JP2009049905A - Stream processing server apparatus, stream filter type graph setting apparatus, stream filter type graph setting system, stream processing method, stream filter type graph setting method, and computer program - Google Patents

Abstract

An API (WebAPI) that automatically recognizes stream data and performs stream processing processing is arranged on each server on the network, and stream processing processing is performed in cooperation with each server.
A primary analysis unit 101 of a stream processing server device 100 extracts stream data that matches an extraction condition (identification condition) from stream data input from a network, and a secondary processing unit 102 The extracted stream data is processed, and the processed stream data is output to the network based on the destination information. Further, the stream filter type graph setting device 300 generates a stream data extraction condition and a processing condition for each stream processing server device 100 and outputs the stream data to the stream processing server device 100 via a network.
[Selection] Figure 1

Description

  The present invention relates to a technique for processing and editing data that arrives in real time one after another, such as media that is a moving image or audio, and in particular, an API (on the Web) that automatically recognizes and processes stream data that passes. Stream processing server device, stream filter type, which can be arranged on each server on the network, and processing conditions of the Web API arranged on each server on the network can be set from the network side. The present invention relates to a graph setting device, a stream filter type graph setting system, a stream processing method, a stream filter type graph setting method, and a computer program.

  When acquiring upper layer information such as media data transferred over a network, it is desired to increase the functionality of network control in content-based routing that reflects the contents of video and audio.

  In addition, general application developers with little expertise in media processing, etc. can share expression techniques such as skilled creators without preparing expensive video equipment and software on their own, and create complicated server programs It is hoped that it will be possible to create a public program without the need for this, thereby reducing the cost of solution services using media such as remote TV (TV) conferences and telemedicine, and online games. It is desired to activate entertainment areas such as video content sites through development and personal submission.

  As a support technology related to media editing and processing on such a network, for example, there is a “video editing server device, system, and program” that the applicant of the present application is currently applying (see Patent Document 1). This currently pending technology enables media editing by accessing a video editing server on a network and combining desired processing APIs in a graph. Furthermore, by learning the graph creation history of various application developers, an editing template is presented so that even beginners without specialized knowledge can easily edit and edit, and there has never been an expert. It is possible to recommend an editing template.

In addition, as a method for generally learning the structure like the template, there is a method using a kernel function (see, for example, Non-Patent Document 1). This determines the projection from structure to structure using a convolution kernel function with a partial graph structure. However, there is a problem that the calculation efficiency of the direct maximum likelihood graph is not always good even though the calculation amount is large and it can be used as a classifier.
Japanese Patent Application No. 2006-230619 Suzuki, Sasaki, Maeda, "Hierarchical Acyclic Directed Graph Kernel", IEICE Transactions, D-II, Vol.J88-D-II, No.2 (20050201), pp. 230-240

  By the way, in the conventional Web API, the destination of the server holding the API that the terminal wants to use is grasped, the stream data (video stream etc.) is transferred with the destination specified, and some processing (filter processing) is performed on the Web API server. ). However, this method is complicated because it is necessary to specify the Web API on the user side. In addition, since a Web API that is not a destination cannot acquire input data, it is difficult to link with other services.

  Therefore, as a first problem, a stream processing server device that can arrange on the network a Web API that automatically recognizes and inputs passing stream data and performs processing as a filter (for example, a filter that processes an image). The offer of was desired. It has also been desired to operate a plurality of stream processing server devices in cooperation with each other.

  In addition, when the above Web API is arranged on a server on the network, it is arranged based on a filter graph (a combination of desired processing APIs in the form of a graph). This filter graph is a plurality of candidate filters. It is determined by estimating a template (recommended filter graph) from the graph. For example, the above-described video editing server device disclosed in Patent Document 1 discloses a filter graph template generation method. In this video editing server device template generation method, a loop-free non-circular filter graph is learned data. Therefore, there is a problem that a cyclic structure having a complex loop (for example, a loop structure that returns a processing result to the previous processing) is not appropriately reflected in the template. This is an obstacle particularly when learning a recognition control mechanism that processes data and obtains its feedback and performs control again, or a mechanism that integrates information acquired from stream data.

  Therefore, as a second problem, it has been desired to provide a stream filter type graph setting apparatus that can appropriately reflect a filter graph including a complex loop structure in template estimation.

  The present invention has been made in view of such circumstances, and an object of the present invention is to arrange a Web API for automatically recognizing and processing passing stream data (such as a video stream) on each server on the network. In addition, a stream processing server device, a stream filter type graph setting device, and a stream filter type graph setting that can operate WebAPIs held in a plurality of servers in cooperation with each other and can appropriately estimate a template. A system, a stream processing method, a stream filter type graph setting method, and a computer program are provided.

  The present invention has been made to solve the above problems, and the stream processing server device of the present invention is a stream processing server device arranged on a network, and a stream input unit for inputting stream data from the network An extraction condition for extracting predetermined stream data is set in advance, and a stream extraction unit that extracts stream data that matches the extraction condition from the stream data input to the stream input unit; Processing conditions for processing stream data are set in advance, and a processing unit that processes the stream data extracted by the stream extraction unit based on the processing conditions, and a destination that identifies a destination to which the stream data is transmitted Information is preset and input to the stream input section. The stream data stream data and the processing unit has is processed, characterized by having a a stream output unit that outputs to the network based on the destination information.

  The stream processing server apparatus of the present invention inputs the extraction condition, the processing condition, and the destination information via the network, sets the input extraction condition in the stream extraction unit, and inputs the input And a setting unit that sets processing conditions in the processing unit and sets the input destination information in the stream output unit.

  Also, the stream filter type graph setting device of the present invention extracts stream data from stream data input via a network based on the input extraction condition, and the process in which the extracted stream data is input Processing the stream data based on the conditions, outputting the processed stream data to the network based on the input destination information, the stream processing server device identified by the network identification information, the extraction condition, the processing condition, A stream filter type graph setting device that outputs the destination information via the network, wherein the network identification information and a processing process in the stream processing server device identified by the network identification information are associated with each other Is stored in advance, and the filter storage unit Based on the network identification information included in the filter, the extraction condition and the processing condition are based on the graph storage unit in which the filter graph in which the filter is combined is accumulated, and the filter graph read from the graph storage unit. The graph setting unit generated for each stream processing server device, the extraction condition for each stream processing server device generated by the graph setting unit, and the processing condition are sent to the stream processing server device corresponding to the network identification information. A setting condition output unit for outputting via a network.

