JP2011045460A - Watching robot, watching method, and watching program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a watching robot, a watching method and a watching program, capable of grasping a situation and an operation of a user, by working actively on the user, and capable of enhancing precision of recognition of the grasped situation and operation. <P>SOLUTION: This watching robot 10 includes a reference preparing part 14 for preparing one reference of associating body information acquired from a user information input part 13 for acquiring the body information of the user, with an intention or a state of the user, and the reference preparing part 14 includes a user recognition part 17 for analyzing the intention or the state, based on the acquired body information, a rule generation part 18 for confirming an analytical result analyzed by the user recognition part 17, to the user, to formalize the intention or the state, a rule collection part 20 for collecting other reference expressing an intention or a state of other user having an attribute same or similar to that of the user, and a data mining part 21 for extracting a feature of the other reference to correct the one reference. <P>COPYRIGHT: (C)2011,JPO&INPIT

Description

  The present invention relates to a watching robot that supports a user, a watching method, and a watching program.

  In recent years, with the rapid increase in the elderly, the number of single-living households and households with only the elderly has increased rapidly, and the need for nursing care in the area (at home) has increased. In the past, helper dispatching by administrative agencies and private volunteers has been conducted, but it is not easy to secure a sufficient labor force due to the nationwide shortage of personnel. In addition, it is difficult for home care at home to provide sufficient support due to changes in the family structure, support consciousness, women's advancement into the workplace, and changes in the local community, creating a heavy burden on both caregivers and non-caregivers. . Furthermore, if an accident occurs in the home of the elderly such as a fall / fall, drowning / drowning, suffocation, etc., it will be fatal to the elderly living alone.

  With such an increase in the number of care recipients centering on the elderly, various measures have been proposed. Under such circumstances, it is made from the viewpoints of insurance, medical care, welfare, etc., based on self-help efforts, in order not to cause a need for nursing care, and to prevent and reduce the progress of the need for nursing care. The concept of preventive care, which is a multifaceted approach, is spreading.

  The flow of care prevention includes (1) health promotion / lifestyle improvement, (2) lifestyle-related disease prevention / health checkup (examination), (3) early detection / early medical care of disease, (4) early rehabilitation / self-help independence, There is a view that (5) prevention of disuse symptoms and bedridden prevention are performed in this order. According to this, in order to prevent care, it is necessary to maintain a correct life rhythm, to be aware of one's own health condition, and to rehabilitate and maintain a motor function. And when an accident such as a fall occurs, it is necessary to watch over so that a quick response can be made.

  In order to realize such care prevention and watching without human intervention as much as possible, various methods have been proposed. For example, as shown in Patent Document 1, it has been proposed to perform user home health management via a robot capable of voice recognition and voice synthesis.

JP 2004-337556 A

  However, the above-described conventional robot recognizes a human situation and motion only from information obtained from a passive sensor. Therefore, it is difficult to recognize an accurate situation of a person and an intention of an action, and it is difficult to perform appropriate care prevention.

  The present invention has been made in view of the above circumstances, and not only from passive sensors but also by actively acting on the user, the user's situation and operation can be grasped, and the grasped situation and It is an object of the present invention to provide a watching robot, a watching method, and a watching program that can increase the accuracy of motion recognition.

The present invention employs the following means in order to solve the above problems.
The watching robot according to the present invention includes a reference creation unit that creates one reference for associating the physical information acquired from the user information input unit that acquires the user's physical information with the user's intention or state, and the reference creation unit A user recognition unit that analyzes the intention or state from the acquired physical information, a rule generation unit that confirms an analysis result by the user recognition unit and formalizes the intention or state, and is the same as the user or A rule collection unit that collects other criteria representing the intention or state of another user having similar attributes, and a data mining unit that extracts features of the other criteria and corrects the one criterion It is characterized by.

