CN102348101A

CN102348101A - Examination room intelligence monitoring system and method thereof

Info

Publication number: CN102348101A
Application number: CN2010102412794A
Authority: CN
Inventors: 丁宁
Original assignee: SMARTECH AVANCED RESEARCH SHENZHEN
Current assignee: SMARTECH AVANCED RESEARCH SHENZHEN
Priority date: 2010-07-30
Filing date: 2010-07-30
Publication date: 2012-02-08

Abstract

The invention relates to an examination room intelligence monitoring system and a method thereof. The system comprises: a signal input system, a signal pretreatment system, an examination room behavior single-mode state identification system, an examination room behavior multi-mode state decision and identification system and an examination room monitoring management background system. The signal input system sends a collected sound/video signal to the signal pretreatment system so as to perform pretreatment. After the pretreatment, the signal is sent to the examination room behavior single-mode state identification system and the examination room behavior multi-mode state decision and identification system so as to be performed classification identification and determination. Then a generated identification result is sent to the examination room monitoring management background system so as to realize a purpose of monitoring examination room behaviors. By using the examination room intelligence monitoring system and the method of the invention, an artificial intelligence technology can be used to process video information, audio information and sensor multi-mode state information simultaneously. And then suspected violation behaviors in the examination room can be identified so that an alarm can be given.

Description

A kind of examination hall intelligent monitor system and method

Technical field

The present invention relates to a kind of supervisory control system and method, more particularly, relate to a kind of examination hall intelligent monitor system and method.

Background technology

Along with society is more and more higher to the fair requirement of taking an examination, examinee's behavior identification and monitoring technique obtain increasing attention in the examination process.Patrolling test system at present on the net and carry out in a large number, in application process, how the ten hundreds of examination point in the whole nation, nearly ten million examinee's examination behavior is monitored in real time, is a problem demanding prompt solution.Patrol test system at present on the net and be a kind of event history system, the monitoring of can not realize taking an examination process exception incident and the doubtful unlawful practice of examinee.Present system transfers back to Surveillance center with the video information of multiple cameras.The examination points of patrolling at different levels can be carried out the tour and the inspection of examination point video through network path, but the scope that the people observes is very limited.Therefore, how to study that the doubtful unlawful practice in examination hall is handled, discerned to the using artificial intellectual technology to monitor message and then warning becomes more and more important.

Summary of the invention

The technical problem that the present invention will solve is; To the above-mentioned of prior art the monitoring in real time defective of prompting automatically in examination hall can not be provided, provide a kind of using artificial intellectual technology monitor message to be handled, discerned the examination hall intelligent monitor system and the method for doubtful unlawful practice in examination hall and then warning.

The present invention solve the technical problem and can realize through adopting following technical scheme: construct a kind of examination hall intelligent monitor system, wherein, comprising: signal input system: be used for gathering and sending pilot signal; Signal pre-processing system: be used for the said pilot signal of preliminary treatment and feature extraction and send preliminary treatment and feature extraction after pilot signal; Examination hall behavior single mode recognition system: be used for pretreated pilot signal and characteristic thereof are carried out single mode identification and doubtful unlawful practice classification, and produce recognition result; Examination hall monitoring management background system: be used to receive said recognition result and deal with.

In the intelligent monitor system of examination hall of the present invention, said pilot signal comprises the pilot signal that video monitoring signal, Voice Surveillance signal and/or transducer receive; Said signal input system comprises: video signal collection apparatus: be used for collection, digitlization, compressed encoding and send video monitoring signal; And/or audio signal sample equipment: be used for collection, digitlization, compressed encoding and send the Voice Surveillance signal; And/or sensor acquisition equipment: be used to gather and send the pilot signal that corresponding transducer receives; Said signal pre-processing system comprises: with said video signal collection apparatus corresponding video signals pretreatment unit, regular event filter element, video features extraction unit; And/or with the corresponding audio signal pretreatment unit of said audio signal sample equipment, abnormal sound positioning unit, audio feature extraction unit; Said vision signal pretreatment unit: be used for that said video monitoring signal is carried out signal condition and judge and preliminary treatment; Said regular event filter element: be used for normal action in the examination hall is filtered; Said video features extraction unit: be used for that the video monitoring signal that said vision signal pretreatment unit was handled is carried out feature extraction and handle; Said audio signal pretreatment unit: be used for that said Voice Surveillance signal is carried out signal condition and judge and preliminary treatment; Said abnormal sound positioning unit: be used to estimate abnormal sound orientation in the examination hall; Said audio feature extraction unit: be used for that the Voice Surveillance signal that said audio signal pretreatment unit was handled is carried out feature extraction and handle; Voice Surveillance signal and/or video monitoring signal adopt the Streaming Media mode to send to examination hall monitoring management background system through the Streaming Media interface.

In the intelligent monitor system of examination hall of the present invention, said examination hall behavior single mode recognition system comprises: said video anomaly analysis unit: be used for carrying out the abnormal behaviour classification through the video monitoring signal that the abnormal behaviour model library was handled said video features extraction unit; Or said audio frequency anomaly analysis unit: be used for carrying out the abnormal sound classification through the Voice Surveillance signal that the abnormal sound model library is crossed said audio feature extraction cell processing; Said abnormal behaviour model library and said abnormal sound model library are learnt according to the result of said examination hall monitoring management background system and are upgraded the model of cognition in said abnormal behaviour model library and the said abnormal sound model library.

In the intelligent monitor system of examination hall of the present invention, said examination hall monitoring management background system comprises: message prompting and administrative unit, Streaming Media demonstration and playback unit, behavior recognition unit are provided with and upgrading unit, the configuration of examination hall Behavior Monitor System and administrative unit; Said message prompting and administrative unit: be used to receive the recognition result that said examination hall behavior single mode recognition system is sent with message, and adopt database technology to manage; Said Streaming Media shows and playback unit: be used for Voice Surveillance signal and/or video monitoring signal that the receiving stream media interface is seen off, and show, playback and management; Said behavior recognition unit is provided with and the upgrading unit: be used for the recognizer parameter of abnormal behaviour model library and abnormal sound model library and the parameter interface of other each unit modules are carried out long-range setting, signal input system, signal pre-processing system and examination hall behavior single mode recognition system are carried out whole software upgrading; Behavior Monitor System configuration of said examination hall and administrative unit: be used for said examination hall behavior intelligent monitor system software and hardware is configured and manages, and through connecting examination hall invigilator's long-range transmission instruction of teacher's communication equipment or system identification information.

In the intelligent monitor system of examination hall of the present invention; Comprise that also multi-modal decision-making of examination hall behavior and recognition system comprise: based on the multi-modal integrated unit of decision-making level; Multi-modal recognition rule of examination hall behavior and model library and the multi-modal recognition unit of examination hall behavior; Behavior multi-modal recognition unit in said examination hall is according to the multi-modal integration unit integrates video anomaly analysis unit based on decision-making level; The signal that video anomaly analysis unit and sensor acquisition equipment obtain; Carry out analysis-by-synthesis according to multi-modal recognition rule of examination hall behavior and model library, and will send comprehensive recognition result and send to examination hall monitoring management background system.

