JP2011109563A - Device, method and program for checking work content - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a device for recording conditions of a worker using a wearable sensor group mounted to the worker, and checking the record against a record obtained by inputting work contents in a day by the worker. <P>SOLUTION: This device for checking a work content retains a distribution of feature quantities corresponding to work contents of a person being checked, and is configured to input feature quantities of sensor data representing conditions of the person being checked, and input work contents and start and end times of the work of the person being checked, and checks whether the retained distribution of the feature quantities corresponding to the input work contents matches the distribution of the feature quantities of the sensor data in the period from the start to the end of the work. <P>COPYRIGHT: (C)2011,JPO&INPIT

Description

  The present invention relates to an apparatus, a method, and a program for verifying an operator's work content using a sensor group including an acceleration sensor and the like.

  Conventionally, there is no means for a worker's manager to confirm the work status of an apartment manager or a delivery company other than viewing daily reports etc. that each worker inputs by himself / herself. In this case, it is impossible to objectively determine the work status of the worker, and it is not possible to determine whether or not the worker has performed the work that should be originally performed. If the worker's work state can be managed, it is possible to improve efficiency such as work procedure, work place, and worker arrangement, and a means for managing the worker's work state is desired.

  Until now, there is a technology for identifying human behavior using a wearable sensor. Specifically, the acceleration sensor is fixed at multiple locations on the body, and the data obtained from the acceleration sensor is separated by a certain time width (window), and the average value, standard deviation, energy, etc. of each window are featured. Based on these, learning is performed and estimation is performed. (For example, see Patent Documents 1 and 2)

  In Patent Literature 1, a subject's behavior (walking, sitting, standing, boarding a vehicle) is identified using a three-axis acceleration sensor (X, Y, Z) mounted on a mobile phone. Specifically, the current behavior of the subject is identified by comparing the motion acceleration obtained from the mobile phone device with a reference behavior pattern of each behavior registered in advance.

  In Patent Document 2, first, the posture of the subject is identified before identifying the behavior of the subject. The posture of the subject is assumed to be three states of standing, sitting, and walking. The posture is identified, and actions that can be taken in the identified posture (for example, standing, climbing the elevator, descending the elevator) are identified. ing. At this time, sensors are installed at multiple locations on the body, and it is clarified which sensor is effective for identifying which posture and action, and the sensor used for identification is appropriate. By selecting this, the identification rate is improved.

There is also a technique for identifying a human movement state. (For example, refer nonpatent literature 1).
In Non-Patent Document 1, in addition to dividing time-series information obtained from an acceleration sensor with a window as described above, a moving state is estimated using information on a microphone and GPS. By using a microphone, GPS, or the like, the environment and position information around the subject can be obtained, and the estimation accuracy is improved. For example, if there is GPS time-series information, the moving speed of the subject can be calculated, and it can be distinguished whether the subject is traveling by train, automobile, bicycle, or walking.

JP 2003-46630 A JP 2009-39466 A

"Shaka: User movement state estimation method using mobile phone" A. Kobayashi, Kengo Iwamoto, Satoshi Nishiyama IPSJ Journal Vol50, No1 199-208 (Jan 2009)

  In the conventional technology, as described above, data obtained from an acceleration sensor or the like is divided by a window of several seconds, a feature amount in the window is extracted, and an action is identified based on the extracted feature amount. Therefore, it is based on the premise that the extracted feature quantity is the same when a certain action is divided into multiple windows, or the action of the time divided by each window is identified, and the action for several seconds is identified. ing.

  Therefore, the behavior that is the object of identification is the behavior that consists of repeated operations with a short cycle. For example, walking, running, brushing teeth. Actions such as walking and running are simple actions, so that features clearly appear in the acceleration values generated by these actions, the feature quantities extracted from each window are the same, and the identification accuracy is high.

  However, there are few things that are subject to identification of complex actions that perform another action when performing a certain action, and the accuracy of the identification is low.

  For example, in the work of wiping the window of the apartment cleaning work, in addition to the work of wiping the window, multiple actions such as squeezing the rag, moving with a rag, replacing the water in the bucket are mixed is doing. It is difficult to identify these complex behaviors.