  The stream filter type graph setting device of the present invention calculates a connection probability network indicating a connection probability between filters in the filter graph based on the filter graph stored in the graph storage unit, and calculates the calculated Based on the connection probability network, a template estimation unit that estimates a filter graph including an input processing filter as a template, and the graph setting unit based on the filter graph estimated by the template estimation unit, The extraction condition and the processing condition are generated for each stream processing server device based on network identification information included in the filter.

  Further, in the stream filter type graph setting device of the present invention, the template estimation unit includes a filter for processing input based on the filter graph stored in the graph storage unit and the connection probability network. The existence probability of the filter graph is calculated for each filter graph, and the filter graph having the highest existence probability is estimated as a template.

  In the stream filter type graph setting device of the present invention, in the connection probability network, the link information connecting the filters is a two-dimensional element including a path length element between the filters and a connection probability element between the filters. It is characterized by being described.

  Further, the stream filter type graph setting system of the present invention provides a plurality of stream processing server devices arranged on a network, stream data extraction conditions and processing conditions in the stream processing server device for each stream processing server device. A stream filter type graph setting system configured with a set stream filter type graph setting device, wherein the stream processing server device receives a stream input unit for inputting stream data from the network, and a predetermined stream data. Extraction conditions to be extracted are set in advance, a stream extraction unit that extracts stream data that matches the extraction condition from the stream data input to the stream input unit, and a processing condition that processes the stream data. Forecast The stream data extracted by the stream extraction unit is preliminarily set with a processing unit that performs processing based on the processing condition and destination information that identifies a destination to which the stream data is transmitted. A stream output unit that outputs the stream data input to the input unit or the stream data processed by the processing unit to the network based on the destination information, and the stream filter type graph setting device includes: A filter storage unit in which a filter in which the network identification information and a processing process in the stream processing server device identified by the network identification information are associated is stored in advance, and a filter graph in which the filter is combined is accumulated. Read from the graph storage unit and the graph storage unit Based on the extracted filter graph, the extraction condition and the processing condition are generated for each stream processing server device based on the network identification information included in the filter, and the graph setting unit generates A setting condition output unit configured to output the extraction condition for each stream processing server device and the processing condition to the stream processing server device corresponding to the network identification information via the network.

  The stream processing method of the present invention is a stream processing method in a stream processing server device arranged on a network, and includes a stream input procedure for inputting stream data from the network by a computer in the stream processing server device. A stream extraction procedure in which extraction conditions for extracting predetermined stream data are set in advance, and stream data that matches the extraction conditions is extracted from the stream data input by the stream input procedure; and a stream Processing conditions for processing the data are set in advance, the processing procedure for processing the stream data extracted by the stream extraction procedure based on the processing conditions, and a destination for identifying the destination to which the stream data is transmitted Information And a stream output procedure for outputting the stream data set by the stream input procedure or the stream data processed by the processing procedure to the network based on the destination information. And

  Also, the stream filter type graph setting method of the present invention extracts stream data from stream data input via a network based on the input extraction condition, and the process in which the extracted stream data is input Processing the stream data based on the conditions, outputting the processed stream data to the network based on the input destination information, the stream processing server device identified by the network identification information, the extraction condition, the processing condition, A stream filter type graph setting method in a stream filter type graph setting device for outputting the destination information via the network, wherein the network identification information and the network identification information are executed by a computer in the stream filter type graph setting device. Identified by A procedure for storing in advance in the filter storage unit the filter associated with the processing in the system processing server device, a procedure for storing the filter graph combined with the filter in the graph storage unit, and reading from the graph storage unit Based on the filter graph, the extraction condition and the processing condition are generated for each stream processing server device based on the network identification information included in the filter, and the stream generated by the graph setting procedure A setting condition output procedure for outputting the extraction condition for each processing server device and the processing condition to the stream processing server device corresponding to the network identification information via the network is performed.

  In the computer program of the present invention, a stream input procedure for inputting stream data from the network and an extraction condition for extracting predetermined stream data are preset in the computer of the stream processing server device arranged on the network. A stream extraction procedure for extracting stream data that matches the extraction condition from the stream data input by the stream input procedure, and a processing condition for processing the stream data are set in advance, and The stream data extracted by the stream extraction procedure is preliminarily set with a processing procedure for processing based on the processing conditions and destination information for identifying a destination to which the stream data is transmitted, and is input by the stream input procedure. Stream The stream data processed by chromatography data or the processing procedure, a program for executing a stream output procedure for output to the network based on the destination information.

  Also, the computer program of the present invention extracts stream data from stream data input via a network based on the input extraction condition, and the extracted stream data based on the input processing condition. Processing, outputting the processed stream data to the network based on the input destination information, and sending to the stream processing server device identified by the network identification information, the extraction condition, the processing condition, and the destination information A filter storage unit that stores a filter in which the network identification information and the processing process in the stream processing server device identified by the network identification information are associated with the computer of the stream filter type graph setting device that outputs the data via the network Pre-stored procedure and the file Based on the procedure for accumulating filter graphs combined with each other in the graph storage unit and the filter graph read from the graph storage unit, the extraction condition and the processing condition are based on network identification information included in the filter. A graph setting procedure generated for each stream processing server device, an extraction condition for each stream processing server device generated by the graph setting procedure, and the processing condition are sent to the stream processing server device corresponding to the network identification information to the network. Is a program for executing a setting condition output procedure that is output via the.

  In the stream processing server device of the present invention, stream data that matches the extraction condition (identification condition) is extracted from the stream data input from the network, and the extracted stream data is subjected to processing conditions set in advance. Since the processing (processing by WebAPI) is performed and the processed stream data is output to the network based on the destination information, the processing is performed by automatically recognizing the passing stream data. Servers can be placed on the network.

  In the stream processing server device of the present invention, the extraction condition, the processing condition, and the destination information are input via the network. Accordingly, the stream data extraction condition (identification condition) It is possible to set the processing condition and the transfer destination of the processed result data from the network. Also, by setting a desired processing condition (WebAPI) for each stream processing server device from the network side, a plurality of stream processing server devices are linked (combined with WebAPI) to perform processing on the stream data. Can do.

  Further, the stream filter type graph setting device of the present invention has a graph storage unit in which filter graphs combined with filters are accumulated, and extracts stream data based on the filter graph read from the graph storage unit Since the conditions and processing conditions are generated for each stream processing server device and output to the stream processing server device via the network, the stream data extraction conditions (identification conditions) in the stream processing server device are thereby obtained. The stream filter type graph setting device can set the processing conditions and the transfer destination of the processed result data.

