414 results about "Cross field" patented technology

The invention provides a non-contact free space eye-gaze tracking method suitable for man-machine interaction. The method comprises the steps of positioning and tracking a face and eyes in real time, extracting eye movement biological characteristic information, building an eye movement model based on the eye movement biological characteristic information, building a mapping relational model of the eye movement model and an eye-gazed object and the like. The non-contact free space eye-gaze tracking method suitable for the man-machine interaction relates to multiple crossed fields of image processing, computer vision, pattern recognition and the like and has wide application prospect in the fields of new-generation man-machine interaction, disabled people assisting, aerospace relating field, sports, automobile and airplane driving, virtual reality, games and the like. In addition, the method has great practical significance for improvement of the life and self-care level of disabled people, building of a harmonious society and improvement of independent innovative capability in the national high-and-new technical fields of the man-machine interaction, unmanned driving and the like.

The invention belongs to the crossing field of computer vision and intelligent robots, relates to an indoor three-dimensional scene rebuilding method based on a double-layer rectification method and solves the problems that an existing indoor scene rebuilding method is expensive in required equipment, high in computation complexity and poor in real-time performance. The indoor three-dimensional scene rebuilding method based on the double-layer rectification method comprises Kinect calibration, SURF feature point extraction and matching, mapping from a feature point pair to a three-dimensional space point pair, three dimensional space point double-layer rectification based on random sample consensus (RANSAC) and inductively coupled plasma (ICP) methods and scene updating. According to the indoor three-dimensional scene rebuilding method based on the double-layer rectification method, the Kinect is adopted to obtain environmental data, and the double-layer rectification method is provided based on the RANSAC and the ICP. Indoor three-dimensional scene rebuilding which is economical and rapid is achieved, and the real-time performance of rebuilding algorithm and the rebuilding precision are effectively improved. The indoor three-dimensional scene rebuilding method based on the double-layer rectification method is applicable to the service robot field and other computer vision fields which are relative to the three-dimensional scene rebuilding.

The invention discloses a field-oriented personalized intelligent recommendation system and an implementation method. Aiming at the defect of poor universality of the traditional recommendation manner, a reasonable data organization mode is adopted, so that the recommendation service has a cross-field property, and field conversion can be realized by configuring field information according to the characteristics of different fields; aiming at the problem of data sparseness caused by single user information acquisition source and low quantity in the traditional recommendation manner, a context perception technology for actively acquiring context information of interactive behaviors of a user and the system is introduced, so that the quantity of the acquired user information is greatly increased, and the time distribution of the acquired user information is more reasonable; the acquisition of the user information is directly related with the interactive behaviors of the user and the system; the system can acquire the current condition of the user in real time, so that the recommendation service can dynamically reflect the change condition of user demand; and the recommendation service quality is improved through a recommendation learning model by continuously using the reflection of the user on the recommendation result.

The invention relates to a method of constructing a fine discrete road grid in an urban drainage simulation system and belongs to the cross field of municipal engineering information technology, database technology and geologic information system technology. From the demand on construction of a waterlogging simulation model, under the restraint conditions in road design specifications and on the basis of conventional map measured elevation points, the number and positions of interpolation points are optimized under the drainage simulation system platform by the fitting method of local elevation change curves of cross-longitudinal road section and by road characteristic terrain vector layers according to mathematical model interpolation encryption plane and elevation information of physical road tomography; accordingly, a fine discrete road elevation grid unit is constructed, true tomographic road features are expressed quickly and economically, the demand on precise simulation of waterlogging models is met, especially the dredging and resisting actions of road micro-tomographic features upon waterflow can be correctly expressed, and the method is significant to promotion and application of the drainage simulation technology.

The invention belongs to the crossing field of computer vision and intelligent robot, and discloses a new multi-feature multi-sensor method for a mobile robot to track a moving body. The method comprises the following steps: 1, coarsely positioning a body carrying a passive tag around a radio frequency identification (RFID) system by using the RFID system; 2, initially positioning the body in an image by using an adaptive template matching algorithm based on head and shoulder features; 3, accurately positioning the body in the image by using a multi-feature-based mean-shift algorithm; 4, predicting the moving state of the body by using a extended Kalman filter algorithm; 5, screening the acquired target position information by using a double-layer collaboration positioning mechanism; and 6, controlling the robot to move along with the body by using a robot following control algorithm. By the method, bodies with different poses can be tracked, the problem that the tracking is influenced when a target suddenly turns and is shielded is solved, and the robot can accurately, stably and continuously track the moving body.

