JP2019095552A - Voice analysis system, voice analysis device, and voice analysis program - Google Patents

Abstract

To enable effective responding to a symptom of bullying.SOLUTION: A sound collection part 1 collects a voice in a school classroom, and outputs a voice signal corresponding to the voice. A voice input part 51 receives the voice signal from the voice collection part 1. A voice analysis part 52 detects the occurrence of a predetermined event related to a symptom of bullying from the voice signal, and specifies at least one of a victim and a related person in the event on the basis of at least one of a voiceprint of the voice indicating the occurrence of the predetermined event, the occurrence place of the predetermined event, and a name detected from the voice indicating the occurrence of the predetermined event.SELECTED DRAWING: Figure 1

Description

  The present invention relates to a voice analysis system, a voice analysis device, and a voice analysis program.

  Bullying in groups such as schools is regarded as a problem, and it is required to detect signs of bullying early and take countermeasures.

  Some voice data analysis devices make possible voice early detection of bullying by performing voice recording in a specific time zone in a school, and detecting a slandering keyword from voice data obtained by voice recording (for example, patent) Reference 1).

JP, 2008-165097, A

  However, although the above-mentioned voice data analysis device can automatically detect a bullying action from voice, it is difficult to identify a victim or a related person (a perpetrator, a companion, etc.) of the bullying, and the bullying is May not be able to effectively respond to

  The present invention has been made in view of the above problems, and an object of the present invention is to obtain a voice analysis system, a voice analysis device, and a voice analysis program that enable an effective response to bullying signs.

  A voice analysis system according to the present invention includes a sound collection unit that collects voice in a classroom of a school and outputs a voice signal corresponding to the voice, and a voice analysis device that receives the voice signal from the sound collection unit. . Then, the voice analysis device detects the occurrence of a predetermined event related to a bullying sign from the voice signal, and generates a voiceprint of voice indicating the occurrence of the predetermined event, the occurrence location of the predetermined event, and the occurrence of the predetermined event. At least one of a victim and a related person in the event is identified based on at least one of the names detected from the indicated speech.

  A voice analysis apparatus according to the present invention collects a voice in a classroom of a school and outputs a voice signal corresponding to the voice from the voice collection unit that receives the voice signal from the voice collection unit; Based on at least one of a voice print of voice indicating the occurrence of the predetermined event, a location of the predetermined event, and a name detected from voice indicating the occurrence of the predetermined event. And a voice analysis unit for identifying at least one of a victim and a related person at the event.

  A voice analysis program according to the present invention includes a voice input unit that receives a voice signal from a sound collection unit that picks up a voice in a classroom of a school and outputs a voice signal corresponding to the voice from a computer; A voice print of a voice indicating a occurrence of a predetermined event related to a bullying sign and indicating the occurrence of the predetermined event, a location where the predetermined event occurs, and a name detected from a voice indicating the occurrence of the predetermined event Based on it, it functions as a voice analysis unit that identifies at least one of the victim and the related person in the event.

  According to the present invention, it is possible to obtain a speech analysis system, a speech analysis device, and a speech analysis program that enable an effective response to bullying signs.

  The above or other objects, features and advantages of the present invention will become more apparent from the following detailed description in conjunction with the accompanying drawings.

FIG. 1 is a block diagram showing the configuration of the speech analysis system according to the embodiment of the present invention. FIG. 2 is a view showing an example of the student data 31 in FIG. FIG. 3 is a diagram showing an example of the event occurrence history data 41 in FIG.

  Hereinafter, embodiments of the present invention will be described based on the drawings.

  FIG. 1 is a block diagram showing the configuration of the speech analysis system according to the embodiment of the present invention. The voice analysis system shown in FIG. 1 includes a sound collection unit 1 and a voice analysis device 2. The sound collection unit 1 collects the voice in the classroom of the school and outputs a voice signal corresponding to the voice. For example, the sound collection unit 1 is a microphone or the like installed in a classroom of a school. The sound collection unit 1 includes an amplifier, an electronic circuit such as a communication circuit, a power supply circuit, and the like as necessary. The voice analysis device 2 is a terminal device or an electronic device installed in a school, an external server, or the like. The connection between the sound collection unit 1 and the voice analysis device 2 may be wired or wireless.