  The watching method according to the present invention is a user watching method using a robot, and creates an acquisition step for acquiring the user's physical information and a reference for associating the acquired physical information with the user's intention or state. A reference creation step, and a user recognition step for analyzing the intention or state from the acquired physical information, and a rule for confirming an analysis result with the user and formalizing the intention or state. A generating step, a rule collecting step for collecting other criteria representing the intention or state of another user having the same or similar attribute as the user, and a feature of the other criteria is extracted to correct one criterion A data mining step.

  Further, the watching program according to the present invention is a watching program for watching a user via a computer, wherein the computer acquires the physical information acquired from the user information input unit for acquiring the user's physical information and the user's intention or A user recognition unit that functions as a reference creation unit that creates one reference for associating states, and the reference creation unit analyzes the intention or state from the acquired physical information, and the analysis result by the user recognition unit is a user A rule generation unit that confirms and formalizes the intention or state, a rule collection unit that collects other criteria representing the intention or state of another user having the same or similar attribute as the user, and the other It is characterized by functioning as a data mining unit that extracts a reference feature and corrects the one reference.

  According to the present invention, not only from a passive sensor but also by actively acting on the user, it is possible to grasp the user's situation and action, and to improve the accuracy of recognition of the grasped situation and action. it can.

It is explanatory drawing which shows the inside of the room where the watching robot which concerns on one Embodiment of this invention is arranged. It is a functional block diagram which shows the watching robot which concerns on one Embodiment of this invention. It is a flowchart which shows the watching method which concerns on one Embodiment of this invention. It is explanatory drawing which shows the preparation content of the reference | standard in the watching robot which concerns on one Embodiment of this invention. It is explanatory drawing which shows the preparation method of the reference | standard in the watching robot which concerns on one Embodiment of this invention. It is explanatory drawing which shows the creation content of the reference | standard of the life pattern in the watching robot which concerns on one Embodiment of this invention. It is explanatory drawing which shows the user assistance point by the watching robot which concerns on one Embodiment of this invention. It is explanatory drawing which shows the other user assistance point by the watching robot which concerns on one Embodiment of this invention.

An embodiment according to the present invention will be described with reference to FIGS. 1 to 5.
As shown in FIG. 1, the watching robot 10 according to the present embodiment watches the user U in an intelligent room IR including a door D, a refrigerator R, a bed B, and a scale M.

  As shown in FIG. 2, the watching robot 10 includes a room server 11 installed in the intelligent room IR, a robot body 12 serving as an interface for communication with the user U, and the body of the user U in the intelligent room IR. And a user information input unit 13 for acquiring information.

  The room server 11 includes a ROM (read only memory) (not shown) in which programs and data for performing necessary processing are stored, and a RAM (random access memory) (not shown) for temporarily storing necessary data. ), A CPU (central processing unit) (not shown) that performs processing according to a program stored in a ROM or the like is incorporated.

  The room server 11 includes a reference creation unit 14, a server-side operation processing unit (operation processing unit) 15, and a reference accumulation unit 16 as functional means (program modules). Then, it is connected to an external storage server SS via the Internet W.

  The reference creation unit 14 creates a reference R1 that associates physical information acquired from the user information input unit 13 to be described later with the intention or state of the user U, so that the user recognition unit 17, the rule generation unit 18, A collection unit 20, a data mining unit 21, and a pattern mining unit 22 are provided.

  The user recognition unit 17 includes a gesture recognition unit 23, a health evaluation unit 25, and a personal identification unit 26, and analyzes the intention or state from the acquired physical information. For example, as described in Japanese Patent Application No. 2007-201447, the gesture recognition unit 23 analyzes the image of the face and hand of the user U captured by the camera 32 described later, and determines the orientation and pointing direction of the user U. calculate. The health evaluation unit 25 performs pathological diagnosis or stress evaluation from the biological information of the user U detected by the biological information measurement unit 35 described later, as described in, for example, Japanese Patent Application Laid-Open No. 2004-073520 or Japanese Patent Application No. 2007-234725. I do. The personal identification unit 26 performs personal authentication of the user U by a known technique based on personal information obtained from the personal data acquisition unit 36 to be described later, and identifies personal attributes of the user U stored in advance.

