JP2000106603A - Device and method for voice recognition and response, and voice guidance method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To recognize received voices from plural lines through one voice recognizing part by outputting a voice guidance, recognizing the voice inputted from each call originating terminal in response to that guidance, controlling the output time of the voice guidance outputted to the call originating terminal in accordance with the conditions of interaction with the other call originating terminal and performing recognition processing of voices received from plural call originating terminals in time division. SOLUTION: After call incoming is received on telephone line 1, a voice recognizing/ responding device outputs a guidance voice 1501 to a caller and outputs a guidance voice 1506 to the caller, even as to a telephone line 2 as well. On the side of the telephone line 2, the uttering timing of the caller on the telephone line 1 is predicted and the guidance voice on the side of the telephone line 2 is selected so as not to overlap the uttering timing. When the caller on the side of the telephone line 2 interrupts uttering after listening to the guidance voice only up to the middle, however, a control part performs control so as to temporarily preserve the voice of the caller on the telephone line 2 in a memory and after the end of recognition processing 1:1512, recognition processing 2:1513 is performed.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a voice recognition response apparatus and a voice recognition apparatus for transmitting voice guidance to a partner connected via a communication line and recognizing voice received from the partner in a wired and wireless communication field. The present invention relates to a response method and a voice guidance method.

[0002]

2. Description of the Related Art Conventionally, as a technology for recognizing voice received via a telephone line and providing a guidance service, a "large-scale extension telephone reception system" (Naito et al., Proceedings of the Acoustical Society of Japan)
3-p-27, p215-216, March 1995). In the prior art, one extension number guidance system is connected to one telephone line. The extension number guidance system has a voice recognition unit for recognizing the received voice and a discourse management unit for managing the progress of the dialog between the system and the caller and outputting voice guidance.
Therefore, when providing services to a plurality of lines using this conventional extension number guidance system, it is necessary to prepare the same number of extension number guidance systems as the number of lines. Will have the same number.

One of the purposes of using such a system is to provide information guidance services and private (intra-company) exchanges (hereinafter referred to as P
BX: Private Branch eXchange). In the unmanned information guidance service and automatic exchange service, the caller answers the question from the voice guidance device on the system side so that the caller does not know how to operate and gets confused. And provide services according to the caller's request.
It has been found that the dialogue between the caller and the guidance voice output by the voice guidance device in such a system has the following characteristic tendency.

[0004] While asking from the voice guidance device, the caller listens to the question without speaking.

[0005] The interrogation time from the voice guidance device and the time to answer by the caller are much longer than the interrogation time.

[0006] The voice recognition section of the above system operates only while the caller is speaking, and waits or stops while the voice guidance device is asking the caller. Therefore, the voice recognition unit is in a standby state or a stopped state without operating most of the time.

[0007]

In the above prior art, the voice recognition unit in the extension telephone reception system is in a standby state or a stopped state for most of the time, and there is a problem that the processing efficiency is very low. Was. In addition, since the system is provided with an extension telephone reception system for each line, when providing services for a large number of lines, the same number of devices as the number of lines are required. Therefore, when a large number of lines are targeted, there has been a problem that the system size of the system is increased and the installation cost is increased.

SUMMARY OF THE INVENTION The present invention has been made to solve the above-mentioned problems, and an object of the present invention is to enable a single voice recognition unit to recognize voices received from a plurality of telephone lines so that a small number of voice recognition units can be efficiently used. It is an object of the present invention to provide a voice recognition response device, a voice recognition response method, and a voice guidance method that can be used in a practical manner.

[0009]

In order to solve the above problems, the present invention outputs voice guidance to a plurality of calling terminals connected via a communication line and responds to the voice guidance. An interactive voice recognition / response apparatus that recognizes voice input from each calling terminal, wherein an output time of voice guidance output to each calling terminal is set according to a dialogue state with another calling terminal. A control means for controlling and a voice recognition means for recognizing and processing voices received from a plurality of calling terminals in a time-division manner.

