JP2007049657A - Automatic answering telephone apparatus - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an automatic answering telephone apparatus which is capable of determining the emotion of a calling party from voice signals of an automatically recorded voice message and controlling the reproduction order of automatic answering voice messages in accordance with a value indicating the extent of the emotion of a called party or the importance of the received voice message. <P>SOLUTION: The automatic answering telephone apparatus includes: a voice emotion recognition section 4 for performing voice emotion recognition regarding voice messages recorded in an automatic answering telephone set or in a device within an automatic answering telephone service center and giving the combination of values indicating the extents of emotions; a voice recognition section 9 for determining the values indicating the extents of emotions or the importance of the voice messages by performing voice recognition and text analysis; a data processing section 6 for determining importance from the combination of the values indicating the extents of emotions and determining priorities; a storage section 5 for storing priority information determined by the data processing section 6 together with the voice messages; and an output section 7 for reproducing or transmitting voice messages in accordance with the order of priorities. <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention relates to an answering machine.

  2. Description of the Related Art Conventionally, there is known an answering machine function that magnetically or electrically records a voice message from a caller in a telephone or an apparatus in an answering machine service center when a telephone call associated with a telephone call is not established. Such an answering machine function is to receive a caller's voice message when the receiver reproduces the contents of the answering machine later. However, they simply reproduce recorded voice messages in chronological order.

  Voice messages recorded by the answering machine may contain information that is either important or not important to the recipient, which can be determined by the recipient after listening to all voice messages. However, there is a problem that it is impossible to control the playback order according to the importance in advance.

Patent Document 1 describes a voice emotion recognition system for recognizing the strength of a speaker's emotion in the voice recognition system. According to the system disclosed in Patent Document 1, a word that is a target of speech recognition is output as a speech recognition result from the features of speech, and an emotion strength that indicates the degree of emotion of a speaker who has speech is output. However, there is a problem in that the playback order of the answering voice messages is not controlled according to the emotion of the caller, and the importance of the voice message cannot be determined based on the emotional intensity.

Patent Document 2 describes an emotion recognition system based on text analysis that can extract emotion information that more closely resembles human emotions. According to the system disclosed in Patent Document 2, the strength of emotion can be calculated by using syntax information such as a word including a human emotion included in the text and a destination of the word including the emotion. Voice signal is not used, analysis based on paralinguistic / non-linguistic information contained in the spoken language is not performed, and the order of playing voice messages and the strength of emotion There is a problem that the importance of voice messages cannot be determined.

  Further, Patent Document 3 describes a voice message theme detection device that enables a recipient of a voice message to know a voice message theme (message). According to the apparatus disclosed in Patent Document 3, the subject is merely estimated using the frequency information in the recorded data of the keyword prepared in advance, the caller's emotion is not taken into account, and the emotion is followed. There was a problem that voice message importance cannot be assigned.

  Patent Document 4 describes an answering machine that can estimate a caller's name by voice recognition and selectively reproduce a voice message from the caller. According to the apparatus disclosed in Patent Document 4, speaker estimation is performed only by comparing the voice of the caller prepared in advance with the voice of the speaker in the received voice message. There was a problem that emotion was not taken into account, and voice message importance according to emotion could not be given.

  JP 11-1119791 A   JP 2004-259238 A   JP2004-96502   JP-A-10-65792

  The voice messages recorded in the answering machine or the answering service center are generally reproduced in the order recorded. In that case, the emotion and importance included in each voice message are determined, and the playback order is not determined.

  As described above, according to the conventional technique, neither the useless information nor the important information for the receiver is accompanied by the determination of the priority, so that the receiver has to listen to all voice messages to the end.

  The present invention provides an information transmission service that is efficient for the user by determining the importance of the voice message from the utterance information of the speaker such as prosody, and determining the priority and controlling the playback order. It is an object.

  The present invention analyzes a voice signal of a voice message recorded in a case where a call associated with a call made by a telephone is not established by an emotion voice recognition unit, and determines at least one emotion level for the recorded voice signal. A combination of values is assigned, and combination data of emotion level values for the voice message is stored in the storage unit, and a voice message including a specific emotion is preferentially reproduced or transmitted according to a request of the receiver It is characterized by doing.

  Further, according to the present invention, the importance of the voice message is determined in the data processing unit based on the combination of the emotional degree values, and the voice message considered important for the receiver is preferentially reproduced or transmitted in the output unit. It is characterized by that.

  In addition, the present invention includes the voice message by performing text analysis such as keyword matching on the utterance content text of the caller obtained by voice recognition in parallel with the voice emotion recognition of the voice message. It is characterized by assisting the judgment accuracy of emotion or importance.

  According to the invention, it is possible to preferentially reproduce a recorded voice message including a specific emotion more strongly according to the request of the recipient by voice emotion recognition, and further determine the importance of the content of the recorded voice message, The playback order of messages can be controlled. As a result, the receiver can preferentially obtain important information, and the information acquisition efficiency is improved. In addition, by installing in a message dial at the time of a disaster, a highly urgent message can be preferentially processed, and congestion of the line at the time of a disaster can be alleviated.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings. FIG. 1 is a basic configuration diagram for realizing an embodiment of an answering machine according to the present invention. An embodiment of the answering machine according to the present invention will be described. The flow of processing in the present embodiment is shown in FIG.

