to sail along theinitial boundary, and acquiring laser point cloud data scanned by a three-dimensional laser scanner in navigation and pictures collected by a camera; extracting a water area boundary according to thelaser point cloud data to obtain a first water area shoreline; extracting a water area boundary according to the picture to obtain a second water area shoreline; and coupling the first water area shoreline and the second water area shoreline to obtain a constructed water area shoreline. According to the unmanned ship-based water area shoreline construction method and system, the accuracy of waterarea shoreline construction can be improved.

The invention discloses a text classification method lacking negative examples, and belongs to the technical field of machine learning and text classification. The method comprises the following steps: firstly, determining to-be-classified data texts, and customizing text classification categories; training a TF-IDF model and an LSI model based on the obtained corpus; respectively constructing feature vectors of the text based on the trained TF-IDF model and the LSI model, and constructing a combined text feature vector based on an ensemble method; secondly, training a Basic classifier by adopting an ROC-SVM combination algorithm, training the Basic classifier in combination with a k-means clustering method, and training a label classifier at the same time; and finally, initially classifying the text to be classified by using a Basic classifier, screening by using Elasticsearch, determining candidate classifications, and accurately classifying the document to be classified into one ormore of the custom classifications by using a label classifier. Text data lacking negative examples can be effectively classified, the accuracy is high, the effect is good, and the efficiency is high.

It is equipped with an injector body 4z which has formed therein high-pressure paths 6az, 6bz, and 6cz through which high-pressure fuel flows to a spray hole and stores therein a piezo-actuator 2z (i.e., an opening / closing mechanism) and a back-pressure control mechanism 3z (i.e., an opening / closing mechanism) which open or close the spray hole, a stem 51z (i.e., an elastic body) which is installed in said body 4z and elastically deformed when subjected to pressure of the high-pressure fluid, and a strain gauge 52z (i.e., a sensing device) which converts the degree of strain occurring at the stem 51z into an electric signal and outputs it as a measured-pressure signal. Specifically, the stain gauge 52z is installed on the stem 51z that is constructed as being separate from the body 4z.

The invention discloses an image similarity determination method and device and electronic equipment. The method comprises the steps: calculating the peak signal-to-noise ratio of a to-be-determined image according to the mean square error of the to-be-determined image and a standard image; And if the peak signal-to-noise ratio of the to-be-determined image is smaller than a peak signal-to-noise ratio threshold value, determining the similarity of the to-be-determined image relative to the standard image according to at least one of the brightness similarity, the contrast similarity and the composition similarity of the to-be-determined image relative to the standard image.