  In the stream filter type graph setting device of the present invention, a connection probability network indicating the connection probability between filters in the filter graph is calculated based on the filter graph stored in the graph storage unit, and this connection probability network Based on this, a template (recommended filter graph) is estimated, and based on this template, extraction conditions and processing conditions in the stream processing server device are generated for each stream processing server device. Each time the filter graph is stored in the storage unit, the connection probability network is updated, and a more appropriate template can be estimated.

  In the stream filter type graph setting device of the present invention, the existence probability of the graph is calculated for each graph stored in the graph storage unit based on the connection probability network, and the graph having the highest existence probability is estimated as a template. As a result, the filter graph that meets the conditions specified by the user and has the highest use frequency can be estimated as a template.

  In the stream filter type graph setting apparatus of the present invention, in the connection probability network, link information connecting the filters is described by a path length element between the filters and a two-dimensional element of the connection probability between the filters. As a result, a more appropriate template can be estimated.

  In the stream filter type graph setting system of the present invention, a plurality of stream processing server devices are arranged on the network, and each stream processing server device matches the extraction condition from the stream data input from the network. Stream data is extracted, the extracted stream data is processed, and the processed stream data is output to the network based on the destination information. Further, the stream filter type graph setting device generates a stream data extraction condition and a processing condition for each stream processing server device based on the filter graph read from the graph storage unit in which the filter graph is accumulated, and creates a network. Thus, a server for automatically recognizing passing stream data and performing processing can be arranged on the network. In addition, the stream data extraction condition (identification condition), the processing condition, and the transfer destination of the processed result data can be set from the stream filter type graph setting apparatus.

  Embodiments of a stream processing server device and a stream filter type graph setting device to which the present invention is applied will be described in detail below with reference to the drawings.

[Description of API and filter graph]
First, APIs and filter graphs used in the stream processing server device and the stream filter type graph setting device of the present invention will be described in advance. The data processed in the stream processing server device is stream data such as a moving image stream. The API is also called a Web API because it is arranged on a server on the network, and the API is also called a filter because it performs a stream processing process (filter process) on stream data. In this specification, stream processing (filtering) for stream data is also simply referred to as “processing”.

  The filter graph is a graph in which desired APIs are combined in the form of a graph, and is information describing the order of processing processing for target stream data. Examples of the processing process include a synthesis process, a contrast process, and a rendering process.

  As the composition process, for example, there are two images, one image is displayed on the entire screen, and another image is displayed on one small screen, or a plurality of images are displayed on one screen. There is processing to display at the same time.

  Each video effect process has more detailed conditions as parameters. For example, in edge enhancement filter processing, there is a parameter that specifies how many edges are enhanced. Further, for example, in the two-screen display process, there is a parameter that specifies at what position and in what size a small screen is displayed.

  An API (filter) executes each process of stream data processing such as video, and representative examples of the API include a synthesis process API, a contrast API, a rendering process API, a voice recognition API, a natural recognition API, and a natural process. There are language processing API, XML analysis API, and the like.

  The composition processing API is an API that synthesizes two or more images and outputs one image. There are three types of basic composition processing: layout composition, superposition, and time connection.

  The layout composition is a composition in which different videos are arranged and flowed on the same screen, for example, like multidimensional relay. The parameter includes position information on the screen.

  Superposition is a combination of processes that change while gradually fading out to another image, such as fade-out. A ratio that determines the degree of pixel value composition (α blending) is passed as a parameter. Note that it is common to pass a blend function as a function pointer as a parameter, as in C language.

  The time connection is, for example, a process of playing the video B after the video A ends (for example, the video signal A stops). There are basically no parameters.

  The contrast API is a function that can change the lightness ratio of white (RGB = 255, 255, 255) to black (RGB = 0, 0, 0). Each of the input and output is one image, and the parameter is Ratio (eg 700: 1).

  The rendering process API is a process for obtaining a two-dimensional image by arranging a 3D model three-dimensionally, setting a light source and a camera position, and performing perspective projection (a projection method in which a distant object becomes smaller). . There are various types of rendering methods, such as Z-buffer method, ray tracing method, radiosity method, and photon mapping. Basically, the input is not a video but a non-video such as one modeling space, and one video is output. That is, the rendering process is a general term for processes such as converting data other than video into video.

  In addition, there is an API that outputs other than video. For example, there is an API that returns an average value of all pixel values of a video, and an API (for rendering) that returns a shape model by receiving a video. The filter graph defines a chain structure of APIs as a filter graph, in which parameters necessary for the API are given and an output video or value is used as another API input.

[Description of Stream Processing Server Device]
FIG. 1 is a configuration diagram of a stream processing server apparatus 100 according to an embodiment of the present invention, and is a diagram for explaining an operation at the time of stream analysis. A plurality of stream processing server devices 100 are arranged on the network, and the plurality of stream processing server devices 100 and the stream filter graph setting device 300 constitute a stream filter graph setting system described later (FIG. 8). See).

  The stream processing server apparatus 100 includes a primary analysis unit 101 having a router function, a secondary processing unit 102 that performs filter processing using an API, a setting unit 103, and a result storage unit 130. In addition, the stream processing server device 100 includes a computer system including a CPU, a ROM, a RAM (all not shown), and the like.

  By using the stream filter type graph setting device 300 (details will be described later), the application developer A 104 transmits analysis setting information 105 in which input / output and parameters are described to the setting unit 103 of the stream processing server device 100. The setting unit 103 notifies the primary analysis unit 101 and the secondary processing unit 102 of primary analysis setting information (stream data extraction conditions) 106 and secondary processing setting information (stream data processing conditions) 107, respectively. The primary analysis unit 101 and the secondary processing unit 102 perform analysis (or processing) corresponding thereto. In the following description, the analysis setting information is simply referred to as “analysis setting”, the primary analysis setting information is simply referred to as “primary analysis setting”, and the secondary processing setting information is simply referred to as “secondary processing”. This is called “setting”.

  In the above configuration, a plurality of stream data (moving image streams, etc.) 112 transmitted from the media terminal or the server 140 are input to the primary analysis unit 101, and the primary analysis unit 101 sets the primary analysis setting from the plurality of stream data. A copy that meets the conditions is copied, and this copy data 109 is transmitted to the secondary processing unit 102. The secondary processing unit 102 performs an analysis process (for example, a screen configuration recognition process) on the duplicate data 109 received from the primary analysis unit 101, and stores a processing result 131 such as image processing data in the result storage unit 130. To do. Note that the processing in the secondary processing unit 102 is performed by the API described above.

  The stream input unit, the stream output unit, and the stream extraction unit described above correspond to the primary analysis unit 101, the processing processing unit described above corresponds to the secondary processing unit 102, and the setting unit corresponds to the setting unit 103. . Further, the extraction condition in the above-described stream processing server device corresponds to the primary analysis setting 106, and the processing condition corresponds to the secondary processing setting 107.