In the intelligent monitor system of examination hall of the present invention, multi-modal recognition rule of said examination hall behavior and model library are learnt according to the result of said examination hall monitoring management background system and are upgraded model of cognition and the rule in multi-modal recognition rule of said examination hall behavior and the model library.

In the intelligent monitor system of examination hall of the present invention, said examination hall monitoring management background system comprises: message prompting and administrative unit, Streaming Media demonstration and playback unit, behavior recognition unit are provided with and upgrading unit, the configuration of examination hall Behavior Monitor System and administrative unit; Prompting of said message and administrative unit: be used to receive the recognition result that the multi-modal decision-making of examination hall behavior and recognition system are sent with message, and the employing database technology is managed; Said Streaming Media shows and playback unit: be used for Voice Surveillance signal and/or video monitoring signal that the receiving stream media interface is seen off, and show, playback and management; Said behavior recognition unit is provided with and the upgrading unit: be used for the recognizer parameter of abnormal behaviour model library, abnormal sound model library and multi-modal recognition rule of examination hall behavior and model library and the parameter interface of other each unit modules are carried out long-range setting, whole software upgrading is carried out in signal input system, signal pre-processing system, examination hall behavior single mode recognition system and the multi-modal decision-making of examination hall behavior and recognition system; Behavior Monitor System configuration of said examination hall and administrative unit: be used for examination hall behavior intelligent monitor system software and hardware is configured and manages, and through connecting examination hall invigilator's long-range transmission instruction of teacher's communication equipment or system identification information.

Construct a kind of examination hall intelligent control method; Based on a kind of examination hall intelligent monitor system; Said examination hall intelligent monitor system comprises signal input system, signal pre-processing system, examination hall behavior single mode recognition system and examination hall monitoring management background system; Wherein, comprising: S1, signal input system collection are also sent pilot signal; S2, signal pre-processing system preliminary treatment and the said pilot signal of feature extraction and send preliminary treatment and feature extraction after pilot signal; S3, examination hall behavior single mode recognition system are carried out single mode identification and doubtful unlawful practice classification to pretreated pilot signal and characteristic thereof, and are produced recognition result; S4, examination hall monitoring management background system receive said recognition result and deal with.

In the intelligent control method of examination hall of the present invention, said pilot signal is the pilot signal that video monitoring signal, Voice Surveillance signal and/or transducer receive; Said step S1 comprises: video signal collection apparatus collection, digitlization, compressed encoding also send video monitoring signal; The collection of audio signal sample equipment, digitlization, compressed encoding also send the Voice Surveillance signal, and the collection of sensor acquisition equipment is also sent the pilot signal that corresponding transducer receives; Said step S2 comprises: the vision signal pretreatment unit carries out signal condition to said video monitoring signal and judges and preliminary treatment; The regular event filter element filters normal action in the examination hall, and the video monitoring signal that the video features extraction unit was handled said vision signal pretreatment unit is carried out feature extraction and handled; The audio signal pretreatment unit carries out signal condition to said Voice Surveillance signal and judges and preliminary treatment; The abnormal sound positioning unit is estimated abnormal sound orientation in the examination hall, and the Voice Surveillance signal that the audio feature extraction unit was handled said audio signal pretreatment unit carries out feature extraction to be handled.

In the intelligent control method of examination hall of the present invention; Step S3 comprises: a, the various characteristic quantities that said vision signal feature extraction unit or said audio signal characteristic extraction unit (26) are extracted are selected, weighting and dimension-reduction treatment, the constitutive characteristic vector; B, video anomaly analysis unit select model of cognition corresponding in the abnormal behaviour model library to carry out real-time grading according to said characteristic vector, and audio frequency anomaly analysis unit selects model of cognition corresponding in the abnormal sound model library to carry out real-time grading according to said characteristic vector; What store in said abnormal behaviour model library or the said abnormal sound model library is a series of said model of cognition to behavior of various examination halls and sound that trained, the grader of said model of cognition for adopting machine learning techniques to make up.

In the intelligent control method of examination hall of the present invention; Also comprise: the recognition result that multi-modal decision-making of said examination hall behavior and recognition system receive said examination hall behavior single mode recognition system carries out multi-modal analysis-by-synthesis and sends to said examination hall monitoring management background system processed steps; Said multi-modal analysis-by-synthesis comprises: a, carry out fusion treatment through algorithm, the fusion treatment algorithm comprises: wavelet transformation, weighted average, production rule and Kalman filtering; B, the signal after a step process is carried out the characteristic fusion treatment through algorithm; The feature description Vector Fusion of a plurality of low dimensions is formed the more associating characteristic vector parameter of higher-dimension, and characteristic fusion treatment algorithm comprises: parameter template, clustering methodology, fuzzy set theory, possibility theory, adaptive neural network, physical model, blackboard model and logic template method; C, utilize the model in multi-modal recognition rule of examination hall behavior and the model library; Signal to handling through step b is discerned judgment processing through algorithm, and identification judgment processing algorithm comprises: classical reasoning, Bayesian inference, statistical decision, DS evidence theory, fuzzy theory, neural net, expert system, rough set theory, broad sense evidential reasoning theory, heuristic method, possibility theory and default reasoning.

In the intelligent control method of examination hall of the present invention, also comprise the step of learning and upgrade the information in multi-modal recognition rule and the model library according to the result of said examination hall monitoring management background system.

Adopt examination hall of the present invention intelligent monitor system and method; Has following beneficial effect: adopt signal pre-processing system of the present invention and examination hall behavior recognition system can realize that the using artificial intellectual technology handles monitor message, identification doubtful unlawful practice in examination hall and then warning.

The collecting device of multiple pilot signal and on-the-spot intervention signal generation equipment make that the signal of gathering is more comprehensive, avoid the generation of abnormal signal in the examination hall simultaneously; Signal Pretreatment unit, signal characteristic extraction unit and the compression of abnormal signal analytic unit have reduced the data volume of monitoring, and presort to extracting characteristic; The regular event filter element is realized supervisor's in the examination hall the activity of examining of normally patrolling is filtered, with the interference to examinee's activity recognition of the normal activity of avoiding the supervisor; The abnormal sound positioning unit realizes that this function realizes through microphone array is installed at the scene to the estimation in abnormal sound orientation takes place in the examination hall; Thereby having according to the result of examination hall monitoring management background system, the signal model storehouse of a while behavior single mode recognition system learns to upgrade model library.The setting of examination hall monitoring management background system is invigilated the personnel inspection monitoring effect easily and is in time made correction, setting.Thereby when judging the doubtful unlawful practice in identification examination hall through the pilot signal characteristic, multi-modal decision package and database can also learn to upgrade the judgement criterion of identification according to the result of examination hall monitoring management background system.

Description of drawings

To combine accompanying drawing and embodiment that the present invention is described further below, in the accompanying drawing:

Fig. 1 is the structural representation of the preferred embodiment of examination hall of the present invention intelligent monitor system;

Fig. 2 is the application drawing of the preferred embodiment of examination hall of the present invention intelligent monitor system.