A wireless energy transfer system includes a primary and one (or more) secondary oscillators for transferring energy therebetween when resonating at the same frequency. The long range (up to and beyond 100 m) efficient (as high as and above 50%) energy transfer is achieved due to minimizing (or eliminating) losses in the system. Superconducting materials are used for all current carrying elements, dielectrics are either avoided altogether, or those are used with a low dissipation factor, and the system is operated at reduced frequencies (below 1 MHz). The oscillators are contoured as a compact flat coil formed from a superconducting wire material. The energy wavelengths exceed the coils diameter by several orders of magnitude. The reduction in radiative losses is enhanced by adding external dielectric-less electrical capacitance to each oscillator coil to reduce the operating frequency. The dielectric strength of the capacitor is increased by applying a magnetic cross-field to the capacitor to impede the electrons motion across an air gap defined between coaxial cylindrical electrodes.

The invention discloses a multiple features fused bidirectional recurrent neural network fine granularity opinion mining method. The method comprises the following steps of: capturing comment data of a specific website through internet and carrying out labelling and preprocessing on the comment data to obtain a training sample set; carrying out training by using a Word2Vec or Glove model algorithm to obtain word vectors of the comment data; carrying out vectorization after carrying out part of speech labeling, dependence relationship labeling and the like; and inputting the vectors into a bidirectional concurrent neural network to construct a bidirectional recurrent neural network fine granularity opinion mining model. According to the method, attribute words in fine granularity opinion mining is extracted and emotional polarity judgement is carried out through the training of a model, so that plenty of model training time is further saved and the training efficiency is improved; no professionals are required to carry out manual extraction on the attribute words, so that a lot of manpower cost is saved; and moreover, the model can be trained by using a plurality of data sources, so that cross-field fine granularity opinion analysis can be completed, thereby solving the problem of long-distance emotional element dependency.

The invention provides a cooperative control method for multi-intersection signal lamp based on Q value migration depth reinforcement learning, and belongs to the crossing field of machine learning and intelligent traffic. A multi-intersection traffic network of an area is modeled into a multi-Agent system firstly. Each Agent simultaneously considers the influence of adjacent Agent actions at themost recent moment in the learning strategy process, so that multiple Agents can cooperatively conduct signal lamp control of multi-intersection. Each Agent adaptively controls one intersection through a deep Q network, and a network input is a discrete traffic state code of the original state information of the corresponding intersection. An optimal action Q value of the adjacent Agent at the most recent moment is transferred to the loss function of the network in the process of learning. The cooperative control method for multi-intersection signal lamp based on the Q value migration depth reinforcement learning can improve the traffic flow of the regional road network and the utilization rate of the road and can reduce the queuing length of the vehicle to relieve the traffic jam, and hasno limitation on the structure of each intersection.

The invention discloses a malicious encrypted flow feature analysis method, which belongs to the crossing field of network security and machine learning, and is used for analyzing the HTTPS flow and detecting the malicious threats in the HTTPS flow. The method comprises the following steps of modeling the flow data by taking a connection quaternion as a data structure, analyzing the encrypted flowfrom four levels by taking the connection quaternion as a unit, and extracting the flow-level features, TLS handshake features, X.509 certificate features and context features of the connection quaternion to obtain an initial feature set, adopting a recursive feature elimination method based on a decision tree to screen the initial feature set to obtain an optimal feature set used for the machinelearning model training to thereby obtain a malicious encrypted flow detection model.

A compact, reliable scanning electron-beam x-ray source achieves reduced complexity and cost. In particular, the x-ray source includes an electron beam that is propagated parallel to an x-ray target and is swept across the target in response to a moving magnetic cross field. Rather than scanning the beam by deflecting it about a single point, the point of deflection is translated along the target length, dramatically reducing the volume of the device. The magnetic cross field is translated along the target length using either mechanical systems to move permanent magnets, or electrical systems to energize an array of electromagnets.

The invention discloses a three-dimensional weaving forming method for a composite material, and belongs to the cross field of weaving and manufacturing. The method comprises the following steps of: performing layered design according to the characteristics of a part structure, and optimizing a weaving path; arranging weaving guide posts on a controllable digital arrangement template at intervals according to certain rules; after a guide sleeve passes through the hollow guide posts along the Z direction and is turned outwards, fixing the guide sleeve on the controllable digital template; selecting a silk material for weaving the layer; after the weaving of one layer is finished, lowering the controllable digital arrangement template, and driving the guide sleeve to expose for a distance to form a new layer of arrangement template; weaving layer by layer till the weaving of the whole part is finished; detaching and sewing the part; and soaking the part into resin to form a composite material part. According to the method, the quick prototyping technology and the weaving technology are organically combined to realize three-dimensional weaving of the complex functional part; and meanwhile, composite material preparation and part formation are integrated, so that manufacture of the complex functional part is realized.

