JP4183645B2 - Conversation leader discriminating apparatus and conversation leader discriminating method - Google Patents

Description

  The present invention relates to a conversation leader discriminating apparatus and a conversation leader discriminating method, and more particularly to, for example, a conversation leader discriminating apparatus and a conversation leader discriminating apparatus for discriminating a conversation leader between two or more persons existing in a certain place. Regarding the method.

There has been no prior art on this type of conversation leader discriminating device. An example of the prior art to be approximated is disclosed in Patent Document 1. According to this Patent Document 1, a signal transmitted from a transmission source possessed by a visitor (visitor) is detected at each exhibition hall, thereby identifying the current location of the visitor and the history of the visit. The visitor's interest is estimated based on the specified position data and history data, and the exhibition information provided to the visitor is created based on the estimated interest.
JP-A-11-96230

  In this prior art, the leaders of conversations (conversation leaders) who introduce (explain) exhibits or answer questions are fixed at each exhibition hall. Therefore, the conversation state in the place (exhibition hall) where the conversation leader exists can be determined based on the voice of the conversation leader. For example, when a conversation leader frequently speaks, it can be determined that a one-way conversation such as introducing (explaining) an exhibit is in progress. When the conversation leader's voice is interrupted and a certain silent period (3 to 10 seconds) is detected, and then the conversation leader's voice is detected again, the conversation leader and the visitor are visited. It can be determined that a two-way conversation such as a question-and-answer session is performed with a person. Furthermore, when the voice of the conversation leader is interrupted for a long time (10 minutes or more), it can be determined that no conversation is being performed at the place.

  However, if the conversation leader changes with time, such as a meeting (including the so-called well-end meeting), it is natural that the conversation leader cannot be specified. Could not be easily determined.

  Therefore, a main object of the present invention is to provide a novel conversation leader discriminating apparatus and conversation leader discriminating method.

  Another object of the present invention is to provide a conversation leader discriminating apparatus and a conversation leader discriminating method capable of accurately discriminating a conversation leader.

Claim 1 is a conversation leader discriminating device for discriminating a leader of a conversation between two or more persons existing in a certain place, and a plurality of voices for collecting voices of each of the two or more persons Microphone, personal recognition means for individually recognizing a person present in the place, calculation means for calculating the amount of speech of each person recognized by the personal recognition means in a certain time zone based on the output of the microphone , calculation means Extraction means for extracting a person having an utterance amount exceeding a predetermined value from among the utterance amounts for each person calculated by the above, and when there are a plurality of persons extracted by the extraction means, the utterance for the extracted person The judging means for judging whether or not the utterance amount varies based on the difference in the amount, and when the judging means judges that there is no variation, it is judged that there is no conversation leader. When the variation is that there is determined by the determining means, it comprises a leader discriminating means for discriminating the speech of the person speaking the maximum amount as leader of the conversation, a conversational leader discriminating device.

In the invention of claim 1, the conversation leader discriminating apparatus discriminates a conversation leader between two or more persons existing in a certain place (for example, a conference room or an exhibition hall). A plurality of microphones are provided for collecting each person's voice. Also, the personal recognition means individually recognizes the person existing in the place. The calculation means calculates the utterance amount of each person recognized by the personal recognition means in a certain time zone based on the output of the microphone. Specifically, when there are a plurality of persons in a certain place, the utterance amount of each person in the same time zone is calculated. The extraction means extracts a person whose utterance amount exceeds a predetermined value from the calculated utterance amount. When there are a plurality of persons extracted by the extracting unit, the determining unit determines whether or not there is a variation in the amount of speech based on the difference in the amount of speech for the extracted person. The conversation leader discriminating means discriminates that the conversation leader does not exist when it is judged that there is no variation by the judgment means. The person with the largest utterance amount is determined as the conversation leader.

According to the first aspect of the present invention, the utterance amount in a certain time zone for all persons existing in a certain place is calculated, and when the utterance amount varies, the utterance amount of the person exceeding the predetermined value is calculated . since speech amount is determined highest person as leader of conversation in the leader of the conversation that put the corresponding time period of one field can be accurately determined.