  The voice analysis device 2 receives the voice signal from the sound collection unit 1 and performs predetermined voice analysis processing on the voice signal. Specifically, the voice analysis device 2 detects the occurrence of a predetermined event related to a bullying sign from the voice signal, and based on at least one of the voice print of voice indicating the occurrence of the predetermined event and the occurrence location of the predetermined event. Identify the victim and / or the related person at the event.

  As shown in FIG. 1, the voice analysis device 2 includes a user interface 11, a communication interface 12, a storage device 13, and an arithmetic processing device 14.

  The user interface 11 includes a display device (such as a display panel) that displays various information to the user and an input device (such as a keyboard) that receives user operations. The communication interface 12 is a peripheral device interface, a network interface or the like for communicating with an external device.

  The storage device 13 is a non-volatile storage device (hard disk drive, flash memory, etc.) storing various programs and various data. Here, the storage device 13 stores a voice analysis program 21. Further, in the storage device 13, student data 31, a bullying sign event list 32, and a keyword list 33 are stored. Furthermore, event occurrence history data 41 is stored in the storage device 13.

  The arithmetic processing unit 14 is a computer provided with a central processing unit (CPU), a read only memory (ROM), a random access memory (RAM), etc., and loads a program stored in the ROM or the storage device 13 into the RAM. By being executed by the CPU, it operates as various processing units.

  Here, the arithmetic processing unit 14 operates as the voice input unit 51, the voice analysis unit 52, and the education support unit 53 by being executed by the voice analysis program 21.

  The voice input unit 51 receives a voice signal from the sound collection unit 1. When an analog audio signal is output from the sound collection unit 1, the analog audio signal is converted into a digital audio signal by the analog-to-digital converter, and the digital audio signal is input to the audio input unit 51.

  The voice analysis unit 52 detects the occurrence of a predetermined event related to a bullying sign from the voice signal, and at the same time based on at least one of the voiceprint of voice indicating the occurrence of the predetermined event and the occurrence location of the predetermined event. Identify at least one of the victim and the related person.

  For example, when one of the conditions described later described in the bullying sign event list 32 is satisfied, it is determined that the occurrence of the predetermined event is detected. In addition, whether the speaker of the above-mentioned voice is a victim or a related person is specified from a word (student's name, a specific keyword, etc.) included in the utterance voice, as in the case of specifying an event type described later. .

  For example, the voice analysis unit 52 identifies (a) voice indicating occurrence of a predetermined event from the above-described voice signal, and (b) a student who associates a plurality of students in the classroom with voiceprint data of the plurality of students. Based on the data 31, a student of voiceprint data that matches the voiceprint of the specified voice is identified as a victim or a related person.

  For example, the student data 31 further associates the plurality of students in the classroom with the seat positions of the seats assigned to the plurality of students, respectively. Then, based on the correspondence relationship between the seat position and the student, the voice analysis unit 52 is a voice including a voice when the student speaks (for example, at the morning meeting or at a class time in a predetermined time zone) while sitting at the seat. Voiceprint data of the student is generated from the signal, and the generated voiceprint data and the student are associated with each other and added to the student data 31.

  In addition, for example, the sound collection unit 1 outputs a plurality of audio signals corresponding to the audio collected at a plurality of positions, and the audio input unit 51 receives the plurality of audio signals. Then, the voice analysis unit 52 specifies (a) a voice generation position indicating the occurrence of a predetermined event from the plurality of voice signals, and (b) a seat allocated to a plurality of students and a plurality of students in the classroom On the basis of the student data 31 associating the seat position with each other, the student at the seat position matching the specified occurrence position is identified as a victim or a related person. For example, the sound generation position is specified from the coordinates of a plurality of sound collection positions, the sound collection timing of sound at a plurality of sound collection positions, and the like. Also, for example, the sound collection unit 1 collects sound at a plurality of positions with a plurality of directional microphones, and the sound generation position is specified from the coordinates of the plurality of sound collection positions, the sound collection volume of a plurality of directional microphones, etc. Be done.

