KR100981838B1 - System and method for distribution-based human recognition using message transfer with bit correct conversion - Google Patents

Abstract

A distributed treatment based human recognition system and method are disclosed. In a distributed processing-based human recognition system using a bit correction conversion message exchange method, a data input unit for receiving human recognition data for human recognition, generating and transmitting bit correction conversion data for distributed processing of the human recognition data, and performing distributed processing And a central control unit for receiving and restoring the resultant result data, and a distributed operation processor for restoring the bit correction conversion data to perform distributed processing, and performing bit correction conversion on the distributed result data.

Human recognition, distributed processing, message transmission, pattern preprocessing

Description

System and method for human recognition based on distributed processing using bit-correction conversion message exchange method {SYSTEM AND METHOD FOR DISTRIBUTION-BASED HUMAN RECOGNITION USING MESSAGE TRANSFER WITH BIT CORRECT CONVERSION}

The present invention relates to a distributed processing-based human recognition system and method, and more particularly, to a human recognition system and method for increasing the efficiency of message exchange between processors by reducing the size of data to be transmitted using a bit-correction conversion message exchange method. .

In general, distributed processing is a method of connecting multiple processors through a network to use its data processing power, and is capable of using powerful data processing power by connecting computing power of independent processors in a geographically dispersed area to a network. .

 In addition, the distributed processing is performed by distributing the work to be processed according to the data processing capability of the connected processor, and may improve the processing speed of the data by collecting the finished results again, Can also perform tasks.

In addition, in the process of distributing work and receiving the processed results, data is transferred using a network connected through message exchange between processors. In this process, the network to which messages travel is heterogeneous according to each environment. It is a commercial network that many people use for various purposes.

 In addition, due to the heterogeneous commercial network, message exchange is affected by the ever-changing network situation, and also affects the network environment in use. Therefore, the size of the message is composed in the field using a large amount of pattern information data. As the network load increases, the time required for exchanging messages between processors increases, which acts as an obstacle to smooth message exchange between processors.

Accordingly, what is needed is a human identification system and method for reducing the size of messages for smooth message exchange between processors.

The present invention provides a distributed processing-based human recognition system and method capable of efficiently exchanging messages between processors by exchanging messages by reducing the size of data to be transmitted using a bit correction conversion message exchange method.

The present invention provides a distributed processing-based human identification system and method for reducing the exchange time of a message by exchanging messages by reducing the size of data to be transmitted using a bit correction conversion message exchange method.

The present invention reduces the size of data to be transmitted using the bit-correction conversion message exchange method, thereby exchanging messages, thereby reducing the load on the network and reducing the time required for transmission, and at the same time, minimizing the recognition rate and maintaining the pattern recognition rate. Provides a system and method.

According to an embodiment of the present invention, a distributed processing-based human recognition system includes a data input unit for receiving human recognition data for human recognition, generating and transmitting bit correction conversion data for distributed processing of the human recognition data, and performing distributed processing. A central control unit for receiving and restoring result data, and a distributed operation processor for restoring the bit correction conversion data to perform distributed processing, and performing bit correction conversion on the distributed result data.

According to one aspect of the invention, the central control unit, a data distribution analysis unit for analyzing the distribution of the data for human body recognition, a correction conversion length setting unit for setting the reduced length of the transmission data and the uniformity based on the reduced length It may include a data distribution alignment unit for aligning the distribution of data to maintain a difference.

In the distributed processing-based human recognition method according to an embodiment of the present invention, receiving human recognition data for human recognition, generating bit correction conversion data by performing bit correction conversion on the human recognition data, and converting the bit correction conversion data. Transmitting to the distributed work processor, performing the distributed processing by restoring the received bit correction conversion data, and performing bit correction conversion on the distributed processing result data to the central control unit, and receiving the bit correction conversion result data. Restoring.

In one embodiment of the present invention, by using a bit-correction conversion message exchange method by reducing the size of the data to be transmitted by exchanging messages, distributed processing-based human identification system capable of efficient exchange of messages between processors by reducing the load on the network and A method is provided.

In one embodiment of the present invention, there is provided a distributed processing-based human recognition system and method for reducing the exchange time of the message by exchanging messages by reducing the size of the data to be transmitted using the bit-correction conversion message exchange method.

