JPH07110021B2 - Interactive voice response device - Google Patents

Description

TECHNICAL FIELD The present invention relates to an interactive voice response device that responds in an interactive manner.

[Conventional technology]

The interactive voice response device, depending on the input contents from the line,
It selects a predetermined voice message and sends it to the line. In this device, the means for knowing the intention of the user (caller), that is, the means for inputting from the line is by recognizing the voice of the user or by non-voice such as PB (push button) signal. This is due to the exchange of information.

FIG. 2 is a flowchart showing an example of processing when voice recognition is used. In the figure Is a block for sending audio to the user, Indicates a recording block of the user's voice. As shown in the figure, in order to confirm the user's intention, for example, a response message instructing the utterance of "high" is transmitted, the voice of the user is recognized, and if the recognition result is "high", the next Then, if it is not "high", the same response message is sent again.

On the other hand, FIG. 3 is a flowchart showing a processing example using the PB signal. As shown in the figure, in this processing example, "high"
By transmitting a response message instructing a predetermined PB operation, which means "* 0", and detecting the PB signal, the same processing as the above processing example is performed.

[Problems to be solved by the invention]

However, in the case of processing by voice recognition, even if the user has an intention of "high", if a word other than the recognition target is erroneously uttered, such as "AE" or "dozo", There was a drawback that it recognized that there was input and would send the same response message again. Moreover, as shown in FIG.
"Please say" High "if you like."
For example, it is necessary to insert a redundant message instructing the content of utterance, which is very different from the conversation between humans, and thus it is difficult to use.

Also when processing with PB signals, etc., it is necessary to operate according to a predetermined procedure such as pressing "* 0" when it is "high" as shown in FIG. There is a drawback that the response message is repeated due to an erroneous operation, and since the user is forced to perform the special PB operation as described above, the user's burden becomes extremely large.

In order to solve these shortcomings, it is necessary to recognize the input content by everyday words with no restrictions on vocabulary, number of words, etc., as input means, but the current speech recognition technology requires a very large hardware scale. However, it is very expensive and difficult to apply to a telephone or the like.

In view of these problems, an object of the present invention is to provide a voice input response device which can respond to an input voice produced by a daily speech and has a simple structure.

[Means for solving problems]

The structure of the present invention will be described with reference to FIG. The present invention is
In an interactive voice response device that receives an input voice from a line, sends a response message based on the input voice, and responds in an interactive manner, an input voice detecting means 1 for detecting the input voice from the line, and a duration of the input voice. Measuring means 2 for measuring the length (voice time length) T, voice time threshold information storage means 3 for storing voice time threshold information, and response message information storage means for storing a plurality of response message information. 4, a comparing means 5 for comparing the voice time length T with a voice time length threshold value, and a response message sending means 6 for selecting a response message based on the comparison result and outputting it to the line.

[Action]

The operation of the present invention will be described with reference to FIG. The input voice of the other party from the line is detected by the input voice detecting means 1. Then, the measuring unit 2 measures the voice duration T. FIG. 4 is an explanatory diagram showing the correlation between the content of the input voice and the voice duration, where the horizontal axis represents the voice duration T and the vertical axis represents the appearance frequency (%). As shown in the figure, in general,
There is a correlation between the content of the input voice issued following the response message and the length of the voice time length T. Therefore, by comparing the length of the voice time length T with the voice time length threshold value by the comparison means 5, it is possible to select an appropriate response message corresponding to the input voice, and the response message sending means 6 selects. The response message is sent to the line.

〔Example〕

An embodiment of the present invention will be described with reference to FIGS.