Embodiment

Do further to detail below in conjunction with the most preferred embodiment shown in the accompanying drawing:

The present invention has made up the examination hall intelligent monitor system of a cover to the doubtful unlawful practice identification in examination hall; This system has adopted the compound monitor mode of video monitoring, Voice Surveillance and the monitoring of general purpose transducer communications platform; Predefined interested monitored object behavior is monitored; According to rule-based knowledge base; Automatically examinee's behavior in the identification examination hall, and make real time record and warning.

The mankind have the ability that combines priori to carry out integrated treatment the resulting information of the various sense organs of human body (eye, ear, nose, four limbs etc.) (comprising image, sound, smell, sense of touch etc.), just can make an estimate and judge surrounding environment and occurent thing.This processing procedure is complicated, also is adaptive, and it is converted into the valuable explanation to environment with various information.The core of the examination hall intelligent monitor system that the present invention relates to is exactly to make full use of the information that a plurality of information sources provide; Expectation is as the process of the various information of human brain integrated treatment; Through reasonable use to these information; Redundancy or complementary information are made up according to certain rule, to obtain consistency explanation or description monitored object and incident thereof.

As the preferred embodiment of a kind of examination hall of the present invention intelligent monitor system, as shown in Figure 1, comprise signal input system 1, signal pre-processing system 2, examination hall behavior single mode recognition system 3 and examination hall monitoring management background system 5; Signal input system 1 is sent to signal pre-processing system 2 with the pilot signal of gathering and carries out preliminary treatment; Then pretreated signal is sent to examination hall behavior single mode recognition system 3; Discern the judgement classification by examination hall behavior single mode recognition system 3; And generation recognition result; Triggering is installed in the driving mechanism of the on-the-spot equipment in examination hall or other mode passages and works in coordination with interlock and in time intervene; Thereby the relevant information that is produced is sent to examination hall monitoring management background system 5, realize the purpose of examination hall behavior monitoring.This system realizes simultaneously the examination process in a plurality of examination halls being monitored in real time; Through real-time pilot signal is carried out the real-time intelligent analysis; Realization is discerned the doubtful unlawful practice in the examination process automatically real-time, and relevant information that will this doubtful unlawful practice sends to examination hall monitoring management background system 5.

As the preferred embodiment of a kind of examination hall of the present invention intelligent monitor system, as shown in Figure 1, said signal input system 1 comprises video signal collection apparatus 11, audio signal sample equipment 12 and general purpose transducer collecting device 13.Video signal collection apparatus 11 is used for collection, digitlization, compressed encoding and sends video monitoring signal; Audio signal sample equipment 12 is used for collection, digitlization, compressed encoding and sends the Voice Surveillance signal, and general purpose transducer collecting device 13 is used to gather and send the pilot signal that corresponding transducer receives.Use this examination hall intelligent monitor system, carry out Real time identification, prompting and assist process to the doubtful unlawful practice that examinee's examination process in the environment of examination hall takes place.This system can adopt vision signal, audio signal or sensor signal to carry out the take an examination monitoring of behavior of examinee separately; Also comprehensively the result of audio frequency and video and the analysis of multiple sensors signal intelligent carries out multi-modal fusion and judgement, discerns more accurately with examinee's examination behavior.

Said signal pre-processing system 2 comprises vision signal pretreatment unit 21, regular event filter element 22, video features extraction unit 23, audio signal pretreatment unit 24, abnormal sound positioning unit 25 and audio feature extraction unit 26.21 pairs of vision signals that send from video signal collection apparatus 11 of said vision signal converting unit are carried out signal condition judgement and preliminary treatment, are sent to video features extraction unit 23 then and carry out vision signal feature extraction processing; 24 pairs of audio signals that send from audio signal sample equipment 12 of audio signal converting unit are carried out signal condition judgement and preliminary treatment, are sent to audio feature extraction unit 26 then and carry out audio signal characteristic extraction processing; The signal condition arbitration functions of audio-video signal is used for identification signal and whether loses, whether receives interference, is in normal condition to guarantee signal.The Signal Pretreatment major function of audio-video signal is for improving signal to noise ratio, so that the effect of feature extraction unit.Regular event filter element 22 is realized supervisor's in the examination hall the activity of examining of normally patrolling is filtered, with the interference to examinee's activity recognition of the normal activity of avoiding the supervisor.Abnormal sound positioning unit 25 realizes that this function need be installed the pick-up array at the scene to the estimation in abnormal sound orientation takes place in the examination hall.Voice Surveillance signal and/or video monitoring signal adopt the Streaming Media mode to send to examination hall monitoring management background system 5 through Streaming Media interface 20, make things convenient for record data and artificial real-time monitoring.

Preferred embodiment as a kind of examination hall of the present invention intelligent monitor system; As shown in Figure 1, said examination hall behavior single mode recognition system 3 comprises abnormal behaviour model library 31, video anomaly analysis unit 32, abnormal sound model library 33 and audio frequency anomaly analysis unit 34.The video monitoring signal that video anomaly analysis unit 32 was handled through 31 pairs of video features extraction units 23 of abnormal behaviour model library is carried out the abnormal behaviour classification; The Voice Surveillance signal that audio frequency anomaly analysis unit 34 was handled through 33 pairs of audio feature extraction unit 26 of abnormal sound model library carries out the abnormal sound classification; Abnormal behaviour model library 31 learn according to the result of examination hall monitoring management background system 5 with abnormal sound model library 33 and update anomalies behavior model storehouse 31 and abnormal sound model library 33 in model of cognition.Examination hall behavior single mode recognition system 3 embedded intelligent behavior model of cognition can carry out phonetic analysis, motion analysis and doubtful unlawful practice classification to what pretreated audio/video signal and characteristic thereof were carried out examinee and supervisor.During 3 work of examination hall behavior single mode recognition system, vision signal pretreatment unit 21, audio signal pretreatment unit 24, video features extraction unit 23 and audio feature extraction unit 26 are computer or digital signal processing appts.Video signal collection apparatus 11, vision signal pretreatment unit 21, regular event filter element 22 and video features extraction unit 23 are formed the video intelligent analyzing subsystem; The audio intelligent analyzing subsystem is formed in audio signal sample equipment 12, audio signal pretreatment unit 24, abnormal sound positioning unit 25 and audio feature extraction unit 26; Sensor acquisition equipment 13 self is formed the sensor platform analyzing subsystem; Video intelligent analyzing subsystem, audio intelligent analyzing subsystem and general purpose transducer platform analyzing subsystem are sent to examination hall behavior single mode recognition system 3 with each self information through passage separately and handle.

Said video intelligent analyzing subsystem is realized following function through the analysis to vision signal: target is extracted and is followed the tracks of; The abnormal motion behavioural analysis.

Said audio intelligent analyzing subsystem is through the analysis to audio signal; Realize following function: 1) abnormal sound identification: the abnormal sound identification module through to contain in a large number abnormal sound as; Shriek; Collective's clamour; The broken sound of glass; The sample of sob etc. is learnt; And be at ordinary times normal sound but in the examination hall for improper sound as; Laugh; Brouhaha etc.; With the normal sound in the examination hall; As; Tables and chairs move sound; Paper stirs sound during hai roll; The supervisor reads out sound such as rule and carries out pattern classification, thereby has possessed the recognition capability to abnormal sound; 2) abnormal sound location (needing the support of abnormal sound positioning unit 25): through the pick-up array source of abnormal sound is discerned, realized Primary Location to sound source.This function be used for PTZ (Pan/Tilt/Zoom: The Cloud Terrace is comprehensive move and camera lens becomes doubly, zoom control) the collaborative interlock of camera, the guiding ptz camera turns to sound source position, further utilizes the video analysis function to differentiate again.