[0010] More specifically, in the above configuration, by changing the wording of the voice guidance, the output time of the voice guidance is controlled so that the time periods for receiving voices from the calling terminals do not overlap. .

In order to solve the above-mentioned problems, voice guidance is output to a plurality of calling terminals connected via a communication line, and voices input from each calling terminal in response to the voice guidance. A voice recognition response device for recognizing a voice, wherein voice recognition means for recognizing and processing voice received from the calling terminal, and another voice when the voice recognition means receives voice from another calling terminal during the recognition process. It has a memory for storing the voice received from the calling terminal, and a memory reading means for reading out the voice stored in the memory and transmitting it to the voice recognition means after the recognition processing in the voice recognition means is completed.

[0012]

Embodiments of the present invention will be described below with reference to the drawings. First, a configuration example of a voice recognition response device of the present invention and a configuration example of a system including the voice recognition response device of the present invention will be described with reference to FIGS. FIG.
FIG. 1 is a diagram showing a configuration of a voice recognition response device of the present invention.
FIG. 2 is a diagram showing a configuration of a voice input / output unit of the voice recognition response device of the present invention. FIG. 3 is a diagram showing a configuration of a system for realizing an automatic extension switching service using the voice recognition response device of the present invention. In the system shown in FIG. 3, when a caller places a call from a telephone 301 or 302 to a switchboard for performing an automatic extension switching service, the caller is connected to a voice recognition response device 307 via a public network 303 and a PBX 304. The voice recognition response device 307 outputs a guidance voice to the caller. When the caller utters the department name or name of the party to be connected according to the guidance voice of the voice recognition response device 307, the voice recognition response device
The 307 recognizes these department names and names, searches for an extension number corresponding to the recognition result from data prepared in advance in the voice recognition response device 307, and notifies the control unit 308 of the PBX 304 of the searched extension number. The PBX 304 performs extension transfer for a call from the caller based on the extension number received from the voice recognition response device 307. As an example of the transfer method at this time, in general, when transferring an extension during a call, the voice recognition response device 307 automatically performs a transfer operation of issuing an extension number after hooking manually performed by a caller. There is a way to

Next, the detailed configuration of the voice recognition response device 307 will be described with reference to FIG. FIG. 1 is a diagram showing a configuration when two lines are connected to one voice recognition response device. The voice recognition response device includes a plurality of line interfaces 101, 102 (hereinafter, line I) for connecting a plurality of telephone lines.
/ F), and audio input / output units 103 and 104 for processing audio input / output from each line I / F, and an input selection unit 105 for selecting audio signals input from a plurality of audio input / output units. An output selection unit 106 for selecting one of a plurality of audio input / output units as an audio output destination, and a memory 108 for storing input audio.
And a memory control unit 114 for controlling reading and writing of the memory 108
A voice recognition unit 107 that performs recognition processing of the voice selected by the input selection unit 105 or the voice stored in the memory 108;
A voice data unit 109 in which guidance voice is stored, a voice data control unit 115 that controls output of guidance voice from the voice data unit, a monitoring unit 110 that monitors a processing status of the voice recognition unit 107 and a voice guidance output state, A recognition data unit 113 that controls each unit and stores recognition data used for the recognition processing of the voice recognition unit 107;
The control unit 111 controls reading of recognition data and prediction data from the prediction data unit 112 in which data for predicting the timing at which the caller utters is stored. Although FIG. 1 shows an example in which two lines are connected, a case where more lines are connected can be handled by increasing the number of voice recognition units 107 and the number of memories 108 according to the number of accommodated lines. If the total number of voice recognition units and the number of memories is the same as the number of accommodated lines, all lines can be handled even if there is an incoming call from all lines at the same time. For example, when accommodating 10 lines, it is also possible to configure with five voice recognition units and five memories, or with three voice recognition units and seven memory units. When the number of accommodated lines is increased, the line I / F and the voice input / output unit are also added according to the increased number of telephone lines.