FIG. 5 is a block diagram showing an automatic response recording apparatus according to an embodiment of the present invention, 7 is an incoming call detecting unit connected to the office lines L ₁ and L ₂ , and 8 is a 4- or 8-bit microcomputer. A control unit configured, 9 is a hook switch, 10 is a loop control unit connected in parallel with the hook switch 9, 11 is a call circuit unit connected to the office lines L ₁ and L ₂ via the loop control unit 10,
Reference numeral 12 is a response message transmission unit connected to the transmission terminals T ₁ and T ₂ of the call circuit unit 11, 13 is a response message information storage unit that is connected to the response message transmission unit 12 and stores a plurality of response message information, 14 Is a partner voice detection unit connected to the receiving terminals R ₁ and R ₂ of the calling circuit unit 11, 15 is a partner message recording unit that is also connected to the receiving terminal of the calling circuit unit 11, and 16 is the duration of the partner voice or A clock for measuring silent time length,
Reference numeral 17 is a time length threshold information storage unit for storing time length threshold information such as voice time length thresholds T _L , T _H (T _L <T _H ), silence determination threshold and occurrence end threshold. Is.

The operation of the above structure will be described with reference to FIG. First, when an incoming call is received, the incoming call detection unit 7 detects it and outputs an incoming call signal to the control unit 8. When the incoming signal is received, the control unit 8 operates the loop control unit 10 to close the loop after a lapse of a predetermined time, and ends the automatic incoming call operation.

FIG. 6 is a flow chart showing the response control in the apparatus of FIG. 5, and the following operation will be explained based on FIGS. 5 and 6. After the loop is closed, the response message sending unit 12 is activated to send the first message (for example, "Yes, I am Suzuki") stored in advance in the response message information storage unit 13 to the call circuit unit 11.

At this time, after sending the first message, the other party message recording unit 15 is operated to record the voice of the user (caller) obtained from the receiving terminals R ₁ and R _{2 of} the communication circuit unit 11.

At the same time, the presence / absence of the user's voice is monitored by the partner voice detection unit 14. The time length of the silent section is measured by the clock 16, and if the voice is detected before this time length exceeds the silence determination threshold value, the silent section is a pause between the voice section and the next voice section. The decision is made and the voice is monitored again to detect the start of a silent section of the voice. When the time length of the silent section exceeds the silent threshold, it is determined that the user's message has ended, and the response message to be sent next is selected based on the duration (voice duration). Control is started.

If the above first response message is "Yes, I am Suzuki", the following user message may be as follows.

When the user asks back "Eh?" Or "Hello?", For example, because the user could not hear the response message sent, or the content of the response message seemed unnatural. In this case, the voice time length T becomes extremely short.

 If it is silent for the same reason as above.

In the case of a message containing only the other party's name, such as "Is there Mr. Ichiro Suzuki?"
In this case, the voice time length T becomes relatively short.

In the case of a message that includes both the user's name and the name of the person who has a message, such as "I'm Tanaka from Tokyo, would you like Mr. Ichiro Suzuki?"
In this case, the voice time length T becomes relatively long.

Therefore, in the present embodiment, the voice duration T is measured by the clock 16 and compared with the voice duration thresholds T _L and T _H so that T ≦ T _L ,
It is classified into three categories of T _L <T <T _H and T _H ≤T.

Then, when T ≦ T _L , it is determined that the above case or is satisfied, and the first message is transmitted again.

When T _L <T <T _H , it is judged that it is the above case, and a response message for asking the user name such as "I'm sorry, which one is it?" Is selected and sent out. Record the voice of. Then, for example, a response message such as "Please tell us what you want to do" is transmitted, and the voice of the user is recorded again.

Further, when T _H ≦ T, it is judged that the above case,
Without asking for the user name as when T _L <T <T _H
A response message for inquiring about the requirement is sent, and the same operation is performed thereafter.

As described above, in the present embodiment, the other party voice detection unit 14 monitors the presence or absence of the user's voice, and the clock 16 measures the voice duration T to infer the content of the user's message and respond to it. Since an appropriate response message can be sent out by the user, there is an advantage that an extremely simple automatic response recording device can be provided for the user.

Next, another embodiment of the present invention will be described.