The invention of claim 2 is dependent on claim 1, and the leader discriminating means further discriminates that there is no leader of the conversation when the utterance amount of all persons does not exceed a predetermined value .

In the invention of claim 2, the conversation leader discriminating means discriminates that there is no conversation leader when the utterance amount of all persons does not exceed a predetermined value (a certain threshold).

  In the invention of claim 2, not only the utterance amount is simply maximized, but also the person with the largest utterance amount among the utterance amounts exceeding a certain threshold is determined as the conversation leader. Can be determined. This is because it is considered that it is not appropriate to discriminate the conversation leader among the persons with a relatively small amount of utterances when compared with the entire time period.

The invention of claim 3 is a conversation leader discriminating method for discriminating a leader of a conversation between two or more persons existing in a certain place, and (a) a voice for each of the two or more persons is recorded. (B) Recognize each person present in the venue individually, (c) Based on the voice collected by step (a), each person recognized by step (b) in a certain time zone The amount of utterance is calculated, (d) out of the amount of utterance for each person calculated in step (c), a person having an utterance amount exceeding a predetermined value is extracted, and (e) extracted in step (d) When there are multiple people, determine the presence or absence of utterance variation based on the difference in utterance amount for the extracted person, and (f) when it is determined that there is no variation in step (e) , Determine that there is no conversation leader, When the can is present is determined, to determine the utterance of the person speaking the maximum amount as leader of the conversation, a conversational leader discrimination method.

Also in the third invention, similar to the first invention, the leader of the conversation at a certain time period of a certain field can be accurately determined.

  According to the present invention, the utterance amount in a certain time zone for all persons existing in a certain place is calculated, and the person having the largest utterance amount is determined as the conversation leader. Leaders can be identified.

  The above object, other objects, features and advantages of the present invention will become more apparent from the following detailed description of embodiments with reference to the drawings.

  Referring to FIG. 1, a conversation discrimination device 10 of this embodiment includes a computer 12 such as a PC or a workstation. A plurality of wearable sensors (hereinafter simply referred to as “sensors”) 14, a plurality of tag readers 16, and a database 18 are connected to the computer 12. However, the sensor 14 is connected to the computer 12 directly (or indirectly via a network) by wireless communication. The sensor 14 is a sensor mainly for detecting sound, and includes a microphone 152 such as a vocal cord vibration microphone, as will be described later. The microphone 152 is attached to the user's throat and collects the user's voice. However, a unidirectional microphone can also be used as the microphone 152.

  The tag reader 16 detects unique identification information transmitted from the tag 20 attached to the user, and inputs the detected identification information to the computer 12. Here, the tag 20 is a frequency tag or an IR tag, and a tag reader 16 corresponding to the type of tag to be used is provided. As will be described later, one sensor 14 and one tag 20 are attached to the user, and the sensor 14 and the tag 20 are associated one-to-one (fixedly). As will be described later, the database 18 manages the utterance information input from the sensor 14 and the sensor information input from the tag reader 16, and the conversation leader information (conversation leader information) determined by the computer 12. Manage (store) (see FIG. 6).

  Further, an output device 22 is connected to the conversation state determination device 10 as necessary. Strictly speaking, the output device 22 is connected to the computer 12. The output device 22 is a display such as a CRT or LCD.

  FIG. 2 is a block diagram showing a specific configuration of the sensor 14 shown in FIG. Referring to FIG. 2, sensor 14 includes a housing 140, and CPU 142 is provided inside housing 140. An A / D converter 144, a memory 146, a wireless LAN 148, and a clock circuit 150 are connected to the CPU 142. In addition, a microphone (voice cord vibration microphone) 152 disposed outside the housing 140 is connected to the A / D converter 144.

  As the memory 146, a semiconductor memory can be used, and a disk recording medium such as a hard disk, MD, MO, CD, and DVD can also be used.