1) For a voice message from a caller recorded via the input unit 3 of the receiver's answering machine, the emotional voice recognition unit 4 digitizes the basic frequency and power of the voice message in time series to Emotion analysis by prosody etc. is performed by measuring and digitizing.
2) Emotion analysis quantifies each emotion such as anger, surprise, sadness, joy, fear, contempt, comfort, calmness, discomfort, excitement, cheerfulness, boredom, tension, satisfaction, depression, intimidation, etc. It is stored in the storage unit 5.
3) When a voice message reproduction request is received from the recipient via the operation control unit 8, the caller's utterance content text obtained by voice recognition of the voice message held in the storage unit 5 Text analysis by keyword matching, etc., and if the receiver is set to perform playback according to a specific emotion strength in advance, the emotion strength is determined based on the keyword frequency information and If playback is set according to the importance of the voice message, the importance is determined.
4) When there is a voice message reproduction request from the recipient via the operation control unit 8, the recipient receives in advance a voice message to which a combination of emotion degree values held in the storage unit 5 is assigned. When the playback order according to the intensity of a specific emotion is set, the data processing unit 6 uses the combination of the emotion degree values together with the determination of the emotion intensity by the voice recognition unit 9 When the priority is determined and the priority information is stored in the storage unit 5 and the reproduction order according to the importance of the voice message is set, the data processing unit 6 and the importance recognition in the voice recognition unit 9 are combined. The priority ranking is performed and the priority information is stored in the storage unit 5.
5) The voice message is reproduced by the output unit 7 in accordance with the priority order given above.
6) After playing the voice message, confirm the deletion of the voice message for reception for each voice message. When there is a deletion request from the receiver, the voice message is deleted from the storage unit 5 via the operation control unit 8.

  When there is no deletion request from the receiver, the priority value of the voice message held in the storage unit 5 is set to −1, and subsequent playback requests from the receiver are handled as the lowest priority message. I will do it. This is because the voice message is known to the recipient and is considered unimportant.

  When there is no deletion request from the apparatus receiver obtained by the present invention, the importance value of the voice message held in the storage unit 5 is set to −1, and the subsequent reproduction requests from the receiver are the lowest. It will be handled as a message of importance. This is because the voice message is known to the recipient and is considered unimportant.

  The determination of the importance level of the voice message in the data processing unit 6 is important by using a multidimensional function from a combination of a plurality of emotion degree values given by the voice emotion recognition unit 4 and held in the storage unit 5. A degree value can be obtained. In obtaining the function, first, a listening experiment is performed on a large number of sample voice messages in which a combination of emotion degree values is obtained in advance by voice emotion recognition, and how important the voice message sounds. A number of evaluators are evaluated in five stages. Further, for the combination of emotion level values included in the voice signal of each voice message, the average value of the five levels obtained by the listening experiment is used as the importance value for the combination of emotion level values. A function that approximates the combination of the value of the emotion level included in the voice signal of the sample voice message used in the listening experiment and the importance value obtained by the listening experiment by a heuristic method is obtained. Further, multivariate analysis may be performed using each emotion as an explanatory variable and importance as a reference variable.

  As a form of the apparatus used in the present invention, the answering machine described in the embodiments may be a fixed telephone, a mobile phone, an apparatus in the answering service center, or software equipped with a telephone function installed in a computer. Also, the line network is not limited to a public line network, but may be a line network using the Internet such as an IP network or a private line network in an organization such as a company or a university.

  INDUSTRIAL APPLICABILITY The present invention can be applied to an industrial field in which a device having an answering machine function is manufactured and sold, an industrial field in which an answering machine service is provided in a communication network, or an industrial field in which a message service that conveys safety information in a disaster is provided.

  1 is a basic configuration diagram for realizing an embodiment of an answering machine according to the present invention.   It is a basic lineblock diagram of a voice emotion recognition part.   It is a table | surface which shows the example of the data stored in the memory | storage part.   It is an example of embodiment of the present invention, and is a flow chart of processing of an example in a telephone equipped with an answering machine function.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Main body 2 Bus 3 Input part 4 Voice emotion recognition part 5 Storage part 6 Data processing part 7 Output part 8 Operation control part 9 Voice recognition part

Claims (8)

  The voice signal of the voice message that is recorded when a telephone call is not established is analyzed by emotion voice recognition, and at least one emotion level combination is assigned to the recorded voice signal. An answering machine that preferentially reproduces or transmits a voice message containing a specific emotion more strongly to the receiver's terminal in accordance with the request of the receiver.   2. The importance level of a voice message is determined based on a combination of emotion level values given in claim 1, and the voice message is reproduced or transmitted to a receiver terminal in the order of importance level. Answering machine described in 1.   In parallel with the voice emotion recognition of the voice message according to claim 1, the text analysis is performed on the caller's utterance content text obtained by the voice recognition, thereby assisting the determination accuracy of the importance level of the voice message. The answering machine according to claim 1, characterized in that:   The answering machine according to claim 1, wherein the terminal of the receiver in claim 1 is built in or externally attached to the mobile phone.   The apparatus according to claim 1, wherein the apparatus is built in or externally attached to an apparatus installed for the purpose of providing an answering machine service in an answering machine service center, and is transmitted to a receiver's terminal. The answering machine described.   The answering machine according to claim 1, wherein the line network used in claim 1 is used through a telecommunication line network such as an IP network.   2. The answering machine according to claim 1, wherein a telecommunications line network such as an intranet network in a company or a university is used as the line network used in claim 1.   2. The device according to claim 1, wherein the device is built in or externally attached to a device having a message service function for conveying safety information, etc., especially in the event of a disaster, and is connected through a circuit network from the caller and receiver terminals in the event of a disaster. The answering machine according to claim 1, wherein the voice message of the caller is transmitted to the receiver.

Images