The invention discloses a contour extraction method based on the brightness characteristic and the contour integrity, and belongs to the crossing field of computer vision and pattern recognition. The contour extraction method aims to completely extract a contour of an object out of a complex environmental background. The contour extraction method includes the step of obtaining a maximum energy response diagram, the step of carrying out non-classic receptive field restraint on brightness characteristic modulation, the step of extracting the contour of the object and carrying out postprocessing on a high-low self-adaptive threshold value and the like based on a probability model, and the step of processing the nearest neighbor orientation consistency connection rupture contour based on the contour integrity. A Gabor filter is used for simulating response of a human simple cell classic receptive field to obtain the maximum Gabor energy response diagram; the brightness characteristic of images is used for restraining the maximum Gabor energy response diagram and rejecting grain and other non-target contours; the postprocessing on the high-low self-adaptive threshold value and the like based on the probability model is carried out on the obtained target contour; rupture points of the nearest neighbor orientation consistency connection contour based on the contour integrity are processed. The contour of the object can be extracted well, and a contour diagram of integrity connection is obtained after the contour of the object is processed.

The invention relates to application of eurotium cristatum in the production of Liupu tea, which belongs to the crossing field of food engineering and biotechnology. The production method for the Liupu tea is an improvement on the prior Liupu tea production flow through biotechnology, namely after primary distillation, a bacterial suspension containing eurotium cristatum spores is used for spray inoculation, and after redistillation, tea strains are crushed into pieces for inoculation again. Thus, when the Liupu tea is in composting fermentation, the eurotium cristatum can quickly become predominant bacteria to inhibit the growth of other bacteria, and when the Liupu tea is in ageing, golden flowers grow out frondently and universally so as to improve the quality of the Liupu tea. The application of the eurotium cristatum improves the quality of the Liupu tea, and solves the problems that in the prior art, the quality is unstable, the ageing time is long, the growing golden flowers aredifficult to control and are easy to pollute and the like. Simultaneously, the invention discloses a method for simply and easily obtaining and expanding the cultivation of eurotium cristatum strains.

The invention discloses a method for fast constructing a high-accuracy personalized face model on the basis of facial images, which belongs to the cross field of computer vision and computer graphics, and aims at achieving the purposes of fast reconstructing a personalized 3D geometrical facial model and a high-accuracy 2D facial texture mapping on the basis of real facial images. The method comprises the following steps: firstly, shooting five facial images from multiple angles; secondly, marking positions and corresponding relationships of forty-one characteristic points on the facial images and a neutral facial model; thirdly, aligning the projection of the neutral facial model on a virtual camera focal plane with the facial images; fourthly, carrying out hierarchical geometrical deformation on the neutral facial model to obtain the personalized geometrical facial model; and finally, synthesizing the high-accuracy facial texture mapping according to the five facial images. The face modeling method has the advantages of convenience for data acquisition, high modeling speed and modeling accuracy, and the like, and achieves higher practical values in the fields of video communication, electronic games, man-machine interaction, and the like.

A glassy metal alloy consists essentially of the formula FeaCobNicMdBeSifCg, where "M" is at least one member selected from the group consisting of molybdenum, chromium and manganese, "a-g" are in atom percent, "a" ranges from about 30 to about 45, "b" ranges from about 8 to about 18, "c" ranges from about 20 to about 45, "d" ranges from about 0 to about 3, "e" ranges from about 12 to about 20, "f" ranges from about 0 to about 5 and "g" ranges from about 0 to about 2. The alloy can be cast by rapid solidification into ribbon, cross-field annealed to enhance magnetic properties, and formed into a marker that is especially suited for use in magneto-mechanically actuated article surveillance systems. Advantageously, the marker is characterized by substantially linear magnetization response in the frequency regime wherein harmonic marker systems operate magnetically. Voltage amplitudes detected for the marker are high, and interference between surveillance systems based on mechanical resonance and harmonic re-radiance is virtually eliminated.