Next, the operation of each part of the voice recognition response device according to the present invention will be described. The line I / F 101 and the line I / F 102 make line connections to the telephone line 1 and the telephone line 2 respectively. Specifically, the line I / F automatically receives an incoming call from the telephone line, notifies the control unit 111 of the incoming call,
Connect or disconnect the line according to the instruction from 11. Control unit
In accordance with the instruction of 111, operations such as hooking and dialing are also performed.

The audio input / output units 103 and 104 are connected to the line I / Fs 101 and 10
The analog signal received from 2 is converted into a digital signal and transmitted to the input selection unit 105. At this time, an audio signal is detected from the digital signal converted by the instruction of the control unit 111, and the detection status is notified to the control unit 111. The voice input / output units 103 and 104 also convert the guidance voice received from the output selection unit 106 from digital to analog, and
Send to 1,102.

The input selection unit 105 transmits digital voice signals from the voice input / output units 103 and 104 to the voice recognition unit 107 or the memory 108 in accordance with an instruction from the control unit 111. Also, the digital voice signal read by the memory control unit 114 and stored in the memory 108 is transmitted to the voice recognition unit 107. Then, the control unit 111 is notified of the transmission state.

The output selection unit 106 receives the guidance voice read from the voice data unit 109 by the voice data control unit 115 based on the designation from the control unit 111, selects an output destination, and selects the voice input / output unit 103 or the voice input / output unit. The information is transmitted to the output unit 104.

The speech recognition unit 107 is equipped with a word speech recognition system for unspecified speakers. In the voice recognition process, the control unit 111 fetches necessary recognition dictionary data from the recognition data unit 113 and stores the data in the voice recognition unit 10.
7 and then input selector 10 according to the instruction of controller 111.
This is performed by recognizing the digital voice signal input from step 5. After this processing is completed, the voice recognition unit 107 notifies the control unit 111 of the result of voice recognition.

The memory 108 temporarily stores digital audio signals from the audio input / output units 103 and 104. Specifically, when a digital voice signal from the voice input / output unit 104 is input while the digital voice signal from the voice input / output unit 103 is being recognized by the voice recognition unit 107, the voice recognition unit 107 Since the digital audio signal cannot be recognized from the audio input / output unit 104 during use, the input selection unit 105 selects the memory 108,
The digital audio signal from the audio input / output unit 104 is temporarily stored in the memory 108.

The voice data section 109 stores various guidance voices for guiding a caller. A plurality of guidance voices with different guidance lengths are accumulated for the same question content.

The monitoring section 110 monitors the operation states of the speech recognition unit 107 and the memory 108 and notifies the control section 111 of the operation state.

The control unit 111 includes line I / Fs 101 and 102, voice input / output units 103 and 104, an input selection unit 105, and an output selection unit 106
Then, the voice recognition unit 107 and the monitoring unit 110 are controlled. The control unit 111 predicts the utterance timing of the caller sent from the plurality of lines, and uses the same guidance voices having the same contents but different guidance lengths stored in the voice data unit 109 to selectively transmit the guidance voices from the plurality of lines. The utterance timing of the incoming caller is shifted so that the voice recognition processing does not overlap. The database used for prediction is the prediction data section 11
Mount on 2. In addition, recognition target dictionary data required for voice recognition is mounted in the recognition data unit 113.

FIG. 2 is a functional block diagram of the voice input / output units 103 and 104 of the voice recognition / response apparatus of the present invention. The audio input / output unit 103 has an analog / digital signal conversion (A / D) 201
, A digital / analog signal converter (D / A) 202, an echo canceller 203, and a voice detection unit 204. The analog / digital signal converter (A / D) 201 converts an analog signal from the line I / F into a digital signal, and the digital / analog signal converter (D / A) 202 converts the digital signal from the output selector 106 into an analog signal. Convert to a signal. When the line I / F is connected to a digital line, the analog / digital signal conversion (A / D) 201 and the digital / analog signal conversion (D / A) 20
1 is not mounted.