  However, it is possible to test whether a certain action has actually been performed. For example, with respect to a daily report created by a worker, each work described in the daily report is verified whether or not the work was performed at that time. This does not require identification of the work content, but is specialized for testing.

  If the work contents can be verified, it is possible to find a declaration including the worker's falsehood, and to expect an increase in work efficiency.

  Therefore, the present invention records a worker's state using a wearable sensor group attached to the worker, and an apparatus, a program, and a program for verifying the record and a record of the worker's daily work content input Aims to provide a method.

  In order to achieve the above object, a work content verification apparatus according to the present invention inputs a first input means for inputting a feature amount of sensor data indicating a state of a subject, the work content of the subject, and a start / end time of the work. A second input means; a holding means for holding a distribution of feature values corresponding to the work content of the subject; a feature value distribution corresponding to the input work content from the holding means; Means for taking out a distribution of feature quantities in the time from the start to the end of the work from the feature quantities, and testing whether the distributions of both feature quantities match.

  In addition, the first input means is one or more sensors that are attached to the subject and measures the sensor data, means for collecting the sensor data from the one or more sensors, and the sensor data It is also preferable to input the feature amount of the sensor data from a sensor data measuring device provided with means for extracting the feature amount of the sensor and means for transmitting the feature amount to the input means.

  The feature amount is preferably an average, variance, or energy of the sensor data in a predetermined time interval.

  The sensor is preferably an acceleration sensor, a geomagnetic sensor, a gyro sensor, an atmospheric pressure sensor, a GPS, or a microphone.

  In order to achieve the above object, a method for validating work content according to the present invention includes a first input step of inputting a feature amount of sensor data indicating the state of a subject, the work content of the subject, and the start / end time of the work. A second input step of inputting, a holding step of holding a distribution of feature values corresponding to the work content of the subject, and a distribution of feature values corresponding to the input work content from the held distribution of feature values And taking out the distribution of the feature quantity in the time from the start to the end of the work from the feature quantity of the sensor data, and testing whether the distribution of both feature quantities is suitable.

  In order to achieve the above object, a program for examining work contents according to the present invention comprises: first input means for inputting a feature amount of sensor data indicating a state of a subject; work contents of the subject; and start and end times of the work. A second input means for inputting; a holding means for holding a distribution of feature values corresponding to the work content of the subject; a feature value distribution corresponding to the input work content from the holding means; A feature distribution in the time from the start to the end of the work is taken out from the feature amount of the data, and the computer is made to function as a means for testing whether the distribution of both feature amounts matches.

  According to the apparatus, program, and method of the present invention, the manager of the worker can objectively determine the work status of the worker, and can determine whether the worker has performed the work that should be originally performed. It is possible to improve the work efficiency of workers.

The operation | movement outline | summary of the system which test | inspects the work content in this embodiment is shown. The structure of the sensor data measuring device in this embodiment is shown. The structure of the work content verification apparatus in this embodiment is shown. The result of having compared the frequency distribution of the dispersion | distribution of the synthetic acceleration data of the acceleration data acquired with the triaxial accelerometer fixed to the cleaner's waist for every operation | work is shown.

  The best mode for carrying out the present invention will be described in detail below with reference to the drawings. FIG. 1 shows an outline of the operation of a system for examining work contents in the present embodiment.

  According to the present embodiment, the system for validating work content includes a sensor data measuring apparatus 1 that is worn by a worker, extracts a feature amount of the work content of the worker, and transmits the feature amount. The daily work report preparation device 2 having a function of inputting the work content of the job and transmitting the input content to a predetermined server, the sensor data of each worker and the daily work report of each worker are compared, and the daily work report is reported And a verification result display device 4 for displaying an inquiry about the work content and an inquiry result.

  FIG. 2 shows a configuration of the sensor data measuring device in the present embodiment. The sensor data measuring device 1 is a small device that is built in a mobile phone or the like and can be worn by a worker. The sensor data measuring device 1 includes a wearable sensor group 11 such as an acceleration sensor, a function 12 for storing sensor data acquired by the wearable sensor group, and a function for extracting a predetermined feature amount from the sensor data and storing the result. 13 and a function 14 for transmitting feature values acquired wirelessly or by wire to a predetermined server.

