239results about How to "Reduce the difficulty of collection" patented technology

The invention discloses a method for culturing high-quality pleurotus geesteranus by using waste pleurotus eryngii residue. The method is characterized in that: when raw materials are prepared and treated, the culture material composition for production, namely the formula composition of raw materials, comprises the following raw materials in part by weight: 400 to 700 parts of waste pleurotus eryngii residue, 200 to 300 parts of sawdust, 300 to 500 parts of corn straw, 100 to 200 parts of cotton seed hull, 200 to 350 parts of bran, 10 to 20 parts of gypsum, 15 to 15 parts of lime, and 10 to 1 5 parts of white sugar. A proper amount of sodium selenite serving as a clearing auxiliary agent is added into each 100 kilograms of dry culture material. By the method, the waste is utilized, the cost of the raw materials is lowered, the resource waste is reduced, environmental pollution is lightened, the sources of the raw materials are easily solved, the process is simple, labor and effort are saved in management, and stable yield and high yield are achieved; and by adding a little amount of sodium selenite (Na2SeO3) serving as the clearing auxiliary agent, the cadmium content of the pleurotus geesteranus finished product is reduced, and the production and the export scale are enlarged.

The invention relates to a method and a system for predicting the diffusion of pollutants. The method comprises the following steps: S1, constructing a Gaussian diffusion model of flue gas dischargedfrom a target flue gas discharge port according to the diffusion condition; S2, collecting an image of the smoke emitted from the target smoke discharge outlet, and performing image processing on theimage to obtain a trajectory equation of the smoke emitted from the target smoke discharge outlet; S3, predicting the spatial concentration distribution of the flue gas discharged from the target fluegas discharge outlet according to the trajectory equation, and correcting the Gaussian diffusion model by comparing the collected real value of the flue gas discharged from the target flue gas discharge outlet; S4, predicting the trajectory diffusion of pollutants in the flue gas discharged from the target flue gas discharge outlet by using the modified Gaussian diffusion model and the association rule algorithm. The method of the invention is based on the Gaussian diffusion model and the image recognition technology, can effectively reduce the difficulty of data acquisition and can optimizethe prediction effect.

The invention discloses a logistics delivery route optimizing generation method and system. The logistics delivery route optimizing generation method comprises the following steps: firstly, obtaining and combining client information in distribution areas to form distribution districts; then, selecting the district which takes the longest distribution working time from a logistics center as a calculation source point according to delivery car capacity; accumulating client delivery capacity according to client linear distribution serial numbers in the district or a next delivery district which takes the shortest working time from the source point until the delivery car capacity is satisfied; and finally, finishing a total delivery amount, and outputting the selected client linear distribution serial numbers as an optimal delivery route. A proper delivery path is calculated by adopting a reverse recursion Dijkstra algorithm, the district is used as a basic calculation unit, a distribution network is simplified, a network calculation amount is greatly lowered, information acquisition difficulties and information acquisition time are greatly lowered and shortened respectively, the amount and the dispatch frequency of the delivery cars are reduced, operation cost is lowered, and distribution efficiency is improved.

The invention discloses a method for improving a hippocampus kelloggi fry cultivating rate. The method for improving the hippocampus kelloggi fry cultivating rate is characterized in that nutrition is fortified on main animal bait for hippocampus- Artemia by using an unsaturated fatty acid emulsifier and microalgae so as to compensate unbalance of the trophic structure of the bait, in particular to increased and balance the content of the unsaturated fatty acid. After the bait is fortified based on the method, the bait is continuously fed to a young hippocampus, thereby significantly improving the bait feeding rate of the hippocampus; and by soaking and cleaning the fortified bait, and the survival rate and the growth rate of the corresponding hippocampus are significantly increased.

Embodiments of the present application disclose a vehicle violation monitoring method based on a block chain technology and a related equipment, which are used for reducing the difficulty of vehicle violation data collection to a certain extent. The method of the embodiment of the application comprises the following steps of using the block chain node equipment to obtain the violation informationuploaded by the monitoring user, register the violation information in the block chain system and broadcast the violation information throughout the network; if it is determined that the violation information passes the preset verification, generating a violation payment contract according to the violation information; identifying a violation user corresponding to the violation information according to the violation information; pusing the illegal payment contract to the illegal user so that the illegal user completes payment according to the illegal payment contract, wherein, when the block chain node device determines that the violation information passes the preset verification, the violation payment contract is generated and sent, so that the violation user completes payment accordingto the violation payment contract, and accordingly the difficulty of vehicle violation data collection is reduced to a certain extent.