  The rule generation unit 18 confirms the current intention or situation with the user U via the robot body 12, and formats the analysis result by the user recognition unit 17 as one reference R1 representing the intention or state. The rule collection unit 20 is configured so that other criteria indicating the intention or state of other users having the same or similar attributes as those of the user U stored in the storage server SS via the Internet W, and the lives of other users including them Collect patterns and behavior patterns.

  The data mining unit 21 performs a mining process using an analysis algorithm, extracts the characteristics of other collected standards, and corrects one standard R1. The pattern mining unit 22 has the same configuration as that of the data mining unit 21 and extracts an optimal lifestyle pattern for the attribute of the user U from the collected lifestyle patterns and behavior patterns.

  The server-side operation processing unit 15 generates information to be presented to the user U based on the created one reference R1 health condition, personal attribute information, and the user U life time in response to the intention and movement of the presented user U. An operation pattern to be transmitted through the operation of the robot body 12 is determined. The reference accumulation unit 16 stores the created reference R1.

  The robot body 12 is transmitted through an intuitive movement with respect to the user U, a robot-side motion processing unit 28 to which the motion processing determined by the server-side motion processing unit 15 is transmitted, a voice output unit 30 that outputs voice. And an operation unit 31 that performs a gesture operation. The robot body 12 is connected to the room server 11 via a wireless LAN or the like.

  The user information input unit 13 includes a camera 32 that acquires an image including the face of the user U, a microphone 33 to which the user U's voice is input, and a biological information measurement unit 35 that detects information such as blood pressure, body temperature, and pulse wave. A personal data acquisition unit 36 for detecting attribute information embedded in an IC tag or the like for specifying an individual, and an open / close sensor 37 disposed in the door D and the refrigerator R in the intelligent room IR. The camera 32, the microphone 33, the biological information measurement unit 35, and the personal data acquisition unit 36 are arranged in the robot body 12. These may be arranged in the intelligent room IR. The open / close sensor 37 constitutes a network and is connected to the user recognition unit 17 so as to obtain the open / close information of the door D and the refrigerator R.

Next, a watching method by the watching robot 10 according to the present embodiment will be described.
For example, when the user U falls, the watching method according to the present embodiment includes an acquisition step (S11) and a reference creation step (S12) as shown in FIG. The reference creation step (S12) includes a user recognition step (S121), a rule generation step (S122), a rule collection step (S123), a data mining step (S124), a pattern mining step (S125), It has.

  In the acquisition step (S11), the operation of the user U in the intelligent room IR is detected via the user information input unit 13. As shown in FIGS. 4 and 5, the robot body 12 captures an image of the user U falling over with the camera 32.

  Then, the process proceeds to a reference creation step (S12). In the user recognition step (S121), the gesture recognition unit 23 analyzes the image of the face and hand of the user U as described in Japanese Patent Application No. 2007-201447, for example. And, for example, the height direction when standing is the Z direction, the head position is 1, the direction when the hand is spread is the X direction or the Y direction, the hand position is 1, and from the standing state to the falling state The relative positions along the flow line are coordinated in the XYZ space by a known method.

  Next, in the rule generation step (S122), for example, the inflection point and its time of the coordinated flow line, and the default standard indicating the fall similarly stored in the rule collection unit 20 in advance are used. When it is the same or similar, it is temporarily determined that the state of the user U is a fallen state. Then, it is confirmed from the robot main body 12 to the user U via the server side operation processing unit 15 and the robot side operation processing unit 28. At this time, if the response from the user U is “No” or there is no response, it is determined that the user U has actually fallen, and one reference R1 indicating the fall is created.

  In the rule collecting step (S123), the one reference R1 is accumulated, and the rule collecting unit 20 collects the standards relating to the fall of another user having the same or similar attribute as the user U from the accumulation server SS. . In the data mining step (S124), based on a known data processing technique, the standard is standardized and optimized, and the standard R1 is corrected. The reference R1 obtained in this way is used as a new reference R1 for fall determination, and is shared with the storage server SS.