The analysis of said sensor platform analyzing subsystem is except that video, sensor signal the audio signal; Sensor acquisition equipment 13 is another main input signals of examination hall intelligent monitor system; It is as effective expansion of video/audio signal; It is many, corresponding fast and characteristics such as dependable performance to give full play to the transducer kind, is specially adapted to some special monitoring places.

Extendible transducer has human-body infrared sensing device, radio frequency signal detector etc., and sensor type, form and quantity can be expanded according to actual needs.

The machine vision of applied video signal collecting device 11 and machine learning techniques, said video intelligent analyzing subsystem carries out real-time analysis to vision signal, according to the motion feature of moving target, identifies the doubtful unlawful practice in all kinds of examination halls in the visual field.

Through to tens thousand of all kinds of audio samples feature learnings, the audio intelligent analyzing subsystem has been set up the model of cognition to a few quasi-representative abnormal sounds, and this subsystem carries out real-time grading to audio signal, determines abnormal sound.

Preferred embodiment as a kind of examination hall of the present invention intelligent monitor system; As shown in Figure 1, examination hall monitoring management background system 5 comprises that message prompting and administrative unit 51, Streaming Media show and playback unit 52, the setting of behavior recognition unit and upgrading unit 53, configuration of examination hall Behavior Monitor System and administrative unit 54.The recognition result that message prompting and administrative unit 51 reception examination hall behavior single mode recognition systems 3 are sent with message, and adopt database technology to manage; Streaming Media shows Voice Surveillance signal and/or the video monitoring signal of seeing off with playback unit 52 receiving stream media interfaces 20, and show, playback and management; The setting of behavior recognition unit is carried out long-range setting with the upgrade 53 pairs of abnormal behaviour model libraries 31 in unit and the recognizer parameter of abnormal sound model library 33 and the parameter interface of other each unit modules, and signal input system 1, signal pre-processing system 2 and examination hall behavior single mode recognition system 3 are carried out whole software upgrading; Examination hall Behavior Monitor System configuration is configured and manages with 54 pairs of said examination hall behavior intelligent monitor system software and hardwares of administrative unit, and through connecting examination hall invigilator teacher communication equipment 6 long-range transmission instruction or system identification information.The setting of examination hall monitoring management background system 5 is invigilated the personnel inspection monitoring effect easily and is in time made correction, setting.

Preferred embodiment as a kind of examination hall of the present invention intelligent monitor system; As shown in Figure 1; Said examination hall intelligent monitor system also comprises multi-modal decision-making of examination hall behavior and recognition system 4, and multi-modal decision-making of examination hall behavior and recognition system 4 comprise multi-modal integrated unit 41, the multi-modal recognition rule of examination hall behavior and model library 42 and the multi-modal recognition unit 43 of examination hall behavior based on decision-making level.Behavior multi-modal recognition unit 43 in examination hall merges the signal that video anomaly analysis unit 32, video anomaly analysis unit 34 and sensor acquisition equipment 13 obtain according to the multi-modal integrated unit 41 based on decision-making level; Carry out analysis-by-synthesis according to multi-modal recognition rule of examination hall behavior and model library 42, and will send comprehensive recognition result and send to examination hall monitoring management background system 5.The single mode behavior recognition result in each examination hall and sensor signal are sent to the multi-modal decision-making of examination hall behavior and recognition system 4 is carried out information fusion, obtains the multi-modal description of in examination process examinee and supervisor's behavior.According to comprehensive description information to the examination hall situation; System adopts rule-based reasoning and decision-making technique that the type of degree, individuality or the group behavior of examinee in the examination hall or supervisor's abnormal behaviour and the anomalous event that other merit attention etc. are discerned, and relevant information message and audio-video signal can be sent to examination hall monitoring management background system 5 through the TCP/IP network shows and alarm.When multi-modal decision-making of this examination hall behavior and recognition system 4 work; The audio-video signal of handling through examination hall behavior single mode recognition system 3 sends the multi-modal decision-making of examination hall behavior to and recognition system 4 is discerned judgement, and behavior single mode recognition system 3 sends the multi-modal decision-making of examination hall behavior to and recognition system 4 is discerned judgement to the pilot signal that sensor acquisition equipment 13 receives through the examination hall.Examination hall behavior single mode recognition system 3 sends to the multi-modal decision-making of examination hall behavior and recognition system 4 data are divided into two types.The first kind is the audio/video flow of compressed encoding, is used in the examination hall monitoring management background system 5 and watches.Second type is the Signal Pretreatment message, and the result that examination hall behavior single mode recognition system 3 will be discerned through the form of message mails to multi-modal decision-making of examination hall behavior and recognition system 4, further classifies and judges.Message format, encryption method and transmission rate can be adjusted according to application requirements, transmission mode.Message content can comprise following content: 1, the time; 2, processing unit numbering; 3, signalling channel numbering; 4, signalling channel state; 5, anomalous event identification grade and content; 6, anomalous event occurrence positions (video) or orientation (audio frequency).Examinee's behavior classification and message design as follows:

1 time: T000000

2 processing units numbering: S0000

3 signalling channels numbering: CH00

4 anomalous events identification grade and content:

4.1 inferior grade: C000 does not trigger warning, keeping records

4.1.1 examinee's quick acting

4.1.2 the doubtful action of examinee for example

4.1.2.1 the doubtful action 1 of examinee: bend to pick up thing

4.1.2.2 the doubtful action 2 of examinee: frequent rotary head

4.1.3 the frequent doubtful action of examinee

4.1.4 laugh, brouhaha take place in the examination hall

4.2 middle grade: B000 reports to the police in triggering, keeping records

4.2.1 the concrete abnormal behaviour classification of examinee

4.2.1.1 the doubtful behavior 1 of examinee: the examinee saws the air

4.2.1.2 the doubtful behavior 2 of examinee: the examinee leaves the seat

4.2.1.3 the doubtful behavior 3 of examinee: the examinee looks back at back row

4.2.2 invigilator teacher abnormal behaviour classification

4.2.2.1 invigilator teacher behavior 1: get into examinee's seating area

4.2.2.2 invigilator teacher behavior 2: stand in the preceding overlong time of camera

4.2.3 shriek, sob, the broken sound of glass take place in the examination hall

4.3 high-grade: A000 triggers high warning, keeping records

4.3.1 examinee colony moves significantly

4.3.2 leaving, examinee colony examines the position

4.3.3 examinee colony clamour takes place in the examination hall

4.4 other: 000000 triggers prompting, keeping records

4.4.1 camera blocks

4.4.2 camera deviation post

4.4.3 missing image

4.4.4 picture noise

4.4.5 image brightness sudden change

4.5 incident occurrence positions: P0000.