  The sensor 14 having such a configuration is attached to the user, and the audio signal input from the microphone 152 is digitally converted by the A / D converter 144, and the audio data that has been digitally converted is provided to the CPU 142. The CPU 142 analyzes the voice data, specifies the utterance start time and utterance end time of the section where the user is actually speaking (utterance section) based on the current time indicated by the clock circuit 150, and stores it in the memory 146. It memorize | stores in the formed table 146t (refer FIG. 3). The CPU 142 inputs the table 146t stored in the memory 146 to the computer 12 via the wireless LAN 148 every predetermined time (for example, 10 minutes).

  In the table 146t shown in FIG. 3, for example, “15h02m50s” means 15: 2: 50. The same applies hereinafter.

  However, as described above, the sensor 14 worn by the user and the tag 20 are fixedly associated with each other. Therefore, the sensor 14 has identification information (“person No.” described later) of the corresponding tag 20. Is input to the computer 12.

  Here, discrimination between the speech section and the silent section of the voice data will be described. In this embodiment, a portion having a level (power) larger than a predetermined threshold (50 dB in this embodiment) is determined as an utterance, and a portion having a power of 50 dB or less is determined as silence. However, in order to accurately determine whether or not the utterance has been started, it is determined that the utterance has started when a portion with a power greater than 50 dB is detected for a predetermined time t1 (in this embodiment, 50 milliseconds). It is. When a silent section is detected, the start time of the silent section is determined as the utterance end time (utterance end time). However, in order not to determine a pause (pause) due to a user's breathing (breath) or the like as a silent interval, the power is 50 dB or less, and if the state continues for a predetermined time t2 (for example, 300 milliseconds) or more, silence is generated. It is determined to be a section.

  Specifically, the CPU 142 provided in the sensor 14 executes an utterance detection process shown in FIG. However, the CPU 142 also executes audio data recording processing in parallel with the utterance detection processing.

  As shown in FIG. 4, when starting the speech detection process, the CPU 142 determines whether or not the speech has been started in step S1. If a voice with a volume of 50 dB or more continues for 50 milliseconds, it is considered that the utterance has started, “YES” is determined in step S1, and in step S3, the utterance is started at a time that is back by 50 milliseconds from the current time indicated by the clock circuit 150. The time is stored in the table 146t.

  In step S5, it is determined whether the utterance has ended. If the period during which sound with a volume of 50 dB or more is not input continues for 300 milliseconds or more, it is considered that the speech has ended, and the process proceeds to step S7. In step S7, the time lapsed by 300 milliseconds from the current time indicated by the clock circuit 150 is stored in the table 146t as the utterance end time. When the process of step S7 is completed, the process returns to step S1.

  As can be seen with reference to the table 146t of FIG. 3, in this embodiment, for the sake of simplicity, less than one second is omitted.

  The conversation state discriminating apparatus 10 having the configuration as shown in FIG. 1 is disposed in a building where a conference room and an exhibition hall are provided, for example. However, it is not necessary to be limited to such a building, and the conversation state determination device 10 can be applied to a place where conversation can occur. As shown in FIG. 5, the tag reader 16 is arranged in a place (here, places A, B, and C) such as a conference room or an exhibition hall, and a person (user) existing in the places A, B, and C is shown. Recognize each. That is, as described above, the user wears the sensor 14 and the tag 20, and therefore the tag reader 16 is in the detection range (within the dotted frame in FIG. 5) of the tag 20 worn by the user. Hereinafter, it may be referred to as “person No.”). The tag reader 16 executes the detection process of the tag 20 every certain time (in this embodiment, 1 second), and the identification information assigned to the detected person No. (hereinafter referred to as “sensor information”). To the computer 12. In this embodiment, for the sake of simplicity, sensor information A is assigned to the tag reader 16 arranged at the location A, sensor information B is assigned to the tag reader 16 arranged at the location B, and It is assumed that sensor information C is assigned to the tag reader 16 arranged. In addition, the user's voice is detected by the sensor 14, and the table 146 t created as described above is input to the computer 12.

  Although the detailed description is omitted, the same applies to the places B and C. In FIG. 5, the computer 12, the sensor 14, and the database 18 are omitted for simplicity. Furthermore, in FIG. 5, the user who exists in the location B and the location C is also omitted for convenience of drawing.