The echo canceller 203 is for preventing a malfunction of the voice detection unit 204 due to the guidance voice from the output selection unit 106 becoming an echo via the telephone line and entering the voice input / output unit 103 again. is there.

The voice detector 204 detects a voice signal from the digital signal from the line I / F and notifies the controller 111 of the voice signal.

Next, the operation of the present invention will be described. First, the utterance timing of the caller is predicted so that the processing of recognizing the voice from the caller sent from a plurality of lines connected to the speech recognition response device, which is a characteristic operation of the present invention, does not overlap. The prediction operation of selecting the guidance voice while doing so will be described with reference to FIGS.

FIG. 4 shows an example of the configuration of data mounted in the prediction data section of the speech recognition response device according to the present invention. As shown in FIG. 4, the data loaded in the prediction data part includes an ID number assigned to each guidance type and a plurality of guidance audio file names corresponding to each guidance audio prepared in a plurality of types for each guidance type. And output time data indicating the guidance length of each guidance voice file for use in the prediction processing.

A specific prediction operation using this data will be described with reference to FIG. FIG. 5 is a diagram showing the relationship between the guidance output state of another line and the prediction result of the prediction target line and a flowchart of the prediction process. The prediction processing is performed by the control unit 111.
Done in FIG. 5A shows the relationship between the guidance output state of another line and the prediction result of the prediction target line, and FIG. 5B shows a flowchart of the prediction process. In FIG. 5A,
At the current time T, the predicted utterance start time of the caller of the other line is T1 + t1, and the predicted utterance end time is T1 + t1 + t2. t1
Is a guidance output time, which can be obtained from the prediction data unit 112 shown in FIG. t2 is a predetermined utterance time of the caller. Approximately 2 to 4 seconds. The output time indicating the length of the i-th guidance voice of the guidance type having a certain ID is t (ID, i). (In the embodiment shown in FIG. 4, three types of guidance voices are prepared for each guidance ID. Therefore, it is expressed as i = 1 to 3). The control unit 111 can obtain the value of t (ID, i) from the prediction data unit 112. For a line to be predicted, a certain ID as guidance output from the current time T
When the i-th guidance voice of the guidance type is used, the predicted utterance start time of the caller is T + t (ID, i) and the predicted utterance end time is T + t (ID, i) + t2 . In the case of this figure, an example is shown in which the speech of the caller of the other line and the line to be predicted differ from each other and the recognition processing does not overlap, so that the recognition can be performed by one voice recognition unit.

Next, the prediction process will be described with reference to FIG. First, the output time of a plurality of files prepared corresponding to the type of guidance target ID is obtained from the prediction data unit 112. In this embodiment, since three types of guidance voices are prepared, t (ID, 1), t (ID, 2), t (ID, 3)
Are obtained (step 501).

Next, an optimum guidance voice satisfying the condition 1 for all the lines other than the line to be predicted is selected from three types. T1 + t1> T + t (ID, i) + t2 of condition 1 is a condition for ending the guidance voice output of the prediction target line and the utterance of the caller before the guidance voice of the other line ends.
T1 + t1 + t2 <T + t (ID, i) is a condition for the utterance of the caller of the line to be predicted to start after the utterance of the caller of the other line ends. If any of these conditions is satisfied, it means that the utterance time does not overlap with other lines (step 502). FIG.
(A) is an example in which the condition of T1 + t1 + t2 <T + t (ID, i) is satisfied, and the utterance of the caller of the prediction target line starts after the utterance of the caller of another line ends. is there.

Next, it is determined in step 502 whether a guidance voice satisfying condition 1 has been obtained (step 50).
3) If requested, end the process (step 50)
Five).