  Then, as shown in FIG. 6, for example, each intention or state over the daily life pattern of the user U such as getting up, watching TV, bathing, wandering, falling, and measuring the health state with the biological information measuring unit 35. We will create standards for recognizing At this time, further, in the pattern mining step (S125), the life patterns of other users having the same or similar attributes are collected from the storage server SS, and a correlation rule that is complicatedly generated by clustering based on a known technique is created. Discover.

  After the reference is created, various support is provided for the life pattern of the user U. For example, FIG. 7 shows a case where another user having the same or similar attribute as the user U can recognize that the health is maintained by exercising during a certain time of day as an association rule. Thus, the case where the life pattern is recommended to the user U will be described.

  In this case, based on the above-described method, a standard for wake-up time, exercise time, and bedtime of a healthy person is created. When the wake-up time is reached, the robot body 12 prompts the user to wake up based on the reference. Similarly, when the exercise time is reached, the robot body 12 prompts the user to exercise, and when the bedtime is reached, the robot body 12 is prompted to go to bed.

  In other words, based on the various standards that have been created, the server-side motion processing unit 15 performs processing related to calling for waking up, exercising, and sleeping, and is transmitted to the robot-side motion processing unit 28. On the robot body 12 side, the voice output unit 30 and the operation unit 31 output voices and operations calling for getting up or exercising. When the user U agrees to exercise, the intention of the user U is grasped by the microphone 33 or the like, and then, for the health evaluation of the user U, for example, the measurement by the biological information measuring unit 35 is promoted, and the health state before the exercise To figure out. Thereafter, the state in which the user U exercises is imaged by the camera 32, and the amount of exercise is measured via the user recognition unit 17.

  In addition, after acquiring personal information including hobbies and preferences as the attributes of the user U and specifying the personal attributes, a standard related to this is created in the same manner, and then the user U's hobbies and preferences A TV program, a stadium, a concert venue, or the like that matches may be searched from the WEB and notified to the user U.

  For example, even if the daily life pattern is the same, each behavior pattern constituting the life pattern may differ depending on attributes such as hobbies and preferences. In such a case, for example, a case where the cleaning robot 40 cleans up tableware will be described with reference to FIG.

  In this case, a reference is created for the order of the tableware cleaning operation in the reference creation step (S12) described above. At this time, first, the tableware is cleared based on the default standard. Here, if the user U has an attribute that he likes tea, for example, and he wants the tea utensils to be cleaned at the end in order to enjoy tea slowly after eating, the user U can cancel the cleanup in the middle of cleaning up. Express instructions or intentions. When this is recognized by the user recognition unit 17, in the rule collection step (S123), a reference for cleaning up other users having the same or similar attributes as the user U is collected. And when a correlation rule is discovered in a pattern mining step (S125), the reference | standard regarding a tidying-up order is corrected into a new thing.

  According to the watching robot 10, the watching method, and the watching program, the user can obtain the physical information related to the intention or the state of the user U when creating the standard for judging the correctness of the recognition of the intention or the state of the user U. It is possible to confirm with U the intention or state and form it as one standard R1, and to create a more accurate judgment standard. At this time, since criteria that indicate the same or similar intentions or states of others are collected and common features are extracted and fed back, the accuracy of the creation criteria can be improved.

  Moreover, based on the created criteria regarding the intention or state, the user U can recognize an accurate daily life pattern, and by comparing with the life pattern obtained from the same or similar others, A life pattern optimal for the attribute can be presented to the user U to support life.

  At this time, not only the lifestyle pattern but also each behavior pattern in the lifestyle pattern can be presented with the optimum user U attribute.

The technical scope of the present invention is not limited to the above embodiment, and various modifications can be made without departing from the spirit of the present invention.
For example, when the user U falls, as shown in FIG. 1, the room server 11 determines that an abnormality has occurred under a predetermined condition, and notifies a third party authorized in advance by an email EM or the like. May be. Moreover, although the watching robot 10 is provided with the room server 11 and the robot main body 12 individually, it is not limited to this and both may be integrated.