  In this embodiment, tag readers 16 are arranged at three locations (locations A, B, and C) to detect users and their utterance information in each location, but there is at least one such location. What is necessary is just four or more places.

  FIG. 6 is an illustrative view showing the contents of the database 18 shown in FIG. The database 18 includes a stay information storage area 30, an utterance information storage area 32, and a conversation leader information storage area 34. The stay information storage area 30 stores a table 30t for stay information, and this table 30t is shown as in FIG. The table 30t stores sensor information, entry time, and exit time corresponding to the identification information (person No.) of the tag 20. When the person number to which sensor information is added is input from the tag reader 16, the computer 12 obtains the current time from the clock circuit 12a, and uses the obtained time as the entry time together with the person number and sensor information in the table. Register (add) to 30t. Therefore, as can be seen from FIG. 7, person numbers and sensor information are registered in the table 30t in the order of early admission time. Further, when the person number and sensor information are no longer input from the tag reader 16, the computer 12 acquires the time at that time from the clock circuit 12a, and writes the time as the exit time in the corresponding item. By referring to such a stay information table 30t, a user (person No.) that exists (stays) in each of locations A, B, and C at a certain time (time zone) is specified (recognized). can do.

  In the utterance information storage area 32, a table 32t for utterance information is stored. As shown in FIG. 8, the table 32t stores an utterance start time that defines an utterance section corresponding to a person No. and an utterance end time corresponding to the utterance section. This table 32t stores the table 146t (FIG. 3) input from the wearable sensor 14 for each user (person No.). That is, the computer 12 creates the table 32t based on the table 146t to which the person number input from the sensor 14 is added.

  In the conversation leader information storage area 34, a table 34t of conversation leader information is stored. This table 34t is created by the conversation leader discriminating process (see FIGS. 11 and 12) described in detail later. Specifically, as shown in FIG. 9, corresponding to each place (sensor information), the conversation leader person number determined for each certain time zone is stored, and the conversation leader The information (conversation leader maintenance / change information) indicating whether or not it has been continued (maintained) or changed since this time period is stored. For example, in the sensor information A (location A), the conversation leader is the user of the person No. 8 between 15:01 and 15:10, and the next 15: 10: 1 to 15:20 It can be seen that the conversation leader is the user of the person No. 9. When the user of the person No. 9 is determined as the conversation leader, it can also be seen that the conversation leader has been changed. Further, it is understood that the conversation leader is the user of the person No. 9 during the next 15: 20: 1 second to 15:30, and the conversation leader is maintained from the immediately preceding time zone. Further, as will be described in detail later, as indicated by the conversation leader's determination result for the sensor information B, depending on the time zone (between 15:20:15 and 15:30), the conversation leader It may be determined that it does not exist (no conversation leader).

  Note that a blank in the conversation leader maintenance / change information column means that the conversation leader maintenance / change is not applicable.

  For example, the conversation leader is determined every time zone (for example, 10 minutes = 600 seconds) at each location (location A, B, C). Since the method for determining the conversation leader is the same at any place, place A will be described, and description of place B and place C will be omitted.

  When the computer 12 sets a time zone for discriminating conversation leaders, the computer 12 extracts users who were present at the location A in the time zone. That is, the person number describing the sensor information A is extracted in the time zone. Specifically, the person number is extracted with reference to the table 30t shown in FIG. For example, when the time zone is set between 15: 1 and 15:10, the person number indicating the sensor information A is extracted in the time zone. In the table 30t shown in FIG. 7, person numbers 3, 5, 8, and 9 are extracted.

  Next, the computer 12 calculates the utterance production rate of the user corresponding to each extracted person No. in the relevant time zone. Here, the utterance production rate is the ratio of the utterance amount (all utterance sections) in the set time zone (here, 600 seconds), and is calculated according to Equation 1.