The invention discloses a human body fall detection method and device. The method is applied in a top mounting scene and comprises the following steps of: obtaining a current frame of image; determining a first area, determined by at least one preset human body recognition part, of a to-be-detected target in the current frame of image according to the human body recognition part; determining a current speed of the to-be-detected target according to the first area in the current frame of image and a second area corresponding to a to-be-detected target in a first image which is located in frontof the current frame and spaced from the current frame by a preset number of frames; judging whether the current speed is greater than a stored normal movement speed range of the to-be-detected target; if the current speed is greater than the normal movement speed range, judging whether an area of the first area is smaller than a stored corresponding area threshold value when the to-be-detected target moves normally or not; and if the area of the first area is smaller than the area threshold value, judging that the to-be-detected target falls. According to the method and device, movement speeds of to-be-detected targets and areas, determined by human body recognition parts, of areas are detected, so that the problems that the acquisition difficulty is high, the human body detection rate islow and the human body fall event detection rate and correctness are relatively low are solved.

The invention provides an offline commodity sales volume information collecting method. The offline commodity sales volume information collecting method comprises the following steps of establishing and storing commodity information; establishing and storing information of a two-dimensional code group; establishing and storing the information of the relation between the two-dimensional code group and commodities; generating and distributing the two-dimensional codes; scanning the two-dimensional codes and visiting a code scanning rewarding system after the commodities are sold; receiving rewarding information; verifying whether check codes are correct or not; verifying serial numbers if the check codes are correct; returning rewarding results if the check codes are not correct; verifying whether the serial numbers are rewarded or not; returning the rewarding results if the serial numbers are rewarded; storing a serial number query record if the serial numbers are not rewarded; storing rewarded information of the serial numbers; transmitting sale information to a sale statistical system; storing code scanning award information; and returning rewarding of the rewarding results. By the offline commodity sales volume information collecting method, the sale information collecting time can be shortened and the cost can be reduced.

The invention provides an Internet financial debit and credit management system and method, which belong to the field of information technology. Through operation of an electronic account opening and signing unit, an electronic data storage unit and a notary office inspection unit under the Internet financial debit and credit management system, the Internet financial debit and credit management method is fully applied to Internet financial management in order to make debit and credit information evidence preserved in time, realize online compulsory enforcement notarization and reduce the difficulty of money collection after loan. Meanwhile, because the default information of dishonest borrowers is input to a third party population information database, the default cost of dishonest borrowers is increased, and certain deterrence can be produced to urge borrowers to make repayment actively. Therefore, through the system and the method, the problem of overdue loans and difficult post-loan disposal in the process of Internet financial operation and management can be solved, and the construction of the Internet financial management system can be promoted.

The invention discloses an interdisciplinary recommending method and system. The method comprises the steps of determining behavior preference data of a user in multiple fields, and then utilizing a recommendation algorithm and the behavior preference data to generate a recommendation result for the user. This shows that the behavior preference data is obtained from multiple fields instead of being obtained from a single field, in this way, data sources in an information recommending process can be obviously drastically enriched, and the data acquisition difficulty is thus lowered; meanwhile, the behavior preference data of the user in multiple fields can more comprehensively embody the individual interest of the user, and thus the follow-up recommendation result can be sharply improved.

The invention discloses a bitcoin address classification method based on an improved random forest. The method comprises the steps of constructing a feature set, forming a data set, obtaining a sampleset with a label, initializing parameters of a learner, iterating the learner, obtaining key features and the like. From the perspective of Bitcoin market supervision, the problem of judging whetherthe user participates in illegal transaction is converted into the Bitcoin address classification problem, and market supervision can be perfected; the sample set is directly obtained through the historical transaction record of the block chain, so that the data collection difficulty is greatly reduced; bitcoin addresses can be classified with the accuracy of about 84%, and only few statistical characteristics are needed. Compared with the prior art, the method has the advantages that the addresses can be well classified and a large number of constructed statistical characteristics are subjected to redundancy elimination; and after the learner completes final training, the key characteristics needing to be extracted finally can be obtained, so that for one address needing to be recognized,the data collection time is shortened, and the time expenditure of address classification is also reduced.