  FIG. 3 shows the configuration of the work content verification apparatus in this embodiment. The work content verification device 3 has a function 31 for receiving feature amounts from the sensor data measuring device 1, a function 31 for receiving daily work reports from the daily work report creation device 2, a function 32 for storing feature amounts and daily work reports, and an inquiry from the display device 4. , The function 33 for transmitting and receiving the inquiry result, the function 34 for maintaining the distribution of the feature amount for each work content of the worker that has been measured in advance by a preliminary experiment, and the distribution of the measured feature amount A function 35 is provided for comparing the distribution of feature values corresponding to the work content of the daily work report and verifying the report content of the daily work report.

  The sensor data measuring device 1 operates as follows. The sensor data measuring device 1 is attached to the body of a subject, and each sensor data is obtained from the wearable sensor group 11. Examples of the target sensor include an acceleration sensor, a geomagnetic sensor, a gyro sensor, an atmospheric pressure sensor, a GPS, and a microphone. Each sensor automatically collects data periodically once activated, and the data is automatically stored in the sensor data storage function 12 for recording sensor data. Since the size of the acquired sensor data is large, the feature amount of the sensor data (the feature amount will be described in detail below) is extracted. After the extraction, the extraction result is stored in the feature amount extraction storage function 13 and the extraction result is transmitted to the work content verification device 3. At this time, instead of automatically transmitting, the result may be transmitted by the subject to the server. This transmission can be performed not only by wireless or wired communication, but also by the subject recording the extraction result on a recording medium such as a memory card and transmitting it from the sensor data measuring device 1 to the work content verification device 3.

  The feature amount of the sensor data is extracted as follows. The time-series data (for example, acceleration data) measured by the wearable sensor group is divided into sliding windows (windows) for a predetermined time, and the feature amount is obtained for each window. This window has a certain time interval (window time) and is shifted at a certain time interval (window shifting time). Note that the windows may overlap. The window time and the window shifting time are determined by conducting a preliminary experiment. The average value, variance value, energy, etc. of the measurement data in this window are extracted as feature quantities. The energy is obtained by performing FFT (Fast Fourier Transform) on the measurement data in the window and squaring the amplitude.

  Note that what is identified in the present invention is an operation in which a plurality of operations are mixed, and is not a repetition of a simple operation. Therefore, the identification performance cannot be improved by using feature values such as an average value, a variance value, and energy in each window as feature values of data obtained from an acceleration sensor or the like. This is because it is unlikely that the average value and the variance value in each window show the same tendency because a plurality of operations are mixed. However, the distribution of feature values from the start to the end of a certain work may show a similar tendency depending on the work.

  FIG. 4 shows a result of comparing the frequency distribution of the dispersion of the combined acceleration data of the acceleration data acquired by the three-axis accelerometer fixed on the waist of a certain cleaning worker for each work. The operations being compared in FIG. 4 include window wiping with a rag (FIG. 4a), window wiping with a mop (FIG. 4b), wall wiping with a rag (FIG. 4c), object wiping with a rag (FIG. 4d), There are a total of six: vacuum cleaner (Fig. 4e) and break (Fig. 4f). For example, the distribution of the distribution of acceleration data generated by cleaning with a vacuum cleaner and the distribution of the distribution of acceleration data generated by taking a break are greatly different. This is because when the cleaner is used for cleaning, it moves around while the cleaner is applied, so that the change in the acceleration value increases, the variance value increases, and the distribution tends to spread. On the other hand, the data at the time of the break is less likely to move around at the time of the break, and the change in the acceleration value is small. Therefore, the distribution of the variance values tends not to be large.

The daily work report input in the daily work report creation apparatus 2 is performed as follows. Each subject prepares a daily work report after the work of one day is completed. The daily work report is performed by the daily work report creation apparatus 2 (for example, a portable terminal or a PC), and the input result of the daily work report is automatically transmitted to the work content verification apparatus 3. The contents to be entered in the daily work report at this time are:
1. Worker name (or unique ID for the worker)
2. 2. Work start time Work end time Work content 5. A work place is required. 2 to 5 are input for each work content.