As shown in Figure 1; As the preferred embodiment of examination hall of the present invention intelligent monitor system, multi-modal recognition rule of examination hall behavior and model library 42 can be learnt and upgrade model of cognition and the rule in multi-modal recognition rule of said examination hall behavior and the model library 42 according to the result of examination hall monitoring management background system 5.Multi-modal decision-making of examination hall behavior and recognition system 4 pairs of vision signals of being gathered, audio signal and other sensor signals are differentiated decision-making; Draw the result of decision to the doubtful unlawful practice in examination hall; The result of decision is sent to examination hall monitoring management background system 5, examination hall monitoring management background system 5 also can be an alarm again.

Multi-modal decision-making of examination hall behavior and recognition system 3 are cores of whole system; Multi-modal decision-making of examination hall behavior and recognition system 3 receive audio/video characteristic information and sensor information; Event property is differentiated on the knowledge aspect according to doubtful unlawful practice decision rule storehouse, examination hall; This system's very first time is prompted to the background monitoring personnel with recognition result; Trigger the tampering devic at the scene of being installed in by the invigilator personnel; The supervisor communicates with the examination hall; Pro-active intervention is carried out in doubtful examination unlawful practice, thus real prison and the control that realizes incident.

The upgrade function of multi-modal recognition rule of examination hall behavior and model library 42 can be learnt and upgrades model and the rule in multi-modal recognition rule of examination hall behavior and the model library 42 according to the result of examination hall monitoring management background system 5.Learning algorithm mainly adopts statistical machine learning model and HMM (HMM) learning models such as neural net, SVMs.

Multi-modal decision-making of examination hall behavior and recognition system 4 are carried out when learning in the following manner:

1, based on the model of knowledge

Heuristics and analysis ratiocination process that the imitation people is adopted in target identification; The information (comprising classification, structure etc.) of target left in advance with the form (syntactic rule, framework, logical relation) of knowledge set up in the good knowledge base, adopt didactic method to carry out identification of targets.Logic template method wherein and fuzzy set theory have embodied the data description that people's understanding experience is carried out, and can differentiate known behavior through the mode of Rulemaking.

2, based on the inference technology of characteristic:

After characteristic carried out association, get the task that the associating characteristic is accomplished each class targets of identification, adopt based on the method for the methods and applications information theory technology of statistical computation event property is carried out reasoning.

3, based on the method for physical model:

Directly calculate clarification of objective according to physical model, set up objective attribute target attribute model different classes of or that have the inhomogeneity parameter according to the classification of objective attribute target attribute, and the information of forecasting that provides through target measurement information and physical model mates to come recognition objective.

4, collaborative interlock:

Work in coordination with and be meant coordinated, the information sharing between the multi-modal passage, accomplish the process of monitor task jointly.Different with multi-modal information fusion, roles' such as the primary and secondary between this each mode of accent with better function, auxiliary and enforcement distribution.

Interlock is meant to the intervention of decision conclusions and implements that purpose is in order to stop further developing of incident.

Multi-modal recognition rule of examination hall behavior and model library 42 are used for storing expertise, can be made up of factual knowledge and reasoning sex knowledge, comprise the rule of description relation, phenomenon and method, and the knowledge of in the expert of system scope, dealing with problems.Knowledge base is carried out reasoning and calculation to the knowledge of storage in advance according to preassigned rule.

Multi-modal recognition rule of examination hall behavior and model library 42 are divided into real-time data base and non-real-time data storehouse, and real-time data base is used for current observed result being provided and merging needed various other data to system, and storage intermediate object program.The supplementary of non-real-time data library storage historical data and relevant environment and target.

Thereby can also learn the renewal criterion according to the result of examination hall monitoring management background system 5 when judging the doubtful unlawful practice in identification examination hall with model library 42 through the pilot signal characteristic with the multi-modal recognition rule of examination hall behavior based on the multi-modal integrated unit of decision-making level 41.

As shown in Figure 1, as the preferred embodiment of examination hall of the present invention intelligent monitor system, when using multi-modal decision-making of examination hall behavior and recognition system 4, examination hall monitoring management background system 5 can be provided with each parameter of multi-modal decision-making of examination hall behavior and recognition system 3; The demonstration of identification judged result; The calling of pilot signal, playback.Examination hall monitoring management background system 5 is that the invigilator teacher provides examination hall monitor mode efficiently with the good man-machine interaction; The management of the input of many group signals and pretreatment system 2, examination hall behavior single mode recognition system 3, the multi-modal decision-making of examination hall behavior and recognition system 4 can be provided through the TCP/IP network, realize the real-time monitoring of simultaneously a plurality of examination hall examination processes.Monitoring management background system in examination hall is according to the grade of anomalous event, and the invigilator teacher on-the-spot to examination sends prompting and the event intervention instruction that has doubtful unlawful practice, can send prompting message to patrolling test system on the net through the TCP/IP network.

Specifically, the examination hall Behavior Monitor System configuration of examination hall monitoring management background system 5 possesses following function with administrative unit 54:

◆ intelligent streaming media server title, IP address setting; Picture quality, result of broadcast, network parameter are set

◆ examination hall management information: examination time started; The examination concluding time; The time set of handing in an examination paper; The hai roll time set; Supervisor's quantity; Parameter setting;

◆ camera calibration: demarcate automatically based on desk; Manual calibrating camera; Manual drawing desk zone

◆ the client operation function: log list selects to play video recording (no DVR pattern); List of devices selects to play real-time video; The invigilator teacher filters.

◆ examination hall behavior recognition unit parameter setting: target size; Susceptibility; Renewal rate;

Examination hall monitoring management background system 4 convenient invigilator's personnel inspection monitoring effects are also in time made correction, setting.

Correct understanding for incident is the one side of monitoring; The prior task of supervisory control system is the pro-active intervention that is for the doubtful unlawful practice in examination hall; The incident that prevents further worsens; Can't incident be done in time to handle to avoid the monitor staff to obtain prompting timely; To such an extent as to lose control to the state of affairs. the examination hall intelligent monitoring is except realizing the picket surveillance for incident; Realize communicating by letter with on-the-spot invigilator teacher through the examination hall communication equipment simultaneously, it can be intervened live event in real time.

The present invention has also made up the examination hall intelligent control method of a cover to the doubtful unlawful practice identification in examination hall; This method is based on a kind of examination hall intelligent monitor system; This supervisory control system comprises signal input system 1; Signal pre-processing system 2; Examination hall behavior single mode recognition system 3 and examination hall monitoring management background system 5; Signal input system 1 is sent to signal pre-processing system 2 with the audio-video signal of gathering and carries out preliminary treatment; Then pretreated signal is sent to examination hall behavior single mode recognition system 3; Discern the judgement classification by examination hall behavior single mode recognition system 3; Thereby the recognition result that is produced is sent to examination hall monitoring management background system 5; Realize the purpose of examination hall behavior monitoring, said method comprises step:

Pilot signal is gathered and sent to S1, signal input system 1;

S2, signal pre-processing system 2 preliminary treatment and the said pilot signal of feature extraction and send preliminary treatment and feature extraction after pilot signal;

S3,3 pairs of pretreated pilot signals of examination hall behavior single mode recognition system and characteristic thereof are carried out single mode identification and doubtful unlawful practice classification, and produce recognition result;

S4, examination hall monitoring management background system 5 receive said recognition result and deal with.