In one embodiment of the present invention, by exchanging messages by reducing the size of data to be transmitted using a bit-correction conversion message exchange method, to reduce the load on the network and to reduce the time required for transmission, while reducing the recognition rate pattern Human recognition systems and methods are provided for maintaining recognition rates.

Hereinafter, with reference to the contents described in the accompanying drawings will be described in detail an embodiment according to the present invention. However, the present invention is not limited to or limited by the embodiments. Like reference numerals in the drawings denote like elements.

1 is a block diagram illustrating a human body recognition system based on distributed processing according to an embodiment of the present invention.

Referring to FIG. 1, the dispersion-based human body recognition device includes a data input unit 120, a central controller 130, and a distributed work processor 140.

The data input unit 120 receives the human body recognition data for the human body recognition from the user 110.

In this case, the data input unit 120 may receive data required for human body recognition, request a processing thereof, and output a result of the processed human body recognition. In addition, the data input unit 120 is an input device for receiving a request for a human body recognition task and information required for human body recognition from the user, a recognition task requester and a human body requesting the human body recognition operation based on the received information and receive the result It may include an output unit for outputting the result of the recognition to the user.

Here, the input unit may receive a request for a human body recognition task from a user, and may receive human body recognition task information including a range, details, and bit lengths to be reduced through bit correction conversion to be used together.

At this time, if the user does not input any other information to the input device, the human body recognition work information may be arbitrarily selected in the range and details of the human body information, and in the case of the bit length to be reduced by bit correction conversion, the data is analyzed in the transmitting step. It is possible to use the most suitable bit length.

In addition, the recognition task requester transmits the information on the range and details of the human body information to be used based on the information input through the input device and the information about the bit length to be reduced through the bit correction conversion to the central control unit 130, The central control unit 130 receives the result of the calculation and processing is completed. Here, the result of the operation and processing is characterized in that it includes the total time required to perform the operation and processing and the recognition rate for the human body information.

The output unit may output a result received from the recognition task requester to the user.

The central control unit 130 generates and transmits the bit correction conversion data for the distributed processing of the human body recognition data, and receives and restores the distributed data.

In this case, the central controller 130 may control operations and processes performed for human recognition, allocate tasks according to required operations and processes, and collect and manage the results generated while performing the operations and processes. Detailed configuration of the central control unit 130 in one embodiment will be described in more detail with reference to FIG.

2 is a diagram showing the detailed configuration of the central control unit 130 in one embodiment of the present invention. Referring to FIG. 2, the central controller 130 may include a data distribution analyzer 210, a compensation conversion length setting unit 220, and a data distribution alignment unit 230.

The data distribution analyzer 210 may generate bit correction conversion information by analyzing a range in which actual data values are distributed in the data to be transmitted and a bit length suitable for the range. That is, the data distribution analyzer 210 determines the distribution of the actual data values, analyzes the most suitable bit length for expressing the distribution of the actual data values, and analyzes the data and analysis information in the data distribution alignment unit 230. To pass.

In this case, the data analysis information may include the smallest value of the data, the difference between the largest value and the smallest value, and a bit length most suitable for representing the analyzed data based on the data so as to express the distribution range of the actual data.

In this case, the analyzed bit length included in the data analysis information is replaced with a value representing a difference in the actual data value represented by the converted bit 1 when the bit length is expressed, which is calculated through Equation 1. Can be.

[Equation 1]

Where range represents the difference in the actual data value represented by converted bit 1, min means the smallest value in the actual data value, max means the largest value in the actual data value, and reduction bit length indicates the analysis. It represents the bit length to be reduced through bit correction conversion designated by the user who received the input through the user or through the user's input, and the difference in the actual data value represented by bit 1 can be obtained by the simple arithmetic operation as above.

In addition, if the user designates a separate bit length, the process of analyzing the most suitable bit length in the data distribution analyzer 210 is omitted, and the bit length specified by the user may use the most suitable bit length in the data analysis information. Can be.

The correction conversion length setting unit 220 may set the correction conversion length using the bit length input from the user or the bit length determined through analysis. That is, when the user specifies the bit length to be applied through input, the corresponding bit length is used. When the bit length is not specified separately, the bit length determined through the analysis of the data distribution analyzer 210 is used. The compensation conversion length can be set.

The data distribution alignment unit 230 may make the distribution of the data constant so that the bit correction may be performed according to the appropriate bit length according to the analyzed information.