  FIG. 2 is a view showing an example of the student data 31 in FIG.

  For example, as shown in FIG. 2, in the student data 31, for each student, the student ID of the student, the voiceprint data of the student, the seat position data indicating the seat position in the classroom of the student, the name of the student, etc. It is included. Voice print data is, for example, feature point data that can be used for voice print authentication.

  Further, the voice analysis unit 52 (a) specifies the keyword included in the voice by the voice recognition process for the voice signal described above, and (b) the speaker of the voice of the voice described above by the emotion recognition process for the voice signal The emotion is identified, and the type of the predetermined event is identified based on the keyword and the emotion. That is, the occurrence of a predetermined event is detected in consideration of the emotion of the detected victim and the emotion of the related person, and the event type is identified.

  The type of the predetermined event is described in advance in the bullying sign event list 32. The bullying symptom event list 32 associates the plurality of predetermined event types with the determination conditions corresponding to the plurality of predetermined event types.

  For example, as the event type, “sneaking”, “inhibition”, “command-based communication”, “smiling slander”, “force”, “fight”, etc. are set.

  For example, with regard to “laughing”, it is a condition that other students (related persons) have made a laugh, etc. after a specific student (victim) says or while the specific student's name is called by a teacher during class etc. It is assumed. The specific student's utterance is identified by voiceprint analysis, and the specific student's name is extracted from the speech in the speech recognition process. In addition, the occurrence of mockery is identified by, for example, emotion recognition processing. Also, being in class may be detected, for example, as the voice signal is continuously without voice, only teacher's voice, or only one student's voice, or a predetermined class Based on the schedule (the start time and the end time of the class), it is determined whether each time point is in class.

  For example, with regard to “inhibition”, it is a condition that the speech of a specific student (victim) is blocked by the speech of another student (related person) in class or the like.

  For example, in the case of "command-type communication", in speech (that is, text data converted from speech signals by speech recognition processing), the "other" of another student (related person) with respect to a specific student (victim) " The condition is that a specific ending such as "eikoi" or "yuna" is detected.

  For example, in the case of "sludge slander", in the speech (that is, the text data converted from the speech signal by speech recognition processing), the keyword list 33 is a statement from another student (related person) to a specific student (victim). The condition is that a particular keyword listed is detected.

  For example, in the case of "compulsion", in a speech (that is, text data converted from speech signals by speech recognition processing), "speaking" from a specified student (victim) to another student (related person), "No" The condition is that a specific keyword such as "Stop" is detected. Here, instead of the specific student's remark, it may be conditioned that a specific emotion (such as crying) is detected by the emotion recognition process.

  For example, with regard to the “question”, it is a condition that a specific sound such as a loud noise or a fighting sound is continuously detected for a predetermined time. In this case, the victim and the related person are identified from the generation position of the specific sound.

  The education support unit 53 is configured to: (a) at least (a1) at least one of the identified predetermined event types; (a2) at least one of the identified victim and the related person; and (a3) an audio data file including the audio signal described above It associates and records in the event occurrence history data 41 as a history, and (b) outputs the history as a report.

  The voice analysis unit 52 identifies a group formed by a plurality of students in the classroom based on the above-described voice signal, and the education support unit 53 identifies (a) a group including a related person, b) The identified group may be indicated in the report.

  For example, the report is output at predetermined intervals, or output as needed when an upper limit described later is satisfied when the history is added to the event occurrence history data 41.

  Also, for example, when the number of histories in which a specific student is a victim is higher than the average number of histories per student, a report to that effect is given to a predetermined person in charge (such as a teacher or counselor). Will be sent.