FIG. 2 is a diagram illustrating an example of setting information for the primary analysis unit and the secondary processing unit.
FIG. 2A shows a first example of the primary analysis setting. As a condition α for extracting (or identifying) input stream data (moving image stream or the like), a port number (for example, 8192) and , A protocol (for example, RTP) is set. In addition, as conditions for transmitting the duplicate data 109 of the stream data conforming to the condition α to the secondary processing unit 102, the IP address (for example, 101.111.120.4) of the secondary processing unit 102 serving as the destination, and the port A number (for example, 4500) is set. That is, the duplicate data 109 of the stream data that meets the condition α is transmitted to the secondary processing unit 102.

  FIG. 2B shows a second example of the primary analysis setting. As a condition β for identifying input stream data, a destination IP address (for example, 131.90.25.76), a port A number (for example, 4096) is set. In addition, as conditions for transmitting the duplicate data 109 of the stream data that matches the condition β to the secondary processing unit 102, the IP address (for example, 101.111.120.4) of the secondary processing unit 102 that is the destination, and the port A number (for example, 4600) is set. That is, the duplicate data 109 of the stream data that matches the condition β is transmitted to the secondary processing unit 102.

  FIG. 2C shows an example of the primary analysis setting 106 received from the external stream filter type graph setting device 300. In the analysis setting example shown in FIG. 2C, the input α (A) is the setting information of the primary analysis setting shown in FIG. 2A, and the input β (A) is shown in FIG. 2B. This is setting information for primary analysis settings. Further, the parameter and output (c) are setting information for the secondary processing setting.

  FIG. 2D shows an example of the secondary processing setting 107, and parameters such as a sampling rate (for example, 15), a discrimination threshold value (for example, 100), and a background color (for example, 126, 78, and 204). Is set. Further, the IP address (for example, 101.111.120.5) and the port number (for example, 3050) of the result storage unit 130 are set as the transmission destination.

  In the analysis setting example shown in FIG. 2C, the setting unit 103 automatically complements the primary analysis settings 106 and 2 even if the IP address in the stream processing server apparatus 100 is not described in detail. It can also be described in the next processing setting 107.

  As described above, in the example shown in FIG. 1, the stream data 112 is recognized, the processing result 131 is accumulated in the result accumulation unit 130, and the application developer B 111 can use the server by using another server or terminal 134. This is an example. The processing result 131 can be set not only in the result storage unit 130 but also in an external DB (database) or other stream processing server device by designating a specific IP address.

FIG. 3 shows a sequence at the time of the stream analysis.
With reference to FIG. 3, the analysis setting 105 is transmitted from the stream filter type graph setting device 300 to the setting unit 103 in the stream processing server device 100 (step S101).
Next, the primary analysis setting 106 is transmitted from the setting unit 103 to the primary analysis unit 101 (step S102), and the setting becomes valid in the primary analysis unit 101. Further, the secondary processing setting 107 is transmitted from the setting unit 103 to the secondary processing unit 102 (step S103), and the setting is validated in the secondary processing unit 102.

  On the other hand, stream data transmitted from the media terminal or server 140 is input to the primary analysis unit 101 (step S104). The primary analysis unit 101 determines whether or not the received stream data conforms to the conditions of the primary analysis setting 106, and if it conforms, the duplicate data 109 is transmitted to the secondary processing unit 102 (step S105).

  In this example, the stream data input to the primary analysis unit 101 passes through the primary analysis unit 101 as it is (without data processing), and is transmitted to the media receiving terminal or the server 141. (Step S106).

  The duplicate data transmitted to the secondary processing unit 102 is processed (processed) by the secondary processing unit 102, and the processing result 131 is transmitted to the result storage unit 130 and stored (step S107).

  As a result, the data stored in the storage unit 130 can be viewed from the server or the terminal 134. In this case, a result acquisition request 132 is transmitted from the server or terminal 134 to the result storage unit 130 (step S108), and information on the processing result 133 stored in the result storage unit 130 is acquired (step S109).

  FIG. 4 is a diagram for explaining the operation at the time of stream processing in the stream processing server apparatus 100 shown in FIG.

  The example shown in FIG. 4 is an example in which stream data (moving image stream or the like) 112 is processed in the secondary processing unit 102 and the processed data 110 is transmitted to another stream processing server device, a media receiving terminal, or the server 141. Is shown.

  Note that the result accumulating unit 130 can be set as the output destination of the processing result (processing data) as in the case of FIG.

FIG. 5 is a diagram illustrating an example of setting information during a stream processing operation.
FIG. 5A shows a first example of primary analysis setting, where a port number (for example, 8192) and a protocol (for example, RTP) are set as a condition α for identifying input stream data. Has been. Further, as conditions for transmitting the detour data 108 of the stream data conforming to the condition α to the secondary processing unit 102, the IP address (for example, 101.111.120.4) of the secondary processing unit 102 serving as the destination, and the port A number (for example, 4500) is set. That is, stream data that meets the condition α is transmitted to the secondary processing unit 102 as the bypass data 108.

  FIG. 5B shows a second example of the primary analysis setting. As a condition β for identifying input stream data, a destination IP address (for example, 131.90.25.76), a port A number (for example, 4096) is set. Further, as conditions for transmitting the detour data 108 of the stream data that matches the condition β to the secondary processing unit 102, the IP address (for example, 101.111.120.4) of the secondary processing unit 102 that is the destination, and the port A number (for example, 6000) is set. That is, the stream data that meets the condition β is transmitted to the secondary processing unit 102 as the bypass data 108.

  FIG. 5C shows an example of the primary analysis setting 106 received from the external stream filter type graph setting apparatus 300. In the analysis setting of FIG. 5C, the input α (A) is the setting information of the primary analysis setting shown in FIG. 5A, and the input β (A) is the primary analysis setting shown in FIG. 5B. This is setting information for analysis settings. The parameter and the transmission destination (c) are setting information for the secondary processing setting.

FIG. 5D shows an example of the secondary processing setting. As parameters, a sampling rate (for example, 15), a discrimination threshold (for example, 100), a background color (for example, 126, 78, and 204),
Is set. Further, the IP address (for example, 87.203.190.8) and the port number (for example, 5060) of the media terminal or the server 141 are set as the transmission destination.

  FIG. 6 shows a sequence at the time of the above stream processing, and shows an example in which the detour data 108 is generated in the primary analysis unit 101. The processing in steps S101 to S104 in the sequence diagram shown in FIG. 6 is the same as the processing in steps S101 to S104 in the sequence diagram shown in FIG. 3, and the processing after step S110 shown in FIG. Different.