The embodiment of the invention provides a touch display screen used for collecting a fingerprint. The touch display screen comprises a touch panel, a fingerprint collection layer and a backlight module, wherein the fingerprint collection layer is located between the touch panel and the backlight module, the projection of the touch panel on the fingerprint collection layer is contained in the fingerprint collection layer, the touch panel determines a finger touch region and changes the state of the region into a light-transmitting state, light of the backlight module reaches the fingerprint through the fingerprint collection layer and the region in the light-transmitting state, and therefore the fingerprint collection layer collects the fingerprint. According to the touch display screen and an electronic device, external light (namely noise) cannot enter the liquid crystal panel through other regions, therefore, the situation that the external light disturbs the fingerprint collectionprocess is avoided, and fingerprint collection difficulty is lowered.

The invention provides a river mouth runoff and tide control reach judgement method. The method comprises the following steps of selecting runoff and tide characteristic indexes; taking tidal range observation stations as target observation stations, utilizing an M-K method, and when the runoff flow of a runoff observation station is greater than a certain flow level and the occurrence frequency of the flow level in a hydrologic P (as shown in the specification) type curve of the corresponding runoff observation station exceeds 90%: if the fact that UF or UB exceeds a confidence interval a=0.95 always exists, calling reaches of the target tidal range observation stations as perennial runoff control reaches, and otherwise, calling the reaches of the target tidal range observation stations as non-perennial runoff control reaches; and for the target tidal range observation stations of the non-perennial runoff control reaches, judging seasonal runoff control reaches and perennial tide control reaches by utilizing the MK method on the basis of flood season or dry season runoff flows and monthly tidal range data of each station. According to the method provided by the invention, theoretical support is provided for the construction of water conservancy projects, so that the implementation of related projects in river mouth areas has better pertinence for the runoff and tide control conditions.

The invention discloses a method and a device for acquiring the position of a terminal, and the method comprises the steps: acquiring an MDT measurement report; judging whether the MDT measurement report contains latitude and longitude information or not; if the MDT measurement report contains latitude and longitude information, determining the position of the terminal according to the latitude and longitude information contained in the MDT measurement report; and if the MDT measurement report does not contain the latitude and longitude information, calling a predetermined position identification model, predicting the latitude and longitude information of the terminal, and determining the position of the terminal according to the predicted latitude and longitude information. According to the invention, the target of obtaining the full coverage of the terminal position can be relatively easily achieved, and the MDT measurement report generally comprises the position information, so thatmodel creation is carried out based on the MDT measurement report, and the sample collection difficulty and cost can be effectively reduced.

The invention discloses an image detection method and device and a storage medium, and the method comprises the steps: obtaining a to-be-detected image of a target object; determining the similarity between the to-be-detected image and a template image in a template image set, wherein the template image set comprises a standard image of a target object generated based on a standard image generation model; obtaining a target template image of which the similarity degree reaches a preset condition from the template image set; determining a detection result of the to-be-detected image according to a comparison condition of the to-be-detected image and the target template image, wherein the standard image generation model comprises an image generation model obtained by training a generative adversarial network model based on an original image pair of a target object in a normal state. According to the invention, the collection difficulty of training images is reduced, and a lot of manpowerand material resources are saved.

The invention discloses a greenhouse fruit and vegetable receiving and conveying robot, which relates to the field of agricultural robots. The greenhouse fruit and vegetable receiving and conveying robot consists of an autonomous moving trolley (1), a rotary disc fruit receiving mechanism and a fruit dumping and unloading mechanism. During operation, the greenhouse fruit and vegetable receiving and conveying robot follows a pick-up robot; and four fruit storage boxes are driven by a rotary disc motor (8) in the rotary disc fruit receiving mechanism to rotate to collect the picked fruits and vegetables. The autonomous moving trolley (1) is guided by electromagnetism to convey the picked fruits and vegetables to a storage or post-production treatment region along a walkway; subsequently, a cam (11) in the fruit dumping and unloading mechanism rotates to press a bottom plate (15) to tilt outwards; and meanwhile, an electromagnet (16) is cut off to open a box door (14) to unload the picked fruits and vegetables. The four fruit storage boxes are driven by the rotary disc motor (8) to rotate, and meanwhile, the fruit dumping and unloading mechanism repeats the action until the fruits and vegetables are completely unloaded. The difficulty in collection of the fruits and vegetables is greatly reduced, the bearing capacity is increased, and automation of collection, transportation and fruit unloading of the picked fruits and vegetables is realized.