[Equation 1]
Utterance production rate (%) = total utterance interval ÷ time slot × 100
However, the utterance amount, that is, the total utterance interval, is the total of the utterance intervals included in the time zone, and is obtained with reference to the utterance information table 32t shown in FIG. Referring to FIG. 8, for example, when considering the user of person No. 1, the user is from 15: 2: 50 to 15: 4: 10 between 15: 1 and 15:10. For 80 seconds and 30 seconds from 15: 6: 10 to 15: 6: 40. Therefore, the total utterance period in this case is 110 seconds, and the utterance production rate of the user is about 18.3% when calculated according to Equation 1.

  The result of calculating the utterance production rate in this way is shown, for example, in FIG. That is, as described above, in the extracted person numbers 3, 5, 8, and 9, the utterance production rates in the time zone are 20%, 10%, 80%, and 60%, respectively.

  The conversation leader is discriminated based on the utterance production rate, but according to a simple method, the user corresponding to the maximum utterance production rate can be discriminated as the conversation leader. That is, in the example shown in FIG. 10, the maximum utterance production rate is 80%, and the conversation leader in the time zone can be determined as the user of the person No. 8. However, in order to accurately determine the conversation leader, the following two conditions are set. Condition (1) is that the utterance production rate is equal to or higher than a predetermined rate (for example, 60%). If there are two or more utterance production rates that satisfy the condition (1), the condition (2) must be varied among the utterance production rates. Specifically, the difference between the utterance production rates is a certain number (for example, 10%) or more.

  Regarding the condition (1), when the maximum utterance production rate is selected, if the utterance production rate is smaller than the predetermined rate, the length of the utterance time is short, and it is appropriate to discriminate as the conversation leader. It is because it is thought that there is not.

  As for condition (2), if there are two or more utterance production rates exceeding a predetermined rate, two or more users are considered to be conversation leader candidates, and if there is no variation, any user This is because it cannot be determined whether or not should be determined as a conversation leader.

  According to such a condition, in the example shown in FIG. 10, it is the persons No. 8 and 9 that satisfy the condition (1), and these utterance production rates satisfy the condition (2). Therefore, the user of the person No. 8 with the highest utterance production rate is determined as the conversation leader. When the conversation leader is determined, as shown in FIG. 9, the numerical value indicated by the person number of the conversation leader is written in the table 34t corresponding to the sensor information and the time zone. Further, it is determined whether or not the person number of the conversation leader in the immediately preceding time zone, that is, the person number indicated by the previous discrimination result, matches the No. indicated by the current discrimination result. If they match, it is determined that the conversation leader is continuing, “Continue” is written in the conversation leader maintenance / change information column, and conversely, if they do not match, the conversation leader It is determined that the user has changed, and “change” is written.

  However, if the conditions (1) and (2) are not satisfied, it is determined that there is no conversation leader in the time zone. In this case, as shown in FIG. 9, “None” is written in the column of the conversation leader person number corresponding to the sensor information and the time zone. At this time, nothing is written in the column of conversation leader maintenance / change information.

  Specifically, the computer 12 shown in FIG. 1 executes the conversation leader determination process shown in FIGS. In this embodiment, as described above, in order to determine the conversation leader for each certain time zone at each location (A, B, C), the conversation leader determination process described below is performed for each location. To be executed.

  When the computer 12 starts the conversation leader determination process, in step S11, the computer 12 sets a predetermined time interval (in this embodiment, 10 minutes = 600 seconds). Here, for example, as shown in FIG. 9, the time zone is set such as from 15: 1 to 15:10 (“15h00m01s / 15h10m00s”). In the subsequent step S13, the number N of determinations is initialized (N = 0). This determination number N is the number of times the conversation leader has been determined, and is omitted in FIG. 1, but is counted by an internal counter of the computer 12.

  Subsequently, in step S15, a person number in which sensor information corresponding to the place where the conversation leader is determined is extracted in the time period. Next, in step S17, the utterance production rate of the person in the time zone is calculated according to Equation 1. In step S19, it is determined whether or not there is a user (person No.) exceeding a predetermined rate (here, 60%).