  That is, in FIG. 6, a stream that meets the conditions of the primary analysis unit 101 is transmitted to the secondary processing unit 102 as detour data 108 (step S110), and after processing such as complex media recognition processing is performed ( In step S111, the processed data 110 is transmitted to the designated stream processing server device, media receiving terminal, or server 141 (step S112).

(Explanation of filter graph)
Next, the configuration of the filter graph will be described. As described above, the filter graph is a combination of desired APIs in a graph, and each API (filter) in the filter graph is assigned to each stream processing server device.
FIG. 7 shows the structure of a filter graph 200 in which APIs held in each stream processing server device (filter) A, B, C, D are connected. Here, the API held in the stream processing server apparatus A has a configuration indicated by the stream filter type Web API 113. The stream filter type Web API 113 schematically includes an input 114a for each analysis setting condition, 2 It can be represented by a filter diagram composed of a parameter 114b for the next processing setting and an output 114c.

  The filter graph 200 includes connection information 201 describing input / output and parameters of the filters A, B, C, and D. As for the connection relationship, the output destination address is that of the destination stream filter type WebAPI (for example, A, B, C, D, etc.), and the input condition of the destination stream filter type WebAPI is the IP address of the output stream filter type WebAPI (for example, A , B, C, D, etc.).

  Also, in the connection information 201, there is a case where “restoration information” is described in the output destination of the filter that is the final process, and this is based on the information tagged with the same restoration information in the input stream. This indicates that the destination is set and transmitted to the original destination again. However, the output destination does not necessarily have to be as described above, and a new IP address and port may be designated.

[Description of Stream Filter Type Graph Setting System and Stream Filter Type Graph Setting Device]
FIG. 8 is a configuration diagram of a stream filter type graph setting system. The stream filter type graph setting system is mainly configured by a stream filter type graph setting device 300 and a plurality of stream processing server devices A, B, and C.

  The stream filter type graph setting device 300 includes an API registration unit 311, a graph creation unit 308, a graph setting unit 306, an analysis setting unit 305, a graph storage unit 307, and a template estimation unit 304. The stream filter type graph setting apparatus 300 includes a computer system including a CPU, a ROM, a RAM (all not shown), and the like.

  As a method of using the stream filter type graph setting apparatus 300, the application developer 104 accesses the graph creation unit 308 using the operation terminal 312, passes the estimation condition 313 to the template estimation unit 304, and uses a template (recommended). Filter graph) 314 is received and modified and combined to create the desired filter graph (eg, filter graph 200 of FIG. 7).

  The created filter graph connection report 201 is passed to the graph setting unit 306. The graph setting unit 306 reads the IP address of each Web API from the registration information 320 registered in the API registration unit 311, complements the connection information 201, and converts it into setting information 316. This setting information 316 is transferred to the analysis setting unit 305, and is disassembled into setting information of each stream processing server apparatus 100 and transmitted. The connection information 201 is stored in the graph storage unit 307.

  The above-described filter storage unit corresponds to the API registration unit 311, the graph storage unit corresponds to the graph storage unit 307, the graph setting unit corresponds to the bluff setting unit 306, and the setting condition output unit includes the analysis setting unit 305. The template estimation unit 304 corresponds to the template estimation unit.

  FIG. 9 is a sequence diagram showing how the filter graph 341 actually set is executed. A stream (moving image stream or the like) transmitted from the transmission terminal or server 351 to the reception terminal or server 352 first reaches the stream processing server apparatus A that is in the process of passing (step S201). At this time, the transmission information of the first stream is extracted as restoration information 353 for later restoration, and is transmitted to the stream processing server apparatus C that is the final process (steps S202 and S203). Further, the stream itself passes through one after another while changing the destination to the stream processing server apparatus B and the stream processing server apparatus C (steps S204 and S205).

  In the stream processing server apparatus C for final processing, a destination is set based on the transferred restoration information 353. For this reason, if the restoration information has not been transmitted yet, a waiting is performed. Then, the stream is transmitted to the receiving terminal or server 352 (step S206).

(Explanation of connection probability network)
Next, a connection probability network used for template estimation in the template estimation unit 304 will be described.
FIG. 10 shows the structure of the connection probability network 400 stored in the graph storage unit 307.
Here, an example of filter graphs 401 and 402 to be learned will be described as an example.

As a first procedure, first, nodes (A, B, C, D) of the filter graphs α, β are searched with depth priority. Note that the subscript numbers ₁ , ₂ , ₃ , and ₄ in the nodes B ₁ , C ₂ , D ₃ , and A ₄ indicate the search order based on the depth (depth in the tree structure). The nodes are searched in the order indicated by the numbers 1, 2, 3, and 4.

The second procedure will be described with reference to FIG.
First, as shown in state 1, when a node f2 that has not been searched yet is found, the update element of f1 (o) → f2 (i) of the connection probability network is added to (+1, +1) together with the immediately preceding node f1. 1). Here, o is a number representing one of the output streams of f1, and i is a number representing one of the input streams of f2. For example, the update element of node B ₁ → node C ₂ is (+1, 1). This simply indicates that the number indicating the order of the newly discovered node f2 is incremented by +1 (left side) and there is one connection (right side).

  Next, as shown in state 2, when a node f2 that has already been searched is encountered, a graph for reaching f2 from the immediately preceding node f1 on the graph that has been searched so far (numbered) is used. The path length s is calculated. If there is no distance on the searched graph as in state 2, the update element of f1 → f2 of the connection probability network is set to (+1, 1).

Next, when there is a distance as in state 3, the sign differs depending on how the distance is traced. When tracing in the output direction from f1, the sign of s is set to be positive, and conversely, when tracing in the input direction, the sign of s is set to be negative. Using this s, the update element of f1 → f2 of the connection probability network is set to (s, 1). For example, the update element of the node D ₃ → the node C ₂ is (+2, 1).

As a third procedure, all update elements of the connection probability network are classified and calculated for each link.
At this time, even on the same link, since it is classified and added by the distance portion (ie, the left side) of (s, 1), it is (s, n _s ) (where n _s is the number of connections).

  Also, the total number of connections in the link

  age,

  Connection probability before update

And Here, _{m s} is the number of connections from the past of f1 to f2 (f1 → f2), M is all of the number that is connected from the past of f1. In this case, the updated connection probability is

  It becomes.