First, the outline of this embodiment will be described with reference to FIG. 7th
The figure is a flow chart showing the basic operation of the present embodiment.
In the above embodiment, the other party's message is not included in the user name and the other party's name if T ≦ T _L or the user's name is not included in the case of T _L <T <T _{H according} to the voice time length T of the other party. If T _L <T <T _H , the message includes the user's name, and if T _H ≤ T, the user name and the other party's name are included. , Some include the user name but not the other party's name. In this case, when the user name is not included and the partner name is included, the voice time length T is about the same, and it is difficult to distinguish from the voice time length T. Therefore, in the present embodiment, as shown in FIG. 7, when T _L <T <T _H, a reply such as “Yes” is sent, and when the user responds to this, the previous message is If it is judged that the message contains only the user name, and that the message after that contains only the name of the other party with a message, and there is no response from the user,
The user's message is determined to include only the name of the other party who has the message.

This embodiment will be described in detail with reference to FIGS. 8 and 9 (a) (b) (c). FIG. 8 is a block diagram showing an automatic response recording apparatus according to another embodiment of the present invention, which includes an incoming call detection section 7, a control section 8, a loop control section 10, a call circuit section 11, a response message transmission section 12, and a response. The message information storage unit 13, the partner voice detection unit 14, the partner message recording unit 15, the clock 16, and the time length threshold value information storage unit 17 are the same as those in FIG. Reference numeral 18 is a hammer transmission unit connected to the transmission terminals T ₁ and T ₂ of the communication circuit unit 11. If the response message information accumulating unit 13 accumulates the echo message information, the echo sending unit 18 can be omitted.

The operation of the above structure will be described with reference to FIG.
First, when there is an incoming call, the automatic incoming call operation similar to that of the embodiment of FIG.
The recording operation of the message voice obtained from the receiving terminals R ₁ and R _{2 of the} call circuit unit 11 is started.

The following operation will be described with reference to FIGS. 8 and 9 (a) (b) (c). 9 (a) (b) (c) is a flowchart showing the response control in the apparatus of FIG. As shown in FIG. 9 (a), after the completion of the automatic incoming call operation, the response message sending section 12 is activated to cause the first response message stored in the response message information accumulating section 13 in advance (for example, "Yes. Will be sent.)).

Next, the other party voice detection unit 14 is operated to monitor the presence and the length of the user's voice. If the user does not issue a message even after the silent threshold value stored in the time length threshold information storage unit 17 has elapsed, the user may not hear the first response message. Then, the first response message is sent again, since it is determined that the speech is in a silent state. If the user does not issue the message even after that, the user judges that he / she does not have the intention to describe the message, repeats this operation a predetermined number of times, and then shifts to the process termination process.

On the other hand, when the user's voice detection unit 14 detects the user's message, the following voice time length T is measured. As shown in FIG. 9 (b), first, the clock 16 is activated to start measuring the voice time length T of the user message. When the user message is determined to be silent by the other party voice detection unit 14, the clock 16 is stopped for a while, the voice time length T counted up to that point is stored, and then the silent time measurement is started. If the silent state continues for the occurrence end threshold time stored in the time length threshold information storage unit 17,
Although it is determined that the user's message has ended at that point, if the voice from the user is detected by the partner voice detection unit 14 before the occurrence end threshold time continues, the voice time saved earlier The length T is reloaded into the clock 16 and the voice time length T is integrated again.

Returning to FIG. 9 (a), description will be made. When the end of the message by the user is confirmed, the control unit 8 reads the voice time length T from the clock 16 and sets the voice time length thresholds T _L and T _H stored in the time length threshold value information storage unit 17. Make a comparison. The subsequent processing is as follows according to the result of the comparison.

When T _H ≦ T Since it can be determined that the user first message describes both the user name and the name of the other party with a message, the response message (from the response message information accumulating unit 13 asks the user for the message ( For example, "I'm out of office now, so please let me know if you have any requirements.") And send it to the station line.

When T _L <T <T _{H It} can be presumed that the first message of the user is either only the user or only the name of the other party who has the message. Therefore, after transmitting the echo (for example, "yes") from the echo output unit 18 to the office line, the partner voice detection unit 14 starts monitoring the user voice. The subsequent processing is divided into the following two depending on the presence / absence of user voice.