The invention discloses a contour and boundary detection algorithm based on a visual color theory and homogeneity inhibition, belongs to the cross field of computer vision and pattern recognition, and aims at extracting the contour and boundary of a target from a complex natural scene. Through the research of the visual information processing mechanism of human eyes, the method carries out the mathematic modeling of a receptive field of a nerve cell of a visual pathway at each stage, and employs the modulation effect of a non-classical receptive field to inhibit the texture edge, thereby highlighting the contour and boundary. The innovation of the invention lies in introducing the visual information processing mechanism of human eyes to a contour and boundary detection model, detecting the color and brightness boundary through setting imbalance vision cone input, maintaining the contour integrity, giving consideration to the homogeneity of the texture region and the homogeneity inhibition on a classical receptive field from the non-classical receptive field, employing the homogeneity inhibition to inhibit the texture edge, and extracting the contour and boundary of a natural image.

A magnetostrictive element for use in a magneto-mechanical marker has a resonant frequency characteristic that is at a minimum at a bias field level corresponding to the operating point of the magnetomechanical marker. The magnetostrictive element has a magnetomechanical coupling factor k in the range 0.28 to 0.4 at the operating point. The magnetostrictive element is formed by applying cross-field annealing to an iron-rich amorphous metal alloy ribbon (45 to 82 percent iron) which includes a total of from 2 to 17 percent of one or more of Mn, Mo, Nb, Cr, Hf, Zr, Ta, V. Cobalt, nickel, boron, silicon and / or carbon may also be included. The metal alloy may include one early transition element selected from the group consisting of Zr, Hf and Ta, and also a second early transition element selected from the group consisting of Mn, Mo, Nb, Cr, and V.

The invention discloses a method for establishing a large-scale cross-field text emotion orientation analysis framework. The method comprises the steps that precise word segmentation is carried out on sample documents in a source field and sample documents in a target field to form two word vector tables; word vectors are clustered, and the fields are aligned; primary sentence modeling is carried out on calibration samples in the source field through the word vectors and serves as input of DCELM, and interlayer abstraction features of text vectors are extracted through convolution operation; convolution layer parameters obtained when the classification effect of a verification set is best are recorded and serve as parameters of a DCELM network convolution layer; finally, interlayer abstraction features of calibration samples, extracted through DCNN, of a small number of target fields are used for training implicit layer parameters of a classifier ELM to establish the large-scale cross-field text emotion orientation analysis framework. By means of the technical scheme, the difference of words for expressing emotion polarities among the fields is eliminated on the sample layer, the defects that on a full-connection layer, local optimization is easily caused and the generalization ability is weak are effectively solved, and the anti-interference performance of a model is improved.

The invention belongs to the crossing field of artificial intelligence and data mining, and particularly relates to a text sentiment analysis method based on a bidirectional long-short term memory neural network. The method comprises the following steps: in a text mapping word vector mapping framework, expressing a text as a word vector matrix; Constructing an improved convolutional neural network, and performing feature extraction on the mapped word vector matrix; Training the extracted features by using a bidirectional long-short term memory recurrent neural network until the loss function is minimum; And adding a global mean pooling layer and a classification layer behind the recurrent neural network so as to output sentiment classification of the text. According to the method, deep learning is applied to text sentiment analysis, the limitation of manual extraction of text sentiment characteristics is avoided, the advantages of a convolutional neural network and a recurrent neural network are combined in the field of deep learning, and the network is improved, so that the method can be better applied to a text processing task.

The present invention discloses a method for obtaining path, a path computing unit and a system thereof. The invention is used for technical field of communication. The method provided by the invention comprises a step that at least two paths calculating fields between a path calculating initializing field and a path calculating terminating field execute interactive calculation for obtaining the path between two end nodes, wherein, at least one path calculating field obtains the path information of the first path. The path information which is used for calculating the second path is calculated, generated and transmitted to the next path calculating path according to the path information of the obtained first path. The path information which is used for calculating the second path comprises the path information of path between two boundary nodes. The two boundary nodes are related with the path calculating terminating field and furthermore one boundary node is passed by the first path. The method, the adopted device and the adopted system can increase the calculation success rate of cross-field separation path at the state that each PCE has no full-mesh topology.

