JP2556978B2 - Interactive answering machine - Google Patents

Description

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an answering machine, and more particularly, the present invention is for determining whether or not a caller has finished speaking a message. The present invention relates to an interactive answering machine which is devised to rationally control the end of utterance detection.

(Prior Art) An answering machine normally closes a loop after detecting an incoming call and sends a response message after sending a response message as shown in the flowchart of FIG. 2 (a). The recording is performed and the operation of releasing the loop is performed. In this case, the caller had to speak out the message at once, so he had to sort out his name, telephone number, message, etc. in a short time and start speaking. Therefore, it causes a great psychological burden, and as a result, there is a drawback that the recorded content of the message is insufficient or that many callers disconnect the message without recording the message. Therefore, in order to solve these problems, as shown in FIG. 2 (b), the applicant's application in which the caller's name, telephone number, message content, etc. are interactively guided and recorded one by one An interactive answering machine of Japanese Patent Application No. 60-27566 has been devised.

When recording a message of interest in an interactive manner, it is necessary to monitor the presence or absence of voice from the calling party and detect whether or not the calling party has finished speaking. In the case of human-to-human dialogue, it is possible to know the end of utterance from the semantic content of the other party's voice, but in the case of human-machine, it is extremely difficult to recognize the semantic content of human voice in real time. Have difficulty.
Therefore, as shown in FIG. 3, the presence or absence of the voice of the caller is monitored, and when the length of the silent portion exceeds a predetermined value (τ in FIG. 3), it is judged that the utterance has ended. This is a realistic method.

In FIG. 3, after the response message is transmitted, the start of utterance is monitored in a loop, and the end of utterance is monitored in a loop. The principle of ending vocalization in FIG. 3 is shown as follows.

The silence observed after the start of vocalization is either a temporary pause or the end of vocalization. Therefore, from the viewpoint that it is rare that the length of the silent part contained in the human voice exceeds a certain value (τ) statistically, from the viewpoint that the silent part longer than τ is observed, the silent part is It can be judged that it is the end of vocalization, not a temporary pose.

(Problems to be Solved by the Invention) However, the conventional method in which the utterance end detection time is set with a constant value τ has the following problems.

(1) In an interactive answering machine, as shown in FIG. 2 (b), the caller's name, telephone number, message, etc. are interactively guided and recorded one item at a time. In this case, for example, comparing the pose included in the voice that speaks the caller name with the pose included in the voice that speaks the message, the latter voice has a larger number and length in both poses. it is conceivable that. Therefore, when the same end-of-speech detection time is used for both, the reliability of end-of-speech detection is not the same, and there are more detection errors in the case message than in the case of the caller name. Is expected to occur.

(2) It is considered rare that the message ends after a very short time after the start of utterance. Therefore, the probability of ending the utterance should be different between the silence that occurs immediately after the caller starts speaking (eg, within 1 second) and the silence that occurs after a considerable amount of time has elapsed (eg, 10 seconds after) the beginning of speech. Is. Therefore, regardless of when the silence occurred,
When a fixed utterance end detection time (τ) is used, it is expected that the detection reliability varies depending on the position of the generated silence.

In order to solve these problems, an object of the present invention is to make the utterance end detection time variable depending on the content of the question and depending on the silent time observed after the start of utterance. .

(Means for Solving Problems) In order to solve the problems, an interactive answering machine is provided with an incoming call detecting means for detecting a calling signal from a central line,
After the incoming call is detected, the loop closing means for closing the loop, the response message storing means for storing a plurality of response messages, the response message transmitting means for transmitting the response message voice to the line, and the message message of the caller are displayed. From a voice message detecting means for recording a message message recording means for recording, a message message reproducing means for reproducing the message message, a voice detecting means for detecting whether or not the caller is speaking the message message, and the voice detecting means. Information, the utterance end detection means for detecting the end of utterance when the length of silent time observed in the caller's voice reaches a predetermined utterance end detection time, and the response message is transmitted after the automatic incoming call. In an interactive answering machine in which the line is released after repeating the recording of the message and the message in an interactive mode, the utterance end detection time is set to the value of the immediately preceding response message. With varied according to volume,
Based on the information from the voice detection means, an elapsed time measuring means for measuring an elapsed time from when the caller starts speaking is provided, and the elapsed time after the start of the speech obtained from the elapsed time measuring means is set to the elapsed time. A control means for adapting and reducing the utterance end detection time is provided and configured.

(Operation) Since the present invention is configured as described above, the problems in the prior art are solved, and depending on the content of the question, and depending on at what time after the start of the utterance, the utterance ends. The detection time can be variable.

Embodiment An embodiment of the present invention will be described below with reference to the drawings.