An example of this is shown in FIG. As shown in FIG. 12, an update element from node D (output 2) to node C (input 3) is obtained. For example, update elements from the node f1 to the node f2 are “(+1, 1), (+1, 1), (−1, 1), (+2, 1), (−1, 1), (+1, 1”. ), (+2, 1) ”, classify by the distance portion s and add together to obtain an update element (s, n _s ) from the node D (output 2) to the node C (input 3). (−1, 2), (+1, 3), (+2, 2).

And the connection probability before update, “( ₋₁ , m ₋₁ / M), (0, m ₀ / M), (+1, m ₁ / M), (+2, m ₂ / M), where m _s is the number of previous connections (before learning) to s, M is the number of updated element connections from D (output 2) to all nodes, and the updated connection probability is (−1, (m ₋₁ + N ₋₁ ) / (M + N)), (+2, (m ₂ + n ₂ ) / (M + N)).

  As shown in FIG. 10 by the above calculation method, an update element can be obtained from the filter graph α and the filter graph β, and a connection probability network 400 between nodes can be obtained from the update element.

(Explanation of template estimation method)
A template (recommended filter graph) can be estimated based on the connection probability network 400 described above. For example, as shown in FIG. 13, based on the connection probability network 400, the existence probability of the filter graph α (in this example, 1.23 × 10 ⁻⁴ ) and the existence probability of the filter graph β (in this example, 7 .41 × 10 ⁻⁴ ) and a filter graph β having a larger existence probability is estimated as a template.

Next, a template estimation method using this connection probability network will be described.
First, a function P (g) representing the existence probability of the filter graph g accumulated in the past is defined. The existence probability P (g) of the filter graph g is calculated as follows.

  As a first procedure, a node (API) existing in the filter graph g is searched with depth priority, with one of the external inputs as a root, and a (searched) label is assigned to the searched node. Go. Here, a graph composed only of the filter API to which a label is attached is defined as g ′. Further, the existence probability P (g ′) = 1 of the initial state (g ′ = φ) is set.

As a second procedure, in the first procedure, when the m-th input of the node f2 is reached during the search for the n-th output O _{f1, n} of the filter API: f1, the filter graph g ′ includes O _f1, A link of _n → I _{f2, m} is added and a new filter graph g ′ is set. Also, the current node f2 is newly set as the node f1.

As a third procedure, if a label has not yet been assigned in the second procedure, a connection probability p: (1, p) of O _{f1, n} → I _{f2, m} on the connection probability network is acquired, and P ( g ′) × p is set to a new existence probability P (g ′).

As the fourth procedure, when the m-th input of the node f2 already labeled in the second procedure is reached, O _{f1, n} on g ′ → The distance s of I _{f2, m} is calculated, and the connection probability p: (s, p) of O _{f1, n} → I _{f2, m} on the connection probability network is obtained, and P (g ′) × p is newly set. The existence probability P (g ′) is set.

  As a fifth procedure, the procedure 2 to the procedure 4 are repeated until g = g ′, and the final existence probability P (g) is calculated.

The template is estimated as follows using the existence probability P (g).
Here, a set of past filter graphs is G = {g (1), g (2),..., G (K)}, | G | = K, and a specific stream is set as a condition from the application developer. It is assumed that a set of filter-type WebAPIs F = {f (1), f (2),..., F (L)}, | F | = L is specified.

  Then, g (i) (here, 1 ≦ i ≦ K) including all elements of the filter API set F is selected in the order stored in the array, and the existence probability P (g) for the selected g (i) is selected. (I)) is calculated. At this time, instead of calculating the entire array, only a specific number k is calculated, and the template having the maximum probability is determined as a template. Further, in order to speed up the calculation in the next template estimation, the contents of the array are rearranged in descending order of the existence probability P (g (i)).

  For example, as shown in FIG. 13, in the connection probability network 400, the filters A, B, C, and D are designated, and as the estimation conditions, the external input probability 1/10 at the node A and the external input probability at the node B When 2/10, external input probability 3/10 at node D, and external output probability 1/8 are specified, the existence probabilities of filter graphs α and β are the external input and external output at each node, It is calculated as follows by sequentially multiplying the connection probability at each node.

Presence probability of filter graph α = (1/10) × (2/10) × (2/3) × (2/3) × (1/3) × (1/3) × (1/8) = 1 .24 × 10 ⁻⁴

Presence probability of filter graph β = (3/10) × (2/10) × (1/3) × (1/3) × (2/3) × (2/3) × (2/8) = 7 .41 × 10 ⁻⁴
Thereby, the filter graph β having a high existence probability is estimated as a template.

  The existence establishment P (g) may be calculated by the first to fifth procedures described above, but the existence establishment P (g) may be calculated as follows. First, as described with reference to FIG. 11, “If the node f2 that has already been searched is encountered as shown in the state 2, the immediately preceding node on the graph that has been searched so far (numbered) is displayed. In the case of calculating the path length s for reaching f2 from f1, the value of the calculated path length s is stored in the graph storage unit. Next, for the link l included in the past filter graph g, the connection probability p (l, s) is acquired from the connection probability network based on the path length s stored in the graph storage unit, and the acquired connection probability p Based on (l, s), the presence establishment P (g) is calculated by the following equation.

  By calculating the existence establishment P (g) as described above, it is possible to calculate the existence establishment P (g) without calculating the path length s again.

  FIG. 14 shows a comparison with the conventional method. In the conventional method 502, since the update element and the connection probability network are described by one-dimensional elements, even if the filter graph to be learned (for example, the filter graph 501) is almost constant, templates having different isotope structures are used. 504 may be estimated.

  On the other hand, in the case of the method 505 of the present invention, the template 506 having a more appropriate structure can be estimated because the update element and the connection probability network are described by two-dimensional elements including the path length.

  FIG. 15 shows a service in which advertisements are changed for each place (or region) when IP multicast is performed using the stream filter type graph setting system of the present invention. A filter graph set from the stream filter type graph setting device 300 is as shown in a filter graph 602.

  In this case, the filter graph 602 includes a video analysis API 610 for analyzing the video stream, a synthesis API (A) 611 for synthesizing the video stream and the advertisement α (CMα), and a video stream and the advertisement β (CMβ). A composition API (B) 612 that composes an image and a composition API (C) 613 that composes an image of a video stream and an advertisement γ (CMγ). Then, the stream filter type graph setting device 300 assigns the composite API (A), the composite API (B), and the composite API (C) to each stream processing server device 100 via the network.

  In each stream processing server device 100, a frame where an advertisement is to be inserted and a position on the video are specified from the video stream, and separate advertisements α, β, and γ are synthesized by individual synthesis filters. Users A, B, and C view separate advertisements α, β, and γ.