The invention relates to the field of data processing, and provides a label labeling method and device based on artificial intelligence, electronic equipment and a medium, and the method comprises thesteps: receiving a plurality of to-be-labeled pictures comprising a target body, carrying out the preprocessing of the to-be-labeled pictures, and obtaining a plurality of preprocessed pictures, classifying the preprocessed pictures of which the similarity values are greater than or equal to a first preset threshold value into one class; calculating each preprocessed picture in the same type of preprocessed pictures with a preset picture to obtain a similarity value set corresponding to each preprocessed picture, and taking the preset picture of which the similarity value is greater than or equal to a second preset threshold value in each type of similarity value set as a recommendation result of the preprocessed pictures; and displaying the recommendation result to the user, so that theuser selects a preset label of a preset picture with the highest matching degree with the specified preprocessed picture as a target label of a target body in the specified preprocessed picture. According to the invention, the difficulty of sample data collection can be reduced, and the recognition accuracy of an artificial intelligence system is improved.

The invention provides a positioning data acquisition method and system. The positioning data acquisition method comprises a map generation mode or a task mode; in the map generation mode, an acquisition platform performs multi-sensor synchronous mapping and positioning according to manual control of a user; according to the established map, the acquisition platform performs map data optimizationpost-processing and map data storage and uploading; in the task mode, according to the map of the current area selected by the user, a mark starting position in the map is selected to assist system initialization; after initialization is completed, the acquisition platform determines the position of the acquisition platform through a multi-sensor fusion positioning algorithm; after initializationis completed, the user inputs a planned target path point to the acquisition platform. According to the invention, the multi-sensor positioning algorithm based on map optimization is used to improve the positioning robustness of the unmanned acquisition platform in a complex environment, so that the data acquisition platform can still perform robust positioning in a complex scene without failure.

The invention relates to a method for analyzing moisturization of a sea level by linear fitting according to adjacent wind speed and wind direction data. According to the method, a math relation formula is constructed between wind situation data and sea level moisturization data according to historical sea level data and astronomic tide data of a sea level monitoring station on the basis of peripheral wind situation stations; and a future forecast value is calculated by using the formula. By the method, required historical information data are single; hydro meteorology departments of different places have the type of data, so that the difficulty in acquisition of the data required by forecasting is greatly and effectively reduced, and a relatively good data environment is supplied to successful running of sea level forecasting; moreover, the method is relatively low in setting cost; the moisturization situation can be analyzed by using the main wind direction of a definite tidal river, so that the arrangement cost and the execution cost are greatly reduced, and forecasting popularization of coastal seal level monitoring stations is facilitated.

The embodiment of the invention discloses a service data processing method and device and related equipment, and the method comprises the steps: obtaining the target sample user information of a target sample user from a target service sample set, and obtaining a target service object having a binding relationship with the target sample user information; obtaining a prediction service object corresponding to the target sample user information from the initial service prediction model; determining a sample reward business volume according to the target service object, the prediction service object and the maximum prediction service volume, and adjusting an initial service prediction model according to the sample reward service volume, the prediction service object and the target sample userinformation; and when the adjusted initial service prediction model satisfies a target convergence condition, determining the adjusted initial service prediction model as a target service predictionmodel. By adopting the method and the device, the service profit between each user and various service objects can be efficiently and accurately predicted.

The invention discloses a method for case detection assisted through voiceprint. The method comprises the following steps: receiving the collected voice to be recognized; analyzing the voice to be recognized, so as to generate corresponding voiceprint; comparing the voiceprint of the voice to be recognized with the original voiceprint in the voiceprint library, and determining the comparison result of the original voiceprint matched with the voiceprint of the voice to be recognized, wherein the voiceprint library is established in advance, stores a plurality of original vocal prints and a database of the identification information corresponding to the original voice prints; and determining the identity information of the criminal investigation object corresponding to the voice to be recognized according to the comparison result obtained by matching. The invention provides a method and device for case detection assisted through voiceprint, the telephone receiver, the mobile phone, and the microphone can be used for collection, the cost is reduced, and the use scene is wider; due to the fact that the voiceprint library is adopted, the storage and searching of the voiceprint data areconvenient, and the operation efficiency of the public security system is improved.