  If “NO” in the step S19, that is, if there is no user exceeding the predetermined rate, it is determined that there is no utterance production rate that satisfies the condition (1), and in step S21, there is no conversation leader in the table 34t. The process proceeds to step S31 shown in FIG. On the other hand, if “YES” in the step S19, that is, if there is a user (person No.) exceeding a predetermined rate, it is determined that there is an utterance production rate that satisfies the condition (1), and in step S23 the person concerned Extract No. as the subject of analysis. In step S25, it is determined whether there is any variation in the extracted utterance production rate.

  However, when the utterance production rate extracted in step S23 is one, the determination process in step S25 is not executed and the process proceeds to step S27 as it is.

  If “NO” in the step S25, that is, if it is determined that there is no variation in the extracted speech production rate and the condition (2) is not satisfied, the process proceeds to a step S21. On the other hand, if “YES” in the step S25, that is, if it is determined that there is a variation in the extracted utterance production rate and the condition (2) is satisfied, the utterance production among the extracted analysis subjects in the step S27. The person with the highest rate is identified (determined) as the conversation leader. In step S29 shown in FIG. 12, the determined conversation leader person number is written in the corresponding field of the table 34t, and the process proceeds to step S31.

  In step S31, the determination number N is incremented by 1, that is, the internal counter is incremented. In step S33, it is determined whether the determination number N is 2 or more. If “NO” in the step S33, that is, the determination number N is 1, and when the conversation leader is determined for the first time, the process proceeds to a step S41 as it is. On the other hand, if “YES” in the step S33, that is, if the determination number N is 2 or more, it is determined that the second or subsequent conversation leader is determined, and in step S35, the conversation leader determined this time is the previous time. It is determined whether or not the same person as the determined conversation leader. That is, referring to the table 34t, it is determined whether or not the person No. in the previous determination result matches the person No. in the current determination result.

  If “YES” in the step S35, that is, if the conversation leader determined last time and the conversation leader determined this time are the same person, in the step S37, the column of the conversation leader maintenance / change information in the table 34t is displayed. "Maintain" is written, and the process proceeds to step S41. On the other hand, if “NO” in the step S35, that is, if the conversation leader determined last time is different from the conversation leader determined this time, the conversation leader maintenance / change information column of the table 34t in the step S39. "Change" is written in the field, and the process proceeds to step S41.

  In step S41, the next time zone is set, and the process returns to step S17 shown in FIG. For example, in step S41, 15: 10: 1 to 15:20 is set as the next time zone.

  By repeating such processing, the conversation leader for each time zone in a certain place (place A, B or C) can be determined.

  In this conversation leader discrimination process, assuming that there are a plurality of users in the place where the conversation leader is discriminated, if the person number is extracted in step S15, the process proceeds directly to step S17. However, in step S15, when the extracted person number is 1 or less, that is, when the number of users is 1 or less, since the conversation is not established, the conversation leader cannot be determined. Therefore, a step (process) for determining whether or not there are two or more person numbers is provided between step S15 and step S17. If there are two or more person numbers, the process proceeds to step S17. If the number of persons is one or less, the process may proceed to step S23.

  As shown in FIG. 1, when the display device 22 is connected to the computer 12, for example, the determination result of the conversation leader as shown in FIG. 13 can be displayed on the display device 22. In FIG. 13, between 15 o'clock (strictly, 15: 1 sec) and 15:10, the conversation leader of the place A is the user of the person No. 8, and the user of the person No. 1 in the place B Is a conversation leader, and there is no conversation leader at location C.

  According to this embodiment, the voice signal of a user existing in a certain place is detected, the utterance production rate of each user in a certain time zone is calculated, and the maximum utterance production out of the utterance production rates exceeding a predetermined rate. Since the user who becomes the rate is determined as the conversation leader, the conversation leader can be accurately determined.

  In this embodiment, the utterance production rate in a certain time zone is calculated, and the maximum utterance production rate among the utterance production rates satisfying the conditions (1) and (2) based on the utterance production rate. Were identified as conversation leaders. However, the conversation leader may be determined based on the amount of speech (speech time) in a certain time zone. In such a case, the predetermined rate (utterance production rate) in the condition (1) is set as a predetermined amount (utterance amount), and the variation in the condition (2), that is, the numerical value of the difference in the speech production rate is the numerical value of the difference in speech production do it.