  This eliminates the need for individually synthesizing and distributing the advertisements on the distribution server 601 side, thereby reducing the load on the distribution server 601 and the network and reducing the threshold for service introduction.

  FIG. 16 shows a service that extracts and integrates information from various stream data using the stream filter type graph setting system of the present invention, and presents overall important information to the end user. Here, voices such as telephone calls, texts such as searches, and video images such as videophones are integrated and converted into real-time knowledge and synthesized as video on the videophone side. The filter graph is shown at 701.

  The filter graph 701 includes a speech recognition API 710 that recognizes speech data from stream data, a text analysis API 711 that analyzes text data, a video recognition API 712 that recognizes video data, and a presentation that generates information to be presented from each data. An information generation API 713, a rendering API 714 that performs rendering processing, and a synthesis API 715 that synthesizes video data and the rendered data are configured.

  Then, the stream filter type graph setting device 300 assigns individual APIs to the stream processing server device 100 on the network (NW).

  As a result, it is possible to present solution information by using totally different types of stream data, such as what the end user is wondering during a videophone call. For example, it is possible to integrate stream data of moving images and stream data that is text data, and provide the integrated stream data.

  FIG. 17 shows an example of realizing a service for converting the background of the end user into another background and showing it to the other party on the TV phone. In this example, the filter graph 801 set by the stream filter type graph setting apparatus 300 determines a trimming position, a trimming API 811 that performs trimming from image data, an object recognition API 812 that recognizes an object from the trimmed image data, and a trimming position. A position parameterization API 813 for generating a position parameter to be generated from an object recognition result, a replication API 814 for duplicating an image for rendering, a rendering API 815 for performing rendering processing, and a background synthesis processing API 816.

  Then, the stream filter type graph setting device 300 assigns individual APIs to the stream processing server device 100 on the network (NW).

  In this way, when complicated processing such as object recognition is performed in the middle, it is often realized to adapt sequentially while obtaining feedback without completing recognition once. In this case, a structure for adjusting the trimming location such as which portion to focus on, based on the recognition score, as in the filter graph 801 is required.

  Further, by using the stream filter type graph setting system of the present invention, such a service having a complicated loop structure can be used by an application developer without specialized knowledge by using a template.

  Although the embodiment of the present invention has been described above, the stream processing server apparatus 100 shown in FIGS. 1 and 4 and the stream filter type graph setting apparatus 300 shown in FIG. 8 may be realized by dedicated hardware. In addition, a program for realizing the function of each processing unit in the stream processing server apparatus 100 and the stream filter type graph setting apparatus 300 is recorded on a computer-readable recording medium, and the program recorded on the recording medium May be implemented by causing the computer system to read and execute the function. Here, the “computer system” includes an OS and hardware such as peripheral devices.

  Further, it is assumed that an input device, a display device, and the like (none of which are shown) are connected to the stream processing server device 100 and the stream filter type graph setting device 300 as peripheral devices. Here, the input device refers to an input device such as a keyboard and a mouse. The display device refers to a CRT (Cathode Ray Tube), a liquid crystal display device, or the like.

  Further, the “computer system” includes a homepage providing environment (or display environment) if a WWW system is used.

  The “computer-readable recording medium” refers to a storage device such as a flexible medium, a magneto-optical disk, a portable medium such as a ROM and a CD-ROM, and a hard disk incorporated in a computer system. Furthermore, the “computer-readable recording medium” dynamically holds a program for a short time like a communication line when transmitting a program via a network such as the Internet or a communication line such as a telephone line. In this case, a volatile memory in a computer system serving as a server or a client in that case, and a program that holds a program for a certain period of time are also included. The program may be a program for realizing a part of the functions described above, and may be a program capable of realizing the functions described above in combination with a program already recorded in a computer system.

  As mentioned above, although embodiment of this invention was described, the stream processing server apparatus of this invention, a stream filter type | mold graph setting apparatus, and a stream filter type | mold graph setting system are not limited only to the above-mentioned illustration example, Of course, various modifications can be made without departing from the scope of the present invention.

It is a figure for demonstrating the structure and operation | movement of a stream processing server apparatus (stream analysis). It is a figure which shows the example of the setting information with respect to the primary analysis part and secondary processing part at the time of a stream analysis. It is a figure which shows the sequence at the time of a stream analysis. It is a figure for demonstrating the structure and operation | movement of a stream processing server apparatus (stream processing). It is a figure which shows the example of the setting information with respect to the primary analysis part at the time of stream processing, and a secondary process part. It is a figure which shows the sequence at the time of stream processing. It is a figure which shows the example of the structure of a filter graph. It is a figure which shows the structure of a stream filter type | mold graph setting system. It is a figure which shows the example of a stream process by filter graph setting. It is a figure which shows the structure of a connection probability network. It is a figure which shows the calculation method of the distance between nodes. It is a figure for demonstrating the update element of a connection probability network. It is a figure for demonstrating the calculation method of presence establishment of a template. It is a figure which shows the example of a comparison with the conventional method. It is a figure which shows the example of the advertisement composition in the multicast concerning Example 1 of this invention. It is a figure which shows the example of the information presentation by the knowledge integration concerning Example 2 of this invention. It is a figure which shows the example of the feedback type object recognition concerning Example 3 of this invention.

Explanation of symbols

DESCRIPTION OF SYMBOLS 100 ... Stream processing server apparatus, 101 ... Primary analysis part, 102 ... Secondary processing part, 103 ... Setting part, 105 ... Analysis setting information, 106 ... Primary analysis setting Information 107 107 Secondary processing setting information 108 Detour data 109 Duplicated data 110 Processing data 112 Stream data 130 Result accumulating unit 134 .. server or terminal, 140... Media transmission terminal or server, 200... Filter graph, 300... Stream filter type graph setting device, 304. 306 ... Graph setting unit, 307 ... Graph storage unit, 308 ... Graph creation unit, 311 ... API registration unit, 312 ... Operation terminal, 341 ... Fill Graph, 351 ... Transmission terminal or server, 352 ... Reception terminal or server, 400 ... Connection probability network (connection probability NW), 401, 402, 501 ... Filter graph, 504, 506 ... Estimated template, 601 ... distribution server, 602, 701, 801 ... filter graph

Claims (11)