That is, the data distribution arranging unit 230 may receive data and data analysis information about the data from the data distribution analyzing unit 210, and evenly arrange the distribution of actual data values before performing the bit correction conversion. .

Accordingly, after the data distribution analysis and the data distribution alignment, the bit correction transformation may be performed, and the bit correction transformation may be calculated through Equation 2.

[Equation 2]

In this case, the reduction bit data is a value of data converted through bit compensation conversion, and the data value is a value of actual data to perform current bit compensation conversion, and min is the smallest within a range where the actual data values are distributed. A value represents a value, and a range represents a value representing a difference in actual data value represented by the converted bit 1 and corresponds to a result calculated by Equation 1 above.

In addition, reconstruction for receiving the distributed data and returning the converted data having the reduced bit length to the bit length before the bit compensation conversion may be performed by Equation 3 below.

&Quot; (3) "

In this case, the restoration bit data represents data whose bit length is restored, the reduction bit data is a value of data converted through bit compensation conversion, and the range is a value representing a difference in actual data value represented by the converted bit 1. It corresponds to the result calculated by Equation 1, and min means the smallest value within the range in which the actual data values are distributed.

In another embodiment of the present invention, the central controller 130 stores and manages face photo data used for face recognition and data generated in the process of processing and processing face recognition, and distributes work. It can perform the necessary operations and processing after receiving the results of previous or distributed tasks.

To this end, the central controller 130 may include a central data manager, a central computing device, a job manager, a data transmitter, and a data receiver.

Here, the central data manager stores and manages face photo data used for face recognition and data generated in the process of processing and processing face recognition, and provides data necessary for the current work. Can be done.

In addition, the central computing device is provided with the data necessary for the job to be performed currently from the central data manager, performs the operation and processing that must be performed before the job is distributed, the job to perform the job division for distributed processing Decentralized job information, face recognition setting information, and data are delivered to the manager, and after the data is processed by the distributed job manager, the distributed job information is received from the job manager. It can play a role of keeping it.

The task manager receives the distributed task information, the face recognition setting information and the data from the central computing device, and divides the data based on the face recognition setting information and the distributed task information. Sending data, collecting the data received from the data receiver to perform a combination of the data based on the face recognition setting information and the distributed operation information, and then deliver the data to the central computing device.

At this time, the face recognition setting information contains information on the number of the distributed work processing unit 140 and its processing capacity that are currently connected for distributed processing, and may receive the information once received first.

In addition, the distributed job information may include information on the size of data to be transmitted for the job, so that the job can be divided according to the number of the distributed job processing unit 140 for the current job.

Referring back to FIG. 1, the distributed task processor 140 performs the distributed processing by restoring the bit correction conversion data, and returns the result by performing bit correction conversion on the distributed result data. That is, operations and processing may be performed on the job assigned from the central controller 130 and the result may be returned.

That is, the distributed job processor 140 may perform a calculation and processing on a job assigned by the central controller 130 and return a result to the central controller 130. To this end, the distributed job processing unit 140 may include a distributed job data manager, a distributed job calculation processor, a data transmitter, and a data receiver.

Here, the distributed job data manager stores and manages data to be used in later calculations among the result data generated after performing calculation and processing on the data received or the assigned job during the job assignment in the central controller 130. Can be performed.

In addition, the distributed task processor performs calculation and processing according to the task to be performed with the data received from the data receiver, and performs further calculation among the data received from the data receiver and the result data obtained from the calculation and processing. The data to be used may be transferred to the distributed job data manager, and the result of the calculation and processing may be sent to the data transmitter for transmission.

In this case, since the data to be stored in the distributed work data manager is determined to perform operations and processing for face recognition, the distributed work computing processor may transfer the received data or data generated at a specific step to the distributed work data manager by itself. .

That is, the data transmitter is present in the central control unit 130 and the distributed work processor 140, respectively, and performs bit correction conversion on the data to be transmitted between the two devices, reduces the size of the data, and configures the message. It can play the role of transmitting.

In addition, the data receiver is also present in the central control unit 130 and the distributed job processing unit 140, respectively, after returning the bit correction conversion state to the state before the bit conversion for the data received between the two devices, It can play the role of delivering to the device to which it belongs.

As described above, in the distributed processing human body recognition system, by exchanging messages by reducing the size of the data to be delivered, it is possible to reduce the load on the network and shorten the exchange time of messages to enable efficient message exchange between the processors.