FIG. 1 is a block diagram showing the configuration of an embodiment of the present invention. In the figure, 1 is an incoming call detection circuit, 2 is a control unit, 3 is a hook switch, 4 is a loop control circuit, 5 is a response message storage unit, 6 is a call circuit, 7 is a response message transmission unit, and 8 is a message message. A recording section, 9 is a voice detection circuit, 10 is a voice end detection means, 11 is an elapsed time measurement means, 12
Is a message reproducing section.

 The operation of the embodiment will be described below with reference to FIG.

First, when there is an incoming call from the central office line, the incoming call detection circuit 1 detects it and outputs it to the control unit 2. After a lapse of a predetermined time, the control unit 2 activates the loop control circuit 4 connected in parallel with the hook switch 3 to close the loop and terminate the automatic incoming call operation.

Next, the control unit 2 sends the first reply message registered in advance in the reply message storage unit 5 to the office line by operating the reply message sending unit 7 connected to the sending terminal of the call circuit 6. To do. After that, the control unit 2 records the voice uttered by the caller in the message recording unit 8 in the following manner.

After sending the first response message, the control unit 2 monitors whether or not the caller is speaking a message based on the signal from the voice detection circuit 9, and detects whether the caller is silent. Measure the length. The utterance end detection means 10 determines the end of utterance when the measured silence length becomes longer than a predetermined value (utterance end detection time). At the time of determination, the control unit 2 causes the utterance end detection means 10 to change the utterance end variable as shown in FIG. 4 according to the content of the message and the elapsed time after the start of utterance obtained from the elapsed time measurement means 11. Give the detection time. The values in FIG. 4 depend on the contents of the response message on the machine side prior to the caller's voice and at what point in time after the start of utterance the silence to be detected begins.
For example, in the case of a question that can be answered in a short time, such as "I am sorry, which one is the last response message," shorten the initial value of the utterance end detection time as shown in (a) in Fig. 4. , And shortens relatively quickly over time. On the other hand, if there is a possibility that the recording time will be long, such as "Please record your message, please talk", set the initial value of the utterance end detection time to a long value as shown in (b) in the figure, and Reduce gradually. That is, the utterance end detection time is controlled depending on the utterance content and how much time has passed since the caller started uttering, and the utterance end is detected accurately and without redundancy. Immediately after the end of utterance is detected, the second response message is transmitted according to a preset order.

Hereinafter, the dialog recording is advanced by repeating the above-mentioned sending of the response message and the recording of the message. When the recording of all the message is completed, the control unit 2 controls the loop control circuit 4 to open the loop and return to the empty state.

When the owner of this apparatus plays back the recorded message in the room (at home), the control unit 2 uses the message recording unit 12 to record the message recorded in the message recording unit 8. The message can be listened to by playing it back.

(Effects of the Invention) As described above, by using the answering machine of the present invention, it is possible to quickly and reliably detect the end of the caller's utterance. That is, it is possible to prevent the machine from interrupting the operation before the end of utterance and shorten the response time to improve the usability.

[Brief description of drawings]

FIG. 1 is a block diagram showing the configuration of an embodiment of the present invention, FIG. 2 (a) is an operation flowchart of a conventional answering machine, and FIG. 2 (b) is an operation flowchart of a conventional interactive answering machine. FIG. 3 is a processing flowchart for detecting the end of utterance, and FIG. 4 is a graph showing the relationship between the elapsed time after the start of utterance and the end of utterance detection time. ...... Hook switch, 4 …… Loop control circuit, 5 …… Response message storage unit, 6 …… Call circuit, 7 …… Response message sending unit,
8 ... Message recording section, 9 ... Voice detection circuit, 10
…… Voice end detection means, 11 …… Elapsed time measurement means, 12…
… Message message playback section, L ₁ , L ₂ …… Local line.

Claims (1)

(57) [Claims] 1. An incoming call detecting means for detecting a call signal from an office line, a loop closing means for closing a loop after detecting an incoming call, a response message accumulating means for storing a plurality of response messages, and a response message voice. Response message sending means for sending a message to the line, a message message recording means for recording a message message of the calling party, a message message reproducing means for playing the message message, and the caller utters the message message. Speaking ends when the length of the silent period observed in the caller's voice reaches a predetermined end-of-speech detection time, based on the sound detection means for detecting whether or not it is in the middle and the information from the sound detection means. With an utterance end detecting means for detecting a voice call, an interactive answering machine that automatically releases a station line after automatically answering a call, sending a response message and recording a message In the step of changing the utterance end detection time according to the content of the immediately preceding response message, and measuring the elapsed time from when the caller starts uttering based on the information from the voice detection means. An interactive answering machine, comprising: time measuring means, and control means for reducing the utterance end detection time in accordance with the elapsed time after the start of utterance obtained from the elapsed time measuring means.