Among above-mentioned steps S1, the S2; The collection of signal input system 1 and the said pilot signal of signal pre-processing system 2 preliminary treatment comprise: said video monitoring signal is carried out digitlization, compressed encoding processing, signal condition judgement and Signal Pretreatment, carry out feature extraction subsequently and handle; Said Voice Surveillance signal is carried out digitlization, compressed encoding processing, signal condition judgement and Signal Pretreatment, carry out feature extraction subsequently and handle;

Among the above-mentioned steps S3, pretreated pilot signal and characteristic thereof are carried out single mode identification classify with doubtful unlawful practice and comprise:

A, the various characteristic quantities that vision signal feature extraction unit 23 or audio signal characteristic extraction unit (26) are extracted are selected, weighting and dimension-reduction treatment, the constitutive characteristic vector;

B, video anomaly analysis unit 32 select model of cognition corresponding in the abnormal behaviour model library 31 to carry out real-time grading according to said characteristic vector, and audio frequency anomaly analysis unit 34 selects model of cognition corresponding in the abnormal sound model library 33 to carry out real-time grading according to said characteristic vector;

Storage is a series of said model of cognition to behavior of various examination halls and sound that trained in abnormal behaviour model library 31 or the abnormal sound model library 33, the grader of said model of cognition for adopting machine learning techniques to make up.

Among the above-mentioned steps S4, examination hall monitoring management background system 5 receives said identification judged result and deals with and comprises: each parameter that said examination hall intelligent monitor system can be set; The demonstration of identification judged result; Management of the calling of pilot signal, playback and each unit of intelligent monitor system, said examination hall and module etc.

As the preferred embodiment of examination hall of the present invention intelligent control method, comprise that also the recognition result of multi-modal decision-making of examination hall behavior and recognition system 4 reception examination hall behavior single mode recognition systems 3 carries out multi-modal analysis-by-synthesis and sends to said examination hall monitoring management background system 5 processed steps: said multi-modal analysis-by-synthesis comprises:

A, the lower layer signal that examination hall behavior single mode recognition system 3 is sent are carried out fusion treatment through blending algorithm, and said blending algorithm comprises: wavelet transformation, weighted average, production rule and Kalman filtering;

B, the information after step S2 handles is carried out the characteristic fusion treatment through algorithm; The feature description Vector Fusion of a plurality of low dimensions is formed the more associating characteristic vector parameter of higher-dimension, and said algorithm comprises: parameter template, clustering methodology, fuzzy set theory, possibility theory, adaptive neural network, physical model, blackboard model and logic template method;

C, the signal of handling through step S3 is discerned judgment processing through algorithm, said algorithm comprises: classical reasoning, Bayesian inference, statistical decision, DS evidence theory, fuzzy theory, neural net, expert system, rough set theory, broad sense evidential reasoning theory, heuristic method, maybe theory and default reasoning.

Multi-modal decision-making of examination hall behavior and recognition system 3 main completion are judged the identification of information, are cores of the present invention, and in the present embodiment, multi-modal decision-making of examination hall behavior and recognition system 3 major functions and structure are mainly:

(1) multi-modal decision-making of examination hall behavior and recognition system 3 are hierarchical structure in logic, and each layer structure and function are:

A. metadata layer: be used for gathering, transmitting each multi-modal passage;

B. parameter and characteristic layer: be used for extracting and transmitting each mode passage relevant parameter and feature extraction;

C. model layer: the model of cognition that is used to train each passage; Storage and analysis and the recognition function of using each passage;

D. decision-making level: belong to the top layer of whole system framework, be used to gather the identification conclusion that each channel pattern obtains, carry out the comprehensive judgement of event property.

(2) combine above-mentioned (1): multi-modal decision-making of examination hall behavior and recognition system 3 need be carried out fusion treatment to all types of information after the information of receiving, its process may be summarized to be:

A. data level: the data level of lowermost layer; Directly comprised the primitive character fullest, the most effectively described; Because a large amount of property, the complexity of characteristic and the strong correlation between the data etc. of data make that it almost is impossible directly utilizing the fusion of initial data.Therefore the fusion of data level can only be a theoretical part of level framework, and the data Layer blending algorithm has: wavelet transformation, weighted average, production rule, Kalman filtering etc.

B. parameter, characteristic and model level: parameter comprises the comparatively information of low layer of each mode passage, mainly characterizes the general relationship between the element discrete in the metadata, like image parameter.On the other hand, also characterized the parameter of each mode passage itself, these parameters are all influential for the characteristic and the processing method of each mode.

The input data obtain its feature description vector through after the front-end processing respectively for each mode passage, and the processing of merging through characteristic then forms the more associating characteristic vector parameter of higher-dimension with the feature description Vector Fusion (merging) of a plurality of low dimensions.When system modelling, carry out the foundation of model to this high dimension vector, during coupling, utilize this high dimension vector to mate.The blending algorithm of parameter, characteristic layer has: parameter template, clustering methodology, fuzzy set theory, possibility theory, adaptive neural network, physical model, blackboard model, logic template method etc.

C. decision level fusion: the system of Decision Fusion; Input signal obtains characteristic parameter through handling, and carries out the modeling and the identification of single mode then respectively, with the intermediate object program parameter of identification; Merge through the Decision Fusion module, obtain final identification result through multi-modal decision making algorithm then.Decision-making level's blending algorithm has: classical reasoning, Bayesian inference, statistical decision, D-S evidence theory, fuzzy theory, neural net, expert system, rough set theory, broad sense evidential reasoning theory, heuristic method, possibility theory, default reasoning etc.

The cooperation of three levels of information fusion is adjusted according to the characteristic of mode, and the data that obtain like each mode passage are complementary, and then can carry out data Layer and merge, otherwise then merge at characteristic layer or decision-making level.

(3) examination hall behavior single mode recognition system 3 is hierarchical structure in logic, and each layer structure and function are:

(4) combine above-mentioned (1): examination hall behavior single mode recognition system 3 need be carried out fusion treatment to all types of information after the information of receiving, its process may be summarized to be:

Among the present invention, signal input system 1 comprises according to the treated parameter and the characteristic of obtaining that draw analyzed of vision signal with signal pre-processing system 2:

Physical parameter: intrinsic parameters of the camera, video camera is with respect to the external parameter (obtaining through camera calibration) of environment;

Image parameter: comprise color, texture and brightness;

Target geometric properties: foreground blocks position, edge, characteristic point, profile and shape;

Target travel characteristic: relative motion relation between motion vector (the real time kinematics speed, the direction that comprise characteristic point), the target.

Among the present invention, signal input system 1 comprises according to the treated parameter and the characteristic of obtaining that draw analyzed of audio signal with signal pre-processing system 2:

Volume parameters: comprise amplitude, frequency and short-time energy;

Fluctuation parameter: comprise zero-crossing rate and fundamental frequency;

Pronunciation character: comprise MFCC (Mei Er cepstrum coefficient) and LPC (linear prediction cepstrum coefficient);

Prosodic features: comprise intonation, stress and pause.

The conclusion that multi-modal decision-making of examination hall behavior and recognition system 7 obtain each mode passage; On the knowledge aspect; Carry out the character of decision event according to the rule base of different scenes and incident formulation. for example, being personal behavior or group behavior, is slight violation or serious violation etc.