  Also, for example, when the number of histories in which a specific student is a victim is continuously increasing, a report to that effect is transmitted to a predetermined person in charge.

  Also, for example, when the number of histories in which a specific student is a victim increases or decreases rapidly in a predetermined short period (when the increase or decrease in the number in a predetermined short period exceeds a predetermined threshold) , A report to that effect is sent to a predetermined person in charge.

  For example, the education support unit 53 uses the communication interface 12 to transmit the report as an e-mail or the like to the address of the person in charge registered in advance.

  At this time, along with the report, additional information such as an advice on a bullying sign or a risk of future bullying may be transmitted.

  The education support unit 53 also has a function of receiving student information using the user interface 11 and adding the information to the student data 31.

  FIG. 3 is a diagram showing an example of the event occurrence history data 41 in FIG.

  For example, as shown in FIG. 3, a history is added to the event occurrence history data 41 each time an event occurs. For each event, records indicating occurrence date / time, occurrence location data, victim student ID, related student ID, event type, evidence voice data (pass to voice data file), etc. are added to the event occurrence history data 41. The evidence speech data is generated from the speech signal in a predetermined format (here, wav format).

  Next, the operation of the voice analysis system will be described.

  The voice input unit 51 of the voice analysis device 2 starts acquiring the voice signal from the sound collection unit 1 when a predetermined start time (for example, school attendance time, classroom unlocking time, etc.) comes according to a predetermined schedule, and predetermined termination When it is time (for example, after school time, classroom locking time, etc.), the acquisition of the audio signal from the sound collection unit 1 is ended.

  During a period from a predetermined start time to a predetermined end time, the voice analysis unit 52 determines whether voice (such as speech) indicating the occurrence of the above-mentioned predetermined event is detected from the voice signal acquired by the voice input unit 51. Monitor.

  When a voice (such as a speech) indicating the occurrence of the predetermined event described above is detected, the voice analysis unit 52 generates a voiceprint or a generation position for each speech in the voice based on the student data 31 for the predetermined event. Based on the contents of the statement, the student (student ID) who is the speaker of the statement is identified as a victim or a related person, and the education support unit 53 generates the above record of the detected event as an event. It is added to the history data 41.

  In addition, the education support unit 53 analyzes the event occurrence history data 41 regularly or at the time of addition of the above-mentioned record, and when a specific student who satisfies the above condition is detected, the specific student is detected Send a report on the information to a designated person in charge.

  As mentioned above, according to the said embodiment, the sound collection part 1 collects the audio | voice in the classroom of a school, and outputs the audio | voice signal corresponding to the audio | voice. The voice input unit 51 receives the voice signal from the sound collection unit 1. The voice analysis unit 52 detects the occurrence of a predetermined event related to a bullying sign from the voice signal, and at the same time based on at least one of the voiceprint of voice indicating the occurrence of the predetermined event and the occurrence location of the predetermined event. Identify at least one of the victim and the related person.

  As a result, the related parties (victims and related persons) of the bullying behavior are identified, which makes it possible to effectively respond to the bullying signs.

  Note that various changes and modifications to the above-described embodiment will be apparent to those skilled in the art. Such changes and modifications may be made without departing from the spirit and scope of the subject matter and without diminishing the intended advantages. That is, such changes and modifications are intended to be included in the scope of the claims.

  For example, in the above-mentioned embodiment, the above-mentioned group may be specified based on a photography picture of a security camera installed in a school other than the above-mentioned sound signal. In that case, for example, a plurality of students who have been photographed simultaneously and continuously for a predetermined period of time are specified from the photographed image by face authentication or the like, and the specified plurality of students belong to one group. In addition, the above-mentioned predetermined time may be set according to an imaging place (a classroom, a club room, etc.).

  In the above embodiment, in the text data obtained by the speech recognition process for the speech signal of the speech uttered by the person concerned, the name of the student registered in the student data 31 is searched, and the student of the found name is It may be specified as a victim. In that case, for example, a student whose name is found in the text data when occurrence of an event whose event type is “command-based communication” is detected is identified as a victim.