The invention discloses a UAV autonomous landing control system and method for mobile platform, belongs to the crossing field of an aerial robots and computer vision, and relates to the autonomous landing technology of the UAVs. The system comprises the steps: a vision module disposed on a UAV, a landing target detection module, a UAV posture estimation module and a parameter comparison module. The method comprises the steps that when the UAV lands, the vision module obtains a real-time image, and carries out the threshold segmentation of the grayed image; a rotating matrix and a translation matrix of the UAV relative to a landing target are solved according to an image processing result, and the posture angle of the UAV is obtained through the correlation between the rotating matrix and an Euler angle; the posture information of the UAV is estimated in real time, and the flight of the UAV is judged and adjusted. The system combines the UAV technology and an image processing algorithm,achieves the precise landing of the UAV on the mobile platform, is higher in automation degree, is simpler in operation, improves the safety of the UAV, and reduces the accident rate of the UAV.

The invention discloses an automatic SWMM hydraulic model building method based on a rainwater pipe network design computation sheet, and belongs to the crossing field of municipal engineering information technologies and geographic information system technologies. According to the automatic SWMM hydraulic model building method based on the rainwater pipe network design computation sheet, information of an existing rainwater pipe network design computation sheet can be used for synchronously achieving effective automatic GIS conversion of graphic data and attribute data of a pipe network model, and convenience is provided for building of an SWMM hydraulic model. In addition, workloads of conducting tedious hierarchical processing on CAD graphic data according to elements and conducting manual entry on the attribute data in a traditional hydraulic model building method are greatly simplified, the defects that traditional manual entry of the attribute data is tedious in work, low in efficiency and prone to error are avoided, only simple processing needs to be conducted on the rainwater pipe network design computation sheet, the work efficiency is greatly improved, the GIS quality of model data is guaranteed, and then the quality of the SWMM hydraulic model can be guaranteed. According to the method, the operability is achieved, all steps can be completed only by means of simple operations of a GIS platform, and the implementing criterion is guaranteed.

The invention relates to the cross field of geographic information and information technology, in particular to a distributed algorithm for quickly establishing a massive remote sensing image pyramid applied to the construction of National Spatial Data Infrastructure. The method realizes the algorithm for quickly establishing the massive remote sensing image pyramid by means of data distribution storage and parallel computation capabilities based on a PC (Personal Computer) cluster, through the steps of dividing N*M grids, performing pyramid establishment task distribution and scheduling and pyramid establishment and correctly dividing grids and computing grid information, and by means of a communication transfer task between a control node and a computation node; and the method mainly aims to solve the related technical problem of how to divide the N*M grids, etc. The distributed algorithm for quickly establishing the massive remote sensing image pyramid has the following advantages that: based on the data distribution storage and parallel computation capabilities, the speed of generating the massive remote sensing image pyramid is increased; and the problem of black edges causedby image block documents crossing grids is solved; therefore, the purpose of increasing the pyramid establishing speed is achieved and the pyramid establishment accuracy is ensured.

The invention relates to a 3D printing path planning method of a continuous fiber reinforced composite material, and belongs to the cross field of composite materials and additive manufacturing. According to the 3D printing path planning method of the continuous fiber reinforced composite material, finite element simulation technology is utilized to simulate and analyze the stress distribution ofan element under the action of a load, and a printing path of continuous fiber reinforced composite material additive manufacturing is planned according to the stress distribution direction and transmission characteristics of the element and the characteristics of fiber continuity. Compared with a traditional path planning method, the orientation of the continuous fibers can be adjusted pertinently, the bearing capacity of the element is improved to the maximum extent, the usage amount of fiber materials is reduced, the manufacturing cost of the continuous fiber reinforced composite material is reduced, and the high performance, high efficiency, high precision and low cost 3D printing forming of the continuous fiber reinforced composite material are achieved.

The invention, which belongs to the crossing field of artificial intelligence and intelligent transportation, provides a multi-intersection cooperative traffic light control method based on Q-value migration of a multi-task deep Q network. Multiple intersections are modeled into a multi-agent system; a multi-task network that is suitable for controlling of each intersection traffic light is trained at multiple intersections; knowledge learned at the multiple intersections is applied to a target problem, so that the multi-task network have much more knowledge than the single network and the single intersection feature extraction capacity is enhanced; and then on the basis of a cooperative algorithm, traffic lights of all intersections in a target area are controlled cooperatively. Therefore, the traffic flows of all intersections can be balanced to a certain degree; the road utilization rate in the regional traffic can be increased; the vehicle queuing length can be reduced; and the traffic congestion can be alleviated. The method has the high scalability to the traffic network.