Just combine specific embodiment below, be described further:

This instance system has adopted the system architecture of serial parallel coexistence, and the video signal collection apparatus 11 in the signal input system 1 is a video camera; Audio signal sample equipment 12 is pick-up, is used for the audio frequency acquiring signal; Sensor acquisition equipment 13 is various kinds of sensors; Signal pre-processing system 2 is made up of computer with examination hall behavior single mode recognition system or is realized by the Intelligent treatment equipment of digital signal processing platform.

Video camera; Pick-up; Equipment such as transducer are the signal input devices of native system; Be used for gathering video; Audio frequency and other interested signals; The Intelligent treatment equipment of signal pre-processing system 2 available computers frameworks or digital signal processing platform is realized; Be used for the initial data of system's input is analyzed and feature extraction; Wherein the computer hardware framework is mainly used in accurate target tracking result under the middle case of low density case of multi-channel video; Digital signal processing chip framework treatment facility is mainly used in crowd's state; Audio analysis; Basic datas such as general purpose transducer are extracted; Among Fig. 1; Video, audio frequency and the sensor characteristic data of 4 pairs of multi-modal decision-making of examination hall behavior and the recognition systems behavior single mode recognition system 3 through the examination hall are carried out comprehensive distinguishing; Identify the doubtful unlawful practice in examination hall; And then output examination hall monitoring management background system 5, store simultaneously, operation such as warning.Can form network through TCP/IP network or other communication modes between video camera, pick-up and other sensor devices, examination hall behavior single mode recognition system 3, the multi-modal decision-making of examination hall behavior and recognition system 4 and the examination hall monitoring management background system 5 communicates.

Among Fig. 1, the serial framework that the passage lock of each mode adopts is that feature extraction and pattern recognition are split as two independent sectors, has examination hall behavior single mode recognition system 3 and multi-modal decision-making of examination hall behavior and recognition system 4 to accomplish respectively respectively; Among Fig. 1; The adoptable parallel framework of system is that processing unit of each camera arrangement carries out synchronous independent process; Characteristic that extracts and abstract information of coming out only need less data volume to represent; Reduced the burden of information transmission network, for the basic data that extracts for Analysis server require relatively low.Complicated analysis task has been carried out functional decomposition, has alleviated the burden of processing unit, has realized effectively sharing of basic data.

Among Fig. 2, signal input system 1, signal pre-processing system 2, examination hall behavior single mode recognition system 3 and the multi-modal decision-making of examination hall behavior and recognition system 4 can be carried on 1 examination hall intelligent monitoring server hardware platform.The server hardware platform can adopt the PC framework also can adopt the DSP framework.Examination hall monitoring management background system 5 can carry on 1 examination hall intelligent monitoring client hardware platform.The client hardware platform mainly adopts the PC framework.According to serial among Fig. 1+parallel system architecture, 1 examination hall behavior intelligent monitoring server can be realized the real-time monitoring to a plurality of examination halls.In Fig. 2, through examination point TCP/IP network, 1 examination hall intelligent monitoring management background system 4 can be managed Duo Tai examination hall behavior intelligent monitoring server simultaneously, thereby realizes the monitoring to whole examination point.In Fig. 1, system communicates by letter with patrolling test system on the net through examination district TCP/IP network, then can realize the examination hall behavior intelligent real-time monitoring task to wider examination point.

The learning process of the multi-modal decision-making of examination hall behavior and recognition system 3 is identical with the description of the learning functionality in the intelligent monitor system of above-mentioned examination hall.

The above only is embodiments of the invention; Be not so limit claim of the present invention; Every equivalent structure transformation that utilizes specification of the present invention and accompanying drawing content to be done, or directly or indirectly be used in other relevant technical fields, all in like manner be included in the scope of patent protection of the present invention.

Claims

1. examination hall intelligent monitor system is characterized in that: comprising:

Signal input system (1): be used for gathering and sending pilot signal;

Signal pre-processing system (2): be used for the said pilot signal of preliminary treatment and feature extraction and send preliminary treatment and feature extraction after pilot signal;

Examination hall behavior single mode recognition system (3): be used for pretreated pilot signal and characteristic thereof are carried out single mode identification and doubtful unlawful practice classification, and produce recognition result;

Examination hall monitoring management background system (5): be used to receive said recognition result and deal with.

2. examination hall according to claim 1 intelligent monitor system is characterized in that,

Said pilot signal comprises the pilot signal that video monitoring signal, Voice Surveillance signal and/or transducer receive;

Said signal input system (1) comprising:

Video signal collection apparatus (11): be used for collection, digitlization, compressed encoding and send video monitoring signal; And/or

Audio signal sample equipment (12): be used for collection, digitlization, compressed encoding and send the Voice Surveillance signal; And/or

Sensor acquisition equipment (13): be used to gather and send the pilot signal that corresponding transducer receives;

Said signal pre-processing system (2) comprising:

With said video signal collection apparatus (11) corresponding video signals pretreatment unit (21), regular event filter element (22), video features extraction unit (23); And/or

With the corresponding audio signal pretreatment unit (24) of said audio signal sample equipment (12), abnormal sound positioning unit (25), audio feature extraction unit (26);

Said vision signal pretreatment unit (21): be used for that said video monitoring signal is carried out signal condition and judge and preliminary treatment;

Said regular event filter element (22): be used for normal action in the examination hall is filtered;

Said video features extraction unit (23): be used for that the video monitoring signal that said vision signal pretreatment unit (21) was handled is carried out feature extraction and handle;

Said audio signal pretreatment unit (24): be used for that said Voice Surveillance signal is carried out signal condition and judge and preliminary treatment;

Said abnormal sound positioning unit (25): be used to estimate abnormal sound orientation in the examination hall;

Said audio feature extraction unit (26): be used for that the Voice Surveillance signal that said audio signal pretreatment unit (24) was handled is carried out feature extraction and handle;

Voice Surveillance signal and/or video monitoring signal adopt the Streaming Media mode to send to examination hall monitoring management background system (5) through Streaming Media interface (20).

3. examination hall according to claim 2 intelligent monitor system is characterized in that, said examination hall behavior single mode recognition system (3) comprising:

Said video anomaly analysis unit (32): be used for carrying out the abnormal behaviour classification through the video monitoring signal that abnormal behaviour model library (31) was handled said video features extraction unit (23); Or

Said audio frequency anomaly analysis unit (34): be used for carrying out the abnormal sound classification through the Voice Surveillance signal that abnormal sound model library (33) was handled said audio feature extraction unit (26);

Said abnormal behaviour model library (31) and said abnormal sound model library (33) are learnt according to the result of said examination hall monitoring management background system (5) and are upgraded the model of cognition in said abnormal behaviour model library (31) and the said abnormal sound model library (33).