  The present invention is applicable to, for example, a terminal device or an electronic device in a school.

1 sound collection unit 2 voice analysis device 14 arithmetic processing device (an example of a computer)
21 voice analysis program 51 voice input unit 52 voice analysis unit 53 education support unit

Claims (10)

A sound collection unit that collects voices in a classroom of the school and outputs a voice signal corresponding to the voices;
A voice analysis device for receiving the voice signal from the sound collection unit;
The voice analysis device detects, from the voice signal, the occurrence of a predetermined event related to a bullying sign, and a voiceprint of voice indicating the occurrence of the predetermined event, the location of the predetermined event, and voice indicating the occurrence of the predetermined event Identifying at least one of a victim and a related person in the event based on at least one of the names detected from
Voice analysis system characterized by A voice input unit that receives the voice signal from a sound collection unit that collects voice in a classroom of the school and outputs a voice signal corresponding to the voice;
From the voice signal, occurrence of a predetermined event related to a bullying sign is detected, and a voiceprint of voice indicating the occurrence of the predetermined event, location of the predetermined event, and a name detected from voice indicating the occurrence of the predetermined event A voice analysis unit that identifies at least one of a victim and a related person at the event based on at least one;
An audio analysis apparatus comprising:   The voice analysis unit identifies (a) voice indicating the occurrence of the predetermined event from the voice signal, and (b) a student who associates a plurality of students in the classroom with voiceprint data of the plurality of students. The voice analysis device according to claim 2, wherein a student of the voiceprint data matching the voiceprint of the voice identified is identified as the victim or the related person based on data. The student data respectively associates the plurality of students in the classroom with the seat positions of the seats assigned to the plurality of students,
The voice analysis unit generates the voiceprint data of the student from the voice signal including the voice of the student speaking when sitting on the seat, and the generated voiceprint data and the student are mutually compared. Associate and add to the student data,
The voice analysis device according to claim 3, characterized in that The sound collection unit outputs a plurality of audio signals corresponding to audio collected at a plurality of positions,
The voice input unit receives the plurality of voice signals,
The voice analysis unit (a) specifies a voice generation position indicating the occurrence of the predetermined event from the plurality of voice signals, and (b) is assigned to a plurality of students in the classroom and the plurality of students Identifying a student at the seat position matching the identified occurrence position as the victim or the associate based on student data respectively associating the seat position with the seat position;
The voice analysis device according to claim 2, characterized in that:   The voice analysis unit (a) identifies a keyword included in the voice by voice recognition processing on the voice signal, (b) identifies an emotion of a speaker of the voice by emotion recognition processing on the voice signal, 3. The voice analysis device according to claim 2, wherein the type of the predetermined event is specified based on a keyword and the emotion.   (A) at least (a1) the type of the specified predetermined event, (a2) at least one of the specified victim and the related person, and (a3) a voice data file including the voice signal in association with each other The speech analysis apparatus according to any one of claims 2 to 6, further comprising: an education support unit which records as (b) and outputs the history as a report. The voice analysis unit identifies a group formed by a plurality of students in the classroom based on the voice signal,
The education support unit (a) identifies the group in which the related person is included, and (b) indicates the identified group in the report.
The voice analysis device according to claim 7, characterized in that   9. The voice analysis device according to claim 8, wherein the group is identified based on the voice signal or a photographed image of a security camera installed in the school. Computer,
A voice input unit that receives the voice signal from a sound collector that picks up a voice in a classroom of the school and outputs a voice signal corresponding to the voice, and detects occurrence of a predetermined event related to a bullying sign from the voice signal. A victim of voice and an associate of the event based on at least one of a voice print of voice indicating the occurrence of the predetermined event, a location of the occurrence of the predetermined event, and a name detected from voice indicating the occurrence of the predetermined event A voice analysis program that functions as a voice analysis unit that identifies at least one.

Images