The invention discloses an abnormal traffic detection method based on an autoencoder network and belongs to the crossing field of a computer network and machine learning. The method comprises the steps of capturing data through utilization of a traffic data obtaining module; preprocessing the data through utilization of a data preprocessing module; outputting hash codes of the traffic data throughutilization of an improved multi-autoencoder network; and carrying out abnormal detection multi-classification on the obtained hash codes through adoption of a detection module based on a data distribution proportion. According to the method, advantages of a neural network and the hash codes are combined; through utilization of a hash encoding technology, consumed memory space and computing resources are reduced; a KitNET autoencoder network which only can be applied to a binary-classification intrusion detection scene is improved into a multi-classification intrusion detection method; and atraffic hash code classification method based on the data distribution proportion is designed. According to the method, an operation rate and expansibility of an abnormal detection technology can be improved. The method is relatively applicable to solution of the abnormal detection problem that new attack types emerge in endlessly at present.

The invention discloses a VNMS for building dynamic application systems. The VNMS mainly comprises a frame and business units and is characterized in that a distributed system formed by the VNMS or formed by the VNMS and existing application systems on a computer network is displayed to be the application systems which conform to enterprise management modes and habitual business processes and habitual operation of users when mapped to the user level, a frame-type building mode, a BMOA development method and the technologies such as the SOA and the MVC are adopted, development and implementation of a parallel development mode are unified, and an integrated development platform, a cross management mode and the like are adopted, so that the built application systems are simple and clear in structure, the VNMS has the advantage that a huge and complex system is resolved into the simple and small-scale systems, the complexity of business modeling and process orchestration is lowered, the reusability is improved, it is shown that described things are obtained things, dynamic changes of the application systems are met, cross-field, cross-content and cross-platform information resource sharing is achieved, and the VNMS is suitable for developing large-scale software systems and also suitable for developing lightweight enterprise applications.

The invention belongs to the crossing field of space technology and computer vision, and discloses a monocular space target distance-measuring and angle-measuring method. The method comprises the following steps: (1) receiving a one-frame image of a space target shot by a single camera, namely a practically shot image Q; (2) pre-treating the practically shot image Q, extracting the space target from the background, thus obtaining a pretreated image S; and then detecting the target according to the pretreated image S, if the space target is completely located in the view field of the camera, entering the step (3), otherwise, entering the step (5); (3) extracting the feature of the space target and recognizing the posture of the space target; (4) computing the three-dimensional center distance Delta p and the posture angle of the space target; and (5) continuing processing until all images are completely processed. The method provided by the invention is simple in process; according to the method provided by the invention, only the three-dimensional structure and size information of the target are needed to be obtained, the target is not needed to be the cooperative target, and any star target is neither needed to be set on the target; and the method provided by the invention has the characteristics of wide distance-measuring range and high precision.

The invention belongs to the crossing field of a computer application technology and production manufacturing. A natural computing technology is used to optimize a multi-target flexible job shop scheduling problem. A problem that a cooperative hybrid artificial fish swarm algorithm is used to solve multi-target flexible job shop scheduling is provided. The method is characterized in that a foraging behavior with a distribution estimation attribute and an artificial fish attraction behavior are designed to improve an artificial fish swarm model; a cooperation idea is introduced into the model; through multiple population cooperation of the fish swarm, global searching is performed and is cooperated with a simulation annealing algorithm so as to enhance an algorithm local searching capability; aiming at a multi-target problem, an improved epsilon-Pareto dominant strategy is designed to evaluate an individual applicable degree value. The method in the invention has the following advantages that problems of slow later-period convergence, a poor local optimizing ability and the like, which exist in the artificial fish swarm algorithm during a searching process, can be overcome; through cooperative optimization, a pareto solution set with good quality and dispersibility is obtained.

The embodiment of the invention provides a voice control method and device. The method comprises the following steps that: receiving voice information input by a user; converting the voice informationinto a text; according to the text, judging whether the intention of the user is a combined task or not; if the intention of the user is the combined task, mapping the combined task to a plurality ofsub-tasks; extracting the element information of each sub-task; according to each piece of element information and a dialogue context, determining a maximum dialogue state under a current probability; according to the dialogue state, adopting a hierarchical reinforcement learning method to decide a current optimal dialogue behavior; according to the dialogue behavior, generating a corresponding text response; and converting the text response into a text, and outputting the text. By use of the embodiment of the invention, data generated in a human-computer interaction process can be effectively utilized, cross-field complex voice input is identified and comprehended, a search space is reduced, and cross-field complex tasks can be more flexibly processed.