-1 When the user's voice is detected before a certain period of time elapses.

Since it can be determined that the user first message only describes the user name, and the user second message describes the name of the other party with a message, after confirming the end of the user message described above. , Selects a message asking the user's requirement from the response message information storage unit 13 and sends it to the central office line.

-2 When the user's voice is not detected even after a certain period of time.

Since it can be judged that the first message of the user is only the name of the other party with a business requirement, a message asking the user name from the response message information storage unit 13 (for example, "I'm sorry, which one is it?") Select and output to the office line.
After confirming the end of the user message, the response message information storage unit 13 selects a message inquiring about the user's requirement and sends it to the central line.

When T _L ≦ T Since it can be determined that the user's first message does not describe the user's name or the name of the other party with a message, the user selects the azuchi from the azuchi transmitter 18 and outputs it to the station line, and then the user again. The process for the first message is repeated.

By performing the above processing, the device sends a message asking for a message in any case. If there is no voice from the user in response to the message asking for the message, the message asking for the message is repeated a fixed number of times as in the case of sending the first response message, and then the process ends. If voice from the user is detected,
After confirming the end of the user message (Fig. 9 (c)), the response message information accumulating unit 13 selects a message indicating the end of the dialogue (for example, "Thank you very much") and sends it to the central line, and then the loop control unit 10 Is activated to open the loop, the partner message recording section 15 is stopped, and all processing is terminated.

In this way, the other party voice detection unit 14 and the clock 16 extract the duration information of the user message by the user, estimate the meaning of the user message from the result, and it is not possible to judge only by the voice duration information. The meaning of the message can be accurately estimated and understood. If you understand the meaning of the user message, it will be possible to output an appropriate response message to the user message,
There is an advantage that a good interactive response of the man-machine interface can be realized.

〔The invention's effect〕

As described above, according to the present invention, the response message corresponding to the content of the input voice can be transmitted by selecting the response message to be transmitted next based on the duration of the input voice. There is an effect that it is possible to realize an interactive voice response device capable of inputting by uttering everyday words.

[Brief description of drawings]

FIG. 1 is a block diagram showing the configuration of the present invention, FIG. 2 is a flow chart showing an example of processing when voice recognition is used, and FIG.
FIG. 4 is a flowchart showing a processing example using a PB signal,
FIG. 5 is an explanatory diagram showing the correlation between the content of the input voice and its duration, FIG. 5 is a block diagram showing an automatic response recording device according to one embodiment, and FIG. 6 is response control in the device of FIG. 7 is a flowchart showing the basic operation of another embodiment, FIG. 8 is a block diagram showing an automatic response recording device according to another embodiment, and FIG. 9 (a).
8B and 8C are flowcharts showing response control in the apparatus of FIG. 1 ... Input voice detecting means, 2 ... measuring means, 3 ... voice time length threshold information storing means, 4 ... response message information storing means, 5 ... comparing means, 6 ... response message sending means, 7 ... incoming call detection section, 8 ... control section, 10 ... loop control section, 11 ... calling circuit section, 12 ... response message sending section, 13 ... response message information storage section, 14 ... caller voice detection Part, 15 …… Recording message from the other party, 16 …… Clock,
17: Time length threshold information storage unit, 18: Aizu transmission unit.

Claims (1)

[Claims] 1. An input voice detecting means (1) for detecting an input voice from a line in an interactive voice response device which receives an input voice from a line, sends a response message based on the voice, and responds in an interactive manner.
A measuring means (2) for measuring the duration of the input voice, a voice time length threshold information storage means (3) for storing voice time length threshold information, and a plurality of response message information. Response message information storage means (4), comparison means (5) for comparing the duration of the input voice with a voice duration threshold, and a response message for selecting a response message based on the comparison result signal and sending it to the line. An interactive voice response device comprising: a sending means (6).