The invention provides a surface garbage salvage vessel which comprises a vessel body, a garbage salvage device, garbage collecting devices, a garbage conveying device and a control center. The garbage salvage device is arranged at the front end of the vessel body, the garbage collecting devices are arranged on the periphery of the vessel body and are fixedly connected to the vessel body through supporting pillars at bottoms of the garbage collecting devices, garbage dampers are arranged on the periphery of the garbage conveying device, a garbage conveying belt is arranged on the surface of the garbage conveying device, and the control center is arranged in the vessel body. By the arrangement, garbage on the water surface can be cleaned through the salvage vessel, automation degree is high, problems about low manual salvage efficiency, serious potential safety hazard, high salvage cost and small salvage range are effectively solved, the surface garbage salvage vessel is compact in structure and continuous and convenient in operation, difficulty in collecting the garbage on the water surface is reduced, and the environment pollution bought by the garbage of the water surface is alleviated.

The invention provides an optical cable fault point positioning method and positioning device. The positioning method comprises the steps of calculating a length deviation coefficient of a fault optical cable; measuring optical cable lengths between two ends of the fault optical cable and a fault point; correcting the two optical cable lengths according to the length deviation coefficient, therebyobtaining two pieces of fault distance of the fault optical cable; and determining a distance interval between the two pieces of fault distance as a fault location interval of the fault point. According to the method and the device, the fault point distance measurement is carried out at two ends of the optical cable line at the same time and the correction is carried out through adoption of the length deviation coefficient of the fault optical cable, so the fault point of the optical cable is positioned accurately, the optical able line resource data collection and maintenance difficulty is greatly reduced, the fault point positioning accuracy is greatly improved, and the method and the device have very high practical value.

The invention provides a corpus collecting method and device and computer equipment. The corpus collecting method comprises the following steps: receiving a query request input by a user; showing templated response for the user, wherein the templated response comprises a feedback portion and a guiding portion, the feedback portion comprises result feedback to a queried text input by the user, andthe guiding portion comprises guidance of the user to a next input; and receiving corpuses input by the user. A real dialogue scenario can be shown for a corpus collecting person, and collection difficulty of task-guided multiple corpuses is reduced by balance between efficiency and validity of conversation.

The invention relates to a coal bunker. The coal bunker comprises a cylindrical bunker body, wherein the bunker body is vertically arranged; an upper opening is formed in the top of the bunker body; a lower opening is formed in the bottom of the bunker body; the coal bunker is characterized in that buffering plates are uniformly arranged from top to bottom in an inner cavity of the bunker body; the buffering plates are alternately arranged on one group of opposite inner side walls in the bunker body; the buffering plates are obliquely arranged downwards; one side with high height of each buffering plate is hinged to the corresponding inner side wall of the bunker body; elastic support parts are arranged between lower plate surfaces of the buffering plates and the corresponding inner side wall; a gap is defined by the side with low height of each buffering plate and the opposite inner side wall of the bunker body; other sides of the buffering plates are in sliding fit with the inner wall of the bunker body. Small impact force is generated when coal briquettes fall off in the coal bunker; the coal briquettes are not easy to break and liable to collect after falling off.

The invention discloses textile dust collecting equipment with a convenient recovery function. The textile dust collecting equipment comprises a shell, a horizontal plate is fixedly connected betweenthe two sides of the top of the inner wall of the shell, and a moving device is fixedly connected to the top of the horizontal plate; a vertical rod is fixedly connected to the bottom of the moving device, the bottom end of the vertical rod penetrates through the horizontal plate and extends to the bottom of the horizontal plate, and a moving plate is fixedly connected to the end, located at the bottom of the horizontal plate, of the vertical rod; and a sprayer is fixedly connected with the bottom of the moving plate, and a squeezing device is fixedly connected with the right side of the bottom of the horizontal plate. The invention relates to the relevant technical field of textile equipment. According to the textile dust collecting equipment with the convenient recovery function, the sprayer is driven to reciprocate to spray water to silk and wool, the processing effect is good, the situation that due to uneven water spraying, silk and wool are not convenient to collect and fly to cause pollution is avoided, the size of silk and wool is reduced, the problem that collection is affected by silk and wool expansion is solved, the silk and wool collecting difficulty is reduced, and harm to bodies of workers is reduced.