  A method will be described in which the work content verification device 3 uses the feature values described above to verify the work content input in the daily work report and the results obtained from the acquired sensor data. First, for the work content input in the daily work report, a feature amount distribution (symbol) is prepared in advance for each work content. There may be a plurality of symbols for each work content. Symbols are created from the data obtained through preliminary experiments. This symbol can be updated each time worker data is accumulated. In this case, the accuracy of the test is improved.

  The work content verification device 3 divides the feature amount of the sensor data transmitted from the sensor data measuring device 1 by using the work start time and the work end time input by inputting the daily work report. The degree of matching between the distribution of the feature quantity of the divided sensor data and the symbol is used as an index for the test. This test is performed by an existing method such as a chi-square test.

  For example, in the case of 10: 00-10: 30 in the daily work report, “wiping the window with a rag” is described, the distribution of sensor data feature values from 10:00 to 10:30, and “wiping the window with a rag” The degree of fit with the symbol corresponding to is used as an index for the test. In this way, it can be verified whether the operator has actually wiped the window with a rag.

  The operator's manager can confirm the test result. The manager can confirm the test result by inquiring from the test result display device 4 to the work content test device 3 for each work content recorded in the daily report of the worker being managed. Further, the test result can be automatically sent to the predetermined test result display device 4 by the work content verification device 3. An administrator can manage the work status of each worker by using the work content display device 3.

  Moreover, all the embodiments described above are illustrative of the present invention and are not intended to limit the present invention, and the present invention can be implemented in other various modifications and changes. Therefore, the scope of the present invention is defined only by the claims and their equivalents.

DESCRIPTION OF SYMBOLS 1 Sensor data measuring device 2 Work daily report production apparatus 3 Work content verification apparatus 4 Test result display apparatus 11 Wearable sensor group 12 Sensor data storage function 13 Feature quantity extraction storage function 14 Feature quantity transmission function 31 Feature quantity and daily work report reception function 32 Feature Quantity and daily work storage function 33 Inquiry transmission / reception function 34 Feature quantity distribution holding means 35 Verification function

Claims (6)

First input means for inputting a feature amount of sensor data indicating the state of the subject;
Second input means for inputting the work content of the subject and the start / end time of the work;
Holding means for holding a distribution of feature values corresponding to the work content of the subject;
The distribution of feature amounts corresponding to the input work content is extracted from the holding unit, the distribution of feature amounts in the time from the start to the end of the operation is extracted from the feature amount of the sensor data, and the distribution of both feature amounts is A means of testing for conformity;
A work content verification device characterized by comprising: The first input means includes
One or more sensors mounted on the subject and measuring the sensor data;
Means for collecting the sensor data from the one or more sensors;
Means for extracting feature quantities of the sensor data;
Means for transmitting the feature quantity to the input means;
The work content verification device according to claim 1, wherein a feature amount of the sensor data is input from a sensor data measuring device comprising:   The work content verification apparatus according to claim 1, wherein the feature amount is an average, variance, or energy of the sensor data at a predetermined time interval.   The work content verification device according to claim 1, wherein the sensor is an acceleration sensor, a geomagnetic sensor, a gyro sensor, an atmospheric pressure sensor, a GPS, or a microphone. A first input step for inputting a feature amount of sensor data indicating the state of the subject;
A second input step of inputting the work content of the subject and the start / end time of the work;
A holding step for holding a distribution of feature values corresponding to the work content of the subject;
A feature amount distribution corresponding to the input work content is extracted from the held feature amount distribution, a feature amount distribution in the time from the start to the end of the operation is extracted from the feature amount of the sensor data, and both Testing whether the distribution of features is compatible;
A method for validating work contents characterized by having First input means for inputting a feature amount of sensor data indicating the state of the subject;
Second input means for inputting the work content of the subject and the start / end time of the work;
Holding means for holding a distribution of feature values corresponding to the work content of the subject;
The distribution of feature amounts corresponding to the input work content is extracted from the holding unit, the distribution of feature amounts in the time from the start to the end of the operation is extracted from the feature amount of the sensor data, and the distribution of both feature amounts is A means of testing for conformity;
A program that verifies work content, which is characterized by making a computer function.

Images