A stream processing server device arranged on a network,
A stream input unit for inputting stream data from the network;
An extraction condition for extracting predetermined stream data is set in advance, and a stream extraction unit that extracts stream data that matches the extraction condition from the stream data input to the stream input unit;
Processing conditions for processing the stream data are set in advance, and a processing unit that processes the stream data extracted by the stream extraction unit based on the processing conditions;
Destination information for identifying a destination to which stream data is transmitted is set in advance, and stream data input to the stream input unit or stream data processed by the processing unit is sent to the network based on the destination information. A stream output section to output;
A stream processing server device. The extraction condition, the processing condition, and the destination information are input via the network,
The input extraction condition is set in the stream extraction unit,
Set the input processing condition in the processing unit,
A setting unit for setting the input destination information in the stream output unit;
The stream processing server apparatus according to claim 1, comprising: Stream data extracted from the stream data input via the network based on the input extraction condition, the extracted stream data processed according to the input processing condition, and the processed stream A stream that outputs data to the network based on input destination information, and outputs the extraction condition, the processing condition, and the destination information to the stream processing server device identified by the network identification information via the network A filter type graph setting device,
A filter storage unit in which a filter associated with the network identification information and a processing process in the stream processing server device identified by the network identification information is stored in advance;
A graph storage unit storing a filter graph in which the filters are combined;
A graph setting unit that generates the extraction condition and the processing condition for each stream processing server device based on network identification information included in the filter, based on a filter graph read from the graph storage unit;
A setting condition output unit that outputs the extraction condition for each stream processing server device generated by the graph setting unit and the processing condition to the stream processing server device corresponding to the network identification information via the network;
A stream filter type graph setting device. Based on the filter graph stored in the graph storage unit, a connection probability network indicating a connection probability between the filters in the filter graph is calculated, and based on the calculated connection probability network, the input processing process is performed. A template estimation unit that estimates a filter graph including a filter as a template;
Have
The graph setting unit
Based on the filter graph estimated by the template estimation unit, the extraction condition and the processing condition are generated for each of the stream processing server devices based on network identification information included in the filter.
The stream filter type graph setting device according to claim 3. The template estimation unit is
Based on the filter graph stored in the graph storage unit and the connection probability network, the existence probability of the filter graph is calculated for each filter graph including the input processing filter,
Estimating the filter graph with the highest existence probability as a template;
The stream filter type graph setting apparatus according to claim 4. In the connection probability network, link information connecting the filters is described by a two-dimensional element of an element related to a path length between the filters and an element of a connection probability between the filters.
The stream filter type graph setting device according to claim 4 or 5. A plurality of stream processing server devices arranged on a network, and a stream filter type graph setting device that sets stream data extraction conditions and processing conditions for each stream processing server device for each stream processing server device A stream filter type graph setting system,
The stream processing server device
A stream input unit for inputting stream data from the network;
An extraction condition for extracting predetermined stream data is set in advance, and a stream extraction unit that extracts stream data that matches the extraction condition from the stream data input to the stream input unit;
Processing conditions for processing the stream data are set in advance, and a processing unit that processes the stream data extracted by the stream extraction unit based on the processing conditions;
Destination information for identifying a destination to which stream data is transmitted is set in advance, and stream data input to the stream input unit or stream data processed by the processing unit is sent to the network based on the destination information. A stream output section to output;
Have
The stream filter type graph setting device comprises:
A filter storage unit in which a filter associated with the network identification information and a processing process in the stream processing server device identified by the network identification information is stored in advance;
A graph storage unit storing a filter graph in which the filters are combined;
A graph setting unit that generates the extraction condition and the processing condition for each stream processing server device based on network identification information included in the filter, based on a filter graph read from the graph storage unit;
A setting condition output unit that outputs the extraction condition for each stream processing server device generated by the graph setting unit and the processing condition to the stream processing server device corresponding to the network identification information via the network;
A stream filter type graph setting system characterized by comprising: A stream processing method in a stream processing server device arranged on a network,
By the computer in the stream processing server device,
A stream input procedure for inputting stream data from the network;
An extraction condition for extracting predetermined stream data is set in advance, and a stream extraction procedure for extracting stream data that matches the extraction condition from the stream data input by the stream input procedure;
Processing conditions for processing stream data are set in advance, and processing procedures for processing the stream data extracted by the stream extraction procedure based on the processing conditions;
Destination information for identifying a destination to which stream data is transmitted is set in advance, and stream data input by the stream input procedure or stream data processed by the processing procedure is sent to the network based on the destination information. Stream output procedure to output,
A stream processing method characterized in that is performed. Stream data extracted from the stream data input via the network based on the input extraction condition, the extracted stream data processed according to the input processing condition, and the processed stream A stream that outputs data to the network based on input destination information, and outputs the extraction condition, the processing condition, and the destination information to the stream processing server device identified by the network identification information via the network A stream filter type graph setting method in a filter type graph setting device,
By the computer in the stream filter type graph setting device,
A procedure for storing in advance in the filter storage unit a filter associated with the network identification information and the processing in the stream processing server device identified by the network identification information;
A procedure for accumulating a filter graph in which the filters are combined in a graph storage unit;
Based on a filter graph read from the graph storage unit, a graph setting procedure for generating the extraction condition and the processing condition for each stream processing server device based on network identification information included in the filter;
A setting condition output procedure for outputting the extraction condition for each stream processing server device generated by the graph setting procedure and the processing condition to the stream processing server device corresponding to the network identification information via the network;
A stream filter type graph setting method characterized in that is performed. In the computer of the stream processing server device arranged on the network,
A stream input procedure for inputting stream data from the network;
An extraction condition for extracting predetermined stream data is set in advance, and a stream extraction procedure for extracting stream data that matches the extraction condition from the stream data input by the stream input procedure;
Processing conditions for processing stream data are set in advance, and processing procedures for processing the stream data extracted by the stream extraction procedure based on the processing conditions;
Destination information for identifying a destination to which stream data is transmitted is set in advance, and stream data input by the stream input procedure or stream data processed by the processing procedure is sent to the network based on the destination information. Stream output procedure to output,
A program for running Stream data extracted from the stream data input via the network based on the input extraction condition, the extracted stream data processed according to the input processing condition, and the processed stream A stream that outputs data to the network based on input destination information, and outputs the extraction condition, the processing condition, and the destination information to the stream processing server device identified by the network identification information via the network In the computer of the filter type graph setting device,
A procedure for storing in advance in the filter storage unit a filter associated with the network identification information and the processing in the stream processing server device identified by the network identification information;
A procedure for accumulating a filter graph in which the filters are combined in a graph storage unit;
Based on a filter graph read from the graph storage unit, a graph setting procedure for generating the extraction condition and the processing condition for each stream processing server device based on network identification information included in the filter;
A setting condition output procedure for outputting the extraction condition for each stream processing server device generated by the graph setting procedure and the processing condition to the stream processing server device corresponding to the network identification information via the network;
A program for running

Images