4. examination hall according to claim 3 intelligent monitor system; It is characterized in that said examination hall monitoring management background system (5) comprising: the message prompting shows and playback unit (52), the setting of behavior recognition unit and upgrading unit (53), configuration of examination hall Behavior Monitor System and administrative unit (54) with administrative unit (51), Streaming Media;

Said message prompting and administrative unit (51): be used to receive the recognition result that said examination hall behavior single mode recognition system (3) is sent with message, and adopt database technology to manage;

Said Streaming Media shows and playback unit (52): be used for Voice Surveillance signal and/or video monitoring signal that receiving stream media interface (20) is seen off, and show, playback and management;

Said behavior recognition unit is provided with and upgrading unit (53): be used for the recognizer parameter of abnormal behaviour model library (31) and abnormal sound model library (33) and the parameter interface of other each unit modules are carried out long-range setting, signal input system (1), signal pre-processing system (2) and examination hall behavior single mode recognition system (3) are carried out whole software upgrading;

Behavior Monitor System configuration of said examination hall and administrative unit (54): be used for said examination hall behavior intelligent monitor system software and hardware is configured and manages, and through connecting examination hall invigilator's long-range transmission instruction of teacher's communication equipment (6) or system identification information.

5. examination hall according to claim 1 intelligent monitor system is characterized in that, also comprises

Multi-modal decision-making of examination hall behavior and recognition system (4) comprising: based on multi-modal integrated unit (41), the multi-modal recognition rule of examination hall behavior and model library (42) and the multi-modal recognition unit of examination hall behavior (43) of decision-making level,

The multi-modal recognition unit of said examination hall behavior (43) merges the signal that video anomaly analysis unit (32), video anomaly analysis unit (34) and sensor acquisition equipment (13) obtain according to the multi-modal integrated unit (41) based on decision-making level; Carry out analysis-by-synthesis according to multi-modal recognition rule of examination hall behavior and model library (42), and will send comprehensive recognition result and send to examination hall monitoring management background system (5).

6. examination hall according to claim 5 intelligent monitor system; It is characterized in that multi-modal recognition rule of said examination hall behavior and model library (42) are learnt according to the result of said examination hall monitoring management background system (5) and upgraded model of cognition and the rule in multi-modal recognition rule of said examination hall behavior and the model library (42).

7. examination hall according to claim 5 intelligent monitor system is characterized in that,

Said examination hall monitoring management background system (5) comprising: the message prompting shows and playback unit (52), the setting of behavior recognition unit and upgrading unit (53), configuration of examination hall Behavior Monitor System and administrative unit (54) with administrative unit (51), Streaming Media;

Prompting of said message and administrative unit (51): be used to receive the recognition result that the multi-modal decision-making of examination hall behavior and recognition system (4) are sent with message, and the employing database technology is managed;

Said behavior recognition unit is provided with and upgrading unit (53): be used for the recognizer parameter of abnormal behaviour model library (31), abnormal sound model library (33) and the multi-modal recognition rule of examination hall behavior and model library (42) and the parameter interface of other each unit modules are carried out long-range setting, whole software upgrading is carried out in signal input system (1), signal pre-processing system (2), examination hall behavior single mode recognition system (3) and the multi-modal decision-making of examination hall behavior and recognition system (4);

Behavior Monitor System configuration of said examination hall and administrative unit (54): be used for examination hall behavior intelligent monitor system software and hardware is configured and manages, and through connecting examination hall invigilator's long-range transmission instruction of teacher's communication equipment (6) or system identification information.

8. examination hall intelligent control method; Based on a kind of examination hall intelligent monitor system; Said examination hall intelligent monitor system comprises signal input system (1), signal pre-processing system (2), examination hall behavior single mode recognition system (3) and examination hall monitoring management background system (5), it is characterized in that, comprising:

Pilot signal is gathered and sent to S1, signal input system (1);

S2, signal pre-processing system (2) preliminary treatment and the said pilot signal of feature extraction and send preliminary treatment and feature extraction after pilot signal;

S3, examination hall behavior single mode recognition system (3) are carried out single mode identification and doubtful unlawful practice classification to pretreated pilot signal and characteristic thereof, and are produced recognition result;

S4, examination hall monitoring management background system (5) receive said recognition result and deal with.

9. examination hall according to claim 8 intelligent control method is characterized in that, said pilot signal is the pilot signal that video monitoring signal, Voice Surveillance signal and/or transducer receive;

Said step S1 comprises:

Video signal collection apparatus (11) collection, digitlization, compressed encoding also send video monitoring signal,

Audio signal sample equipment (12) collection, digitlization, compressed encoding also send the Voice Surveillance signal,

The pilot signal that corresponding transducer receives is gathered and sent to sensor acquisition equipment (13);

Said step S2 comprises:

Vision signal pretreatment unit (21) carries out signal condition to said video monitoring signal to be judged and preliminary treatment,

Regular event filter element (22) filters normal action in the examination hall,

The video monitoring signal that video features extraction unit (23) was handled said vision signal pretreatment unit (21) is carried out feature extraction and is handled;

Audio signal pretreatment unit (24) carries out signal condition to said Voice Surveillance signal to be judged and preliminary treatment,

Abnormal sound positioning unit (25) is estimated abnormal sound orientation in the examination hall,

The Voice Surveillance signal that audio feature extraction unit (26) was handled said audio signal pretreatment unit (24) carries out feature extraction to be handled.

10. examination hall according to claim 9 intelligent control method is characterized in that step S3 comprises:

A, the various characteristic quantities that said vision signal feature extraction unit (23) or said audio signal characteristic extraction unit (26) are extracted are selected, weighting and dimension-reduction treatment, the constitutive characteristic vector;

B, video anomaly analysis unit (32) select model of cognition corresponding in the abnormal behaviour model library (31) to carry out real-time grading according to said characteristic vector,

Audio frequency anomaly analysis unit (34) selects model of cognition corresponding in the abnormal sound model library (33) to carry out real-time grading according to said characteristic vector;

Storage is a series of said model of cognition to behavior of various examination halls and sound that trained in said abnormal behaviour model library (31) or the said abnormal sound model library (33), the grader of said model of cognition for adopting machine learning techniques to make up.

11. examination hall according to claim 10 intelligent control method; It is characterized in that; Also comprise: the recognition result that multi-modal decision-making of said examination hall behavior and recognition system (4) receive said examination hall behavior single mode recognition system (3) carries out multi-modal analysis-by-synthesis and sends to said examination hall monitoring management background system (5) processed steps, and said multi-modal analysis-by-synthesis comprises:

A, carry out fusion treatment through algorithm, the fusion treatment algorithm comprises: wavelet transformation, weighted average, production rule and Kalman filtering;

B, the signal after a step process is carried out the characteristic fusion treatment through algorithm; The feature description Vector Fusion of a plurality of low dimensions is formed the more associating characteristic vector parameter of higher-dimension, and characteristic fusion treatment algorithm comprises: parameter template, clustering methodology, fuzzy set theory, possibility theory, adaptive neural network, physical model, blackboard model and logic template method;

C, utilize the model in multi-modal recognition rule of examination hall behavior and the model library (42); Signal to handling through step b is discerned judgment processing through algorithm, and identification judgment processing algorithm comprises: classical reasoning, Bayesian inference, statistical decision, DS evidence theory, fuzzy theory, neural net, expert system, rough set theory, broad sense evidential reasoning theory, heuristic method, possibility theory and default reasoning.

12. examination hall according to claim 11 intelligent control method is characterized in that, also comprises the step of learning and upgrade the information in multi-modal recognition rule and the model library (42) according to the result of said examination hall monitoring management background system (5).