The invention discloses a generative adversarial network training method and device, an image enhancement method and device, an electronic device and a storage medium. The training method comprises the steps that a first sample set and a second sample set are acquired, the first sample set comprises first image data and corresponding enhanced image data, and the second sample set comprises secondimage data; inputting the first image data into a generative adversarial network, and calculating a first loss error according to the enhanced image data and an output result of the generative adversarial network so as to adjust network parameters of the generative adversarial network; and inputting the second image data into a generative adversarial network, and calculating a second loss error according to an output result of the generative adversarial network so as to adjust network parameters of the generative adversarial network. According to the method, the generative adversarial networkis trained based on semi-supervised deep learning, and the accuracy and robustness of the generative adversarial network are improved while the sample data collection difficulty is reduced.

The invention discloses a model training method, a human body posture detection method and device, equipment and a medium, and the method comprises the steps: inputting a corresponding image trainingsample under the current iteration into a current posture detection network model, obtaining a corresponding first loss function under the current iteration according to the data form of the image training sample, carrying out reprojection processing on a current output result of the attitude detection network model according to the obtained camera parameters, acquiring a corresponding second lossfunction under current iteration based on a reprojection processing result, and performing back propagation on the attitude detection network model based on a fitting loss function formed by the first loss function and the second loss function to obtain an attitude detection network model for the next iteration. By utilizing the method, the collection difficulty of training samples is reduced, the easiness in implementation of network training is ensured, and meanwhile, the whole model training is completed in an image domain, so that the stability and the rapid convergence of the model training are facilitated.

The invention provides a system and a method for controlling the light source of a slit lamp microscope. In image previewing, the light source is adjusted to the brightness comfortable for eyes to carry out eye controlled focus through operating a brightness adjusting knob; and during image shooting, the brightness of the light source is maximal through the control of a CPU (Central Processing Unit). In image shooting, the duration of the maximal brightness of the light source is the same as the camera exposure time. After the images shooting is finished, the CPU controls the brightness of the light source to be restored to be brightness adjusted in image previewing. The brightness required in image previewing and imaging of the light source of the slit lamp can be considered at the same time, so that the definition of the shot images and the checking efficiency are improved effectively.

The invention discloses a smart socket-based electrical appliance type inference method and device, wherein the method comprises the following steps: constructing a user power data set according to electrical appliance power consumption data, and constructing an electrical appliance power consumption characteristic model library according to the electrical appliance power consumption characteristic; constructing an electrical appliance behavior data set according to the user power data set and the electrical appliance power consumption characteristic model library; constructing a characteristic vector extracting model according to the electrical appliance behavior data set so as to construct a power consumption characteristic vector data set; constructing an electrical appliance type classifier according to the power consumption characteristic vector data set and a machine learning algorithm; detecting the power consumption data of a to-be-measured electrical appliance and determiningthe electrical type of the to-be-measured electrical appliance through the model and the classifier. According to the method, the power data of the user is collected through a smart socket; by extracting the characteristics of the range of the power consumption of the household appliances, the duty ratio of the power consumption and the duration of time of power consumption, the power utilizationbehavior of the user is inferred, an electrical appliance behavior event data set is generated, and an electrical appliance type classifier is constructed based on the machine learning model, so thatthe type of the utilized electrical appliance can be inferred in real time; the robustness is high; and the method is simple and easy to implement.

The invention provides an acoustic model optimization method and device, and belongs to the technical field of voice transfer. The method comprises the following steps: acquiring audio data of a preset length time period before the current moment; extracting effective voice data in the audio data, determining the effective voice duration corresponding to the effective voice data and effectively transferring the text; according to the threshold value range of the effective voice duration, determining a parameter updating rule, training the acoustic model through the effective voice data and theeffective transferring text, so as to update the parameters specified in the acoustic model by the parameter updating rule. Due to the fact that a large amount of relevant audio data is collected inadvance, the collection difficulty corresponding to the mode that the audio data is acquired in real time is small, and the workload is small. In addition, the audio data of the speaking user can be acquired in real time in the actual use scene, so that the obtained audio data and the actual use scene are matched with each other, and are matched with the speaking user, and the optimization effectis better when the acoustic model is optimized.