3 is a flowchart illustrating a human body recognition method based on distributed processing according to an embodiment of the present invention.

Referring to FIG. 3, in step S310, human body recognition data for human body recognition is input. In this case, the human body recognition data may include a bit length to be reduced through a range, details, and bit correction conversion of human body information.

In step S320, bit correction transformation is performed on the human body recognition data to generate bit correction transformation data, and the bit correction transformation data is transmitted to the distributed work processor.

That is, the human body recognition data before conversion increases the load and transmission time of the network due to the large size of the data, so that the data can be bit-corrected and transmitted to reduce the size of the data.

In step S330, the received bit correction conversion data is restored to perform distributed processing, and the bitwise converted result of the distributed processing is returned to the central controller.

That is, after receiving the bit correction conversion data, it is restored to the data before conversion, the data is distributed, and bit correction is converted again for transmission to the central control unit to return the result data.

In step S340, the bit correction result data is received and restored.

By using the above-mentioned human body recognition method, PCA using the result of operation using 64-bit data without bit compression and bit correction and 32 bit and 16 bit data with bit compression and bit correction in the covariance matrix calculation step. The results of measuring the recognition rate after performing all the algorithm processes are shown in Table 1 below.

TABLE 1

Compression & correction Recognition rate 16 bit compression 91.25% 32 bit compression 91.25% No compression (64 bit) 91.25%

In other words, the experimental results show that all three transmission methods recorded the same recognition rate of 91.25%, and it can be seen that reasonable recognition rate can be maintained even if the bit length is reduced through bit compression and correction expressing the data.

Accordingly, a human body recognition system and method for reducing the load on the network and reducing the time required for transmission and at the same time minimizing the recognition rate can be maintained.

4 is a flowchart illustrating a human body recognition method based on distributed processing according to another embodiment of the present invention.

Referring to FIG. 4, in operation S411, a user may set face recognition task information. At this time, if there is no separate user input, the range and details of the face information are randomly selected by the input device.In the case of the bit length to be reduced by bit correction conversion, the most suitable bit length is analyzed by analyzing the data in the transmitting step. Can be used.

In operation S412, a job for face recognition may be requested.

In operation S413, data may be imported based on the range and details of the face information received through the work request.

In step S414, when the job is distributed and performed, it is determined whether there is a job to be performed in advance, and in step S415, when there is a pre-computation, the central processing unit performs calculation and processing on the data. In step S416, if the process is not required, the data is transmitted to the distributed work processor 140 without going through a separate process.

In step S417, the distributed task processor 140 that receives the data performs calculation and processing according to the assigned task.

In step S418, it is determined whether the received data and the result data of the calculation and processing continue to be used later.

In the case of the data that is continuously used with respect to the received data and the result data of the calculation and processing in the step S419, the storage process is performed once, and in the case of the data that is not used in the future in the step S420 It can be transmitted to the central control unit 130 without going through a separate process.

In step S421, the central control unit 130 that receives the data determines whether a separate post-operation is necessary, and in step S422, if post-operation is required, an additional post-process is performed. ), If the post-work is not necessary, the data of the work can be stored without going through a separate process.

In the step S424, if the dispersion processing work for face recognition still remains, the above-described process is repeated, but if the final work is completed and the result is obtained, the face recognition may be immediately terminated.

5 is a flowchart illustrating a data transmission method according to an embodiment of the present invention.

Referring to FIG. 5, in step S510, a data transmitter of a device to which data is to be transmitted may analyze the range in which actual data values are distributed, and may make and record the information as data analysis information.

In operation S520, it may be determined whether a user setting bit is input.

In the case of setting the transmission bit by the user in advance in step S530, the bit length set by the user is determined. In step S540, when the user does not set, the data distribution is analyzed. Based on the information, the appropriate reduced bit length can be set.

In operation S550, the distribution of the data may be aligned to maintain a uniform difference in which the distribution of the data may be represented according to the determined bit length.

In step S560, bit correction transformation is performed on the data whose distribution is aligned.

In step S570, the data subjected to bit correction conversion and the data analysis information generated through the above-described process may be transmitted from the data transmitter of the device to be transmitted to the data receiver of the device receiving the transmission.

In operation S580, the data receiver receiving the data may perform a task of returning the reduced bit length to the state before the bit conversion based on the received data and the data analysis information.