  In this embodiment, since only the conversation leader is determined, the user's voice signal is not recorded. However, if the voice signal is also recorded, after the conversation leader is determined, the determination is made. Based on the voice signal of the conversation leader, the conversation state of the place where the conversation leader was present in that time zone can be determined. Such a conversation state discriminating apparatus and method are described in detail in Japanese Patent Application No. 2004-3976 filed earlier by the applicant of the present application and are not an essential part of the present application. To.

FIG. 1 is an illustrative view showing one example of a configuration of a conversation leader discriminating apparatus of the present invention. FIG. 2 is an illustrative view showing a configuration of the wearable sensor shown in FIG. FIG. 3 is an illustrative view showing a table formed in the memory of the wearable sensor shown in FIG. FIG. 4 is a flowchart showing the utterance detection process of the CPU of the wearable sensor. FIG. 5 is an illustrative view for explaining an application example of the conversation leader discriminating apparatus shown in FIG. 1 embodiment. FIG. 6 is an illustrative view showing the contents of the database shown in FIG. 1 embodiment. FIG. 7 is an illustrative view showing one example of a table of stay information stored in the database shown in FIG. 1 embodiment. FIG. 8 is an illustrative view showing one example of a table of speech information stored in the database shown in FIG. 1 embodiment. FIG. 9 is an illustrative view showing one example of a table of conversation leader information stored in the database shown in FIG. 1 embodiment. FIG. 10 is an illustrative view for explaining a determination method for determining a conversation leader from the calculated utterance production rate. FIG. 11 is a flowchart showing a part of the conversation leader discrimination process of the computer shown in FIG. 1 embodiment. FIG. 12 is a flowchart subsequent to the conversation leader discriminating process shown in FIG. FIG. 13 is an illustrative view showing a display example of the output device shown in FIG. 1 embodiment.

Explanation of symbols

DESCRIPTION OF SYMBOLS 10 ... Conversation leader discriminating device 12 ... Computer 14 ... Wearable sensor 16 ... Tag reader 18 ... Database 20 ... Tag 22 ... Output device 142 ... CPU
146 ... Memory 152 ... Microphone

Claims (3)

A conversation leader discriminating apparatus for discriminating a leader of a conversation between two or more persons existing in a certain place,
A plurality of microphones for collecting audio for each of the two or more persons;
Individual recognition means for individually recognizing a person existing in the place,
Calculation means for calculating the amount of speech in a certain time zone of each person recognized by the personal recognition means based on the output of the microphone ;
Extraction means for extracting a person having an utterance amount exceeding a predetermined value out of the utterance amount for each person calculated by the calculation means;
Determining means for determining whether or not there is a variation in the utterance amount based on a difference in utterance amount for the extracted person when there are a plurality of persons extracted by the extracting means; and
When it is determined by the determination means that there is no variation, it is determined that there is no leader of the conversation, and when it is determined by the determination means that there is variation, the utterance amount that maximizes the utterance amount A conversation leader discriminating device comprising a leader discriminating means for discriminating a person as a conversation leader. The conversation leader discriminating apparatus according to claim 1, wherein the leader discriminating means further discriminates that there is no conversation leader when the utterance amount of all persons does not exceed the predetermined value . A conversation leader discriminating method for discriminating a leader of a conversation between two or more persons existing in a certain place,
(a) collecting audio for each of the two or more persons,
(b) Recognize the person who exists in the place individually,
(c) Based on the voice collected in step (a), the amount of speech in a certain time zone of each person recognized in step (b) is calculated ,
(d) Out of the utterance amount for each person calculated in the step (c), a person having an utterance amount exceeding a predetermined value is extracted,
(e) When there are a plurality of persons extracted by the step (d), determine the presence or absence of variation in the amount of speech based on the difference in the amount of speech for the extracted person, and
(f) When it is determined in step (e) that there is no variation, it is determined that there is no conversation leader, and when the determination means determines that there is variation, the utterance amount is maximum. The conversation leader discriminating method which discriminate | determines the person of the utterance amount used as a conversation leader.

Images