  In addition, the open / close sensor 37 is arranged in the door D and the refrigerator R, but may be arranged in other things, for example, divided into partitions to classify medicines to be taken daily, It may be arranged in a medicine box (not shown) arranged in the intelligent room IR. As a result, the user U obtains information that the medicine is taken out of the medicine box and drinks it through the user information input unit 13, and the same monitoring method is applied to such an operation by applying the monitoring method described above. A reference can be created, and further life support for the user U can be provided.

10 watching robot 13 user information input unit 14 reference creation unit 15 server side motion processing unit (motion processing unit)
17 User recognition unit 18 Rule generation unit 20 Rule collection unit 21 Data mining unit 22 Pattern mining unit R1 Reference U user

Claims (12)

A reference creation unit for creating one reference for associating the physical information acquired from the user information input unit for acquiring the user's physical information with the user's intention or state;
The reference creation unit
A user recognition unit that analyzes the intention or state from the acquired physical information;
A rule generation unit for confirming the analysis result by the user recognition unit and formalizing the intention or state;
A rule collection unit that collects other criteria representing the intention or state of another user having the same or similar attribute as the user;
A data mining unit that extracts features of the other criteria and corrects the one criterion;
Watching robot characterized by having The reference creation unit includes a pattern mining unit that extracts a life pattern optimal for the attribute from a life pattern including the reference of another user having an attribute that is the same as or similar to the attribute of the user,
The watching robot according to claim 1, further comprising an operation processing unit that presents the extracted life pattern to a user. The pattern mining unit extracts an action pattern optimal for the attribute for each of a plurality of action patterns constituting the life pattern,
The watching robot according to claim 2, wherein the motion processing unit presents the optimum behavior pattern to a user.   The watching robot according to claim 1, wherein the rule collection unit collects the other reference stored in a storage server connected via the Internet. A method for watching a user with a robot,
An acquisition step of acquiring physical information of the user;
A standard creation step of creating a standard for associating the acquired physical information with the user's intention or state;
With
The reference creation step includes
A user recognition step of analyzing the intention or state from the acquired physical information;
A rule generation step of confirming the analysis result with the user and formalizing the intention or state;
A rule collection step for collecting other criteria representing the intention or state of another user having the same or similar attributes as the user;
A data mining step of extracting features of the other criteria and correcting one criteria;
A monitoring method characterized by comprising: The standard is created for each user life pattern,
A pattern mining step in which the reference creating step extracts a life pattern optimal for the attribute from a life pattern including the reference of another user having an attribute that is the same as or similar to the attribute of the user;
An operation processing step of presenting the extracted life pattern to the user;
The monitoring method according to claim 6, further comprising: The pattern mining step extracts an optimum behavior pattern for the attribute for each of a plurality of behavior patterns constituting the life pattern,
The watching method according to claim 7, wherein the operation processing unit presents the optimum behavior pattern to a user.   The monitoring method according to claim 5, wherein the rule collecting step collects the other reference stored in a storage server connected via the Internet. A watching program for watching users via a computer,
Computer
Function as a reference creation unit for creating one reference for associating the physical information acquired from the user information input unit for acquiring the user's physical information with the user's intention or state,
The reference creation unit
A user recognition unit that analyzes the intention or state from the acquired physical information;
A rule generation unit for confirming the analysis result by the user recognition unit and formalizing the intention or state;
A rule collection unit that collects other criteria representing the intention or state of another user having the same or similar attribute as the user;
A data mining unit that extracts features of the other criteria and corrects the one criterion;
Watch program characterized by functioning as The standard is created for each user life pattern,
A pattern mining unit that extracts a life pattern optimal for the attribute from a life pattern including the reference of another user having the same or similar attribute as the user attribute;
An operation processing unit for presenting the extracted life pattern to the user;
The monitoring program according to claim 9, comprising: The pattern mining unit extracts an action pattern optimal for the attribute for each of a plurality of action patterns constituting the life pattern,
The watching program according to claim 10, wherein the operation processing unit presents the optimum behavior pattern to a user. The monitoring program according to claim 9, wherein the rule collection unit collects the other reference stored in a storage server connected via the Internet.

Images