As described above, by exchanging messages by reducing the size of data to be transmitted using the bit-correction conversion message exchange method, it is possible to reduce the load on the network and shorten the exchange time of messages, thereby enabling efficient message exchange between processors. It is possible to provide a base human recognition system and method.

In addition, by exchanging messages by reducing the size of the data to be transmitted, it is possible to provide a human body recognition system and method for maintaining the pattern recognition rate by reducing the load on the network, reducing the time required for transmission and at the same time minimizing the recognition rate.

As described above, the present invention has been described by way of limited embodiments and drawings, but the present invention is not limited to the above-described embodiments, which can be variously modified and modified by those skilled in the art to which the present invention pertains. Modifications are possible. Accordingly, the spirit of the present invention should be understood only by the claims set forth below, and all equivalent or equivalent modifications thereof will belong to the scope of the present invention.

1 is a block diagram illustrating a human body recognition system based on distributed processing according to an embodiment of the present invention.

2 is a diagram showing the detailed configuration of the central control unit in an embodiment of the present invention.

3 is a flowchart illustrating a human body recognition method based on distributed processing according to an embodiment of the present invention.

4 is a flowchart illustrating a human body recognition method based on distributed processing according to another embodiment of the present invention.

5 is a flowchart illustrating a data transmission method according to an embodiment of the present invention.

<Explanation of symbols for the main parts of the drawings>

120: data input unit 130: central control unit

140: distributed work processing unit 210: data distribution analysis unit

220: correction conversion length setting unit 230: data distribution alignment unit

Claims (5)

A data input unit for receiving human body recognition data for human body recognition; A data distribution analyzer for analyzing a distribution of data for human recognition; A correction conversion length setting unit for setting a correction conversion length which is a length value reduced when converting the bit correction conversion data based on the distribution analysis result of the data; A data distribution alignment unit to align the distribution of data to maintain a uniform difference based on the distribution analysis result of the data; And A bit correction conversion unit generating bit correction conversion data for distributed processing of the human recognition data by bit correcting and converting the data for human body recognition in which the distribution of data values is evenly arranged in the data distribution alignment unit; A transmission / reception unit for transmitting the bit correction conversion data to a distributed job processing unit and receiving the resultant data distributed from the distributed job processing unit; And A decompression unit for restoring the distributed result data; A distributed work processor which performs the distributed processing by restoring the bit correction conversion data received from the transceiver, and performs bit correction conversion on the resultant data of the distributed processing to return to the transceiver. Distributed treatment-based human recognition system comprising a. The method of claim 1, The bit correction conversion unit and the distributed job processing unit, Distributed processing-based human recognition system, characterized in that for performing the bit correction by the following [Equation 1] and [Equation 2]. [Equation 1]

Is a value representing the difference in actual data value represented by converted bit 1,

Is the smallest value within the range of the original data values,

Is the largest value within the range of the original data values,

Is the bit to be reduced by bit correction Means each, [Equation 2]

here,

Is the value of the data converted through bit correction,

Is the value of original data before performing bit correction,

Is the smallest value within the range of the original data values

Is a value representing the difference in the actual data value represented by converted bit 1 Respectively. The method of claim 1, The bit correction conversion unit or the distributed job processing unit, A distributed processing-based human recognition system according to Equation 3 below, restoring the bit correction conversion data or the received result data. &Quot; (3) &quot;

here,

Is the data whose bit length has been restored,

Is the value of the data converted through bit correction,

Is a value representing the difference in the actual data value represented by converted bit 1,

Is the smallest value within the range of the actual data values Respectively. delete Receiving human body recognition data for human body recognition; Analyzing a distribution of data for human body recognition; Setting a correction conversion length which is a length value reduced when converting the bit correction conversion data based on the distribution analysis result of the data; Arranging the distribution of data to maintain a uniform difference based on the distribution analysis result of the data; Generating bit correction conversion data for distributed processing of the human body recognition data by performing bit correction conversion on the data for human body recognition in which the distribution of data values is evenly arranged in the data distribution alignment unit; Transmitting the bit correction conversion data to a distributed task processor; Restoring the received bit correction conversion data to perform distributed processing, and performing bit correction conversion on the distributed result data to a central control unit; And Receiving and restoring the bit-corrected result data Distributed treatment-based human recognition method comprising a.

Images