If it is not possible to obtain a guidance voice that does not overlap the uttering timing of the caller,
In order to select a guidance with as little overlap in utterance time as possible, a guidance voice that satisfies condition 2 is obtained. Tb
(i) is the initial overlap time of the utterance time of the line to be predicted, MIN (Tb (1), Tb (2), Tb (3)) is the minimum overlap time of the first, Ta (i) Is the overlap time at the end of the utterance time of the line to be predicted, and MIN (Ta (1), Ta (2), Ta
(3)) is a condition for finding a guidance voice that minimizes the overlap time of the end part. By finding guidance speech that satisfies condition 2 for all the lines other than the prediction target line, it is possible to find the guidance speech with the least overlap of the utterance time with other lines (step 50).
Four). When the guidance voice is obtained, the processing is terminated (step 505).

Next, the speech recognition operation will be described. FIG. 6 shows a recognition data unit 11 of the voice recognition response apparatus according to the present embodiment.
3 shows the data to be mounted. The recognition data unit 113 has data called dictionary data to be used for speech recognition processing. The dictionary data has data 601 describing a department name, and data 602 and 603 corresponding to the name and extension number of each department. There is an incoming call from the line, line I / F101
After activating the voice input / output unit 103, the control unit 111 transfers the dictionary data to be recognized to the voice recognition unit 107 from the recognition data unit 113 to the voice recognition unit 107. In the case of the automatic exchange service, first, the dictionary data 601 in which the name of the department is described is transferred from the recognition data unit 113 to the speech recognition unit 107. Then, after the transfer of the recognition target dictionary data 601 in which the department name is described, the speech recognition unit 107 is activated.
Then, as a first recognition process, a voice uttered by the caller is recognized, and a department name requested by the caller is obtained based on the recognition target dictionary data in which the department name is described. Next, based on the department name of the recognition result, recognition target dictionary data 602 or 603 corresponding to the name belonging to the department and the extension number correspond to the recognition data section 11.
3 to the speech recognition unit 107. Then, as a second recognition process, the voice uttered by the caller is recognized to obtain the extension number of the person requested by the caller.

FIG. 7 shows guidance voice data installed in the voice data unit 109 of the voice recognition response device according to the present embodiment. As shown in FIG. 7, the voice data section stores the ID number, file name, and guidance content of each guidance voice. The name of each file is the same as the file name mounted on the prediction data section 112 shown in FIG.

The prediction process and the recognition process have been described.
Next, the entire processing performed by the control unit 111 in the automatic extension switching service will be described. FIG. 8 shows a process of the control unit 111 of the voice recognition response device according to the present embodiment.
The control unit 111 periodically performs the operation shown in FIG.
Is repeated. The cycle is, for example, 10 ms to 20 ms. The control unit 111 first checks the line status of the line I / F 101 for the telephone line 1 (CH1) (step 80).
1). If it is connected (during a call), the operation state of the voice detection unit 204 of the voice input / output unit 103, the input selection unit 105, and the output selection unit 106 is confirmed (step 802). Subsequently, dialog prediction is performed based on the usage status of each line, and guidance output processing is performed (step 803). If the telephone line 1 is not being connected but is in a connection waiting state, the flow shifts to confirmation of the next telephone line. When the checking of the line status of all the lines is completed in this way, the state of each voice recognition unit and the memory is obtained from the monitoring unit 110, and if there is data temporarily stored in the memory, the input selection unit 105 Connects the vacant speech recognition unit to the memory via the, and performs recognition processing (step 80)
7). The control unit 111 executes the above steps 800 to 808
Repeat the above steps periodically to check the status of all telephone lines. Hereinafter, FIG. 8 will be described with reference to FIGS.
Line I / F confirmation processing (step 801), voice input / output confirmation and recognition processing, and memory processing (step 80)
2) The operation of each step of the guidance selection / output processing (step 803) will be described. Note that the dialogue prediction processing has been described with reference to FIG.

First, the line I / F confirmation processing will be described.
FIG. 9 shows detailed control of the line I / F in the control unit 111 of the voice recognition response device according to the present embodiment. First, the control unit 111 checks a line state when a confirmation process is performed one cycle before a certain line I / F (step 901).
If the line I / F was connected to the telephone line one cycle ago, the current line status is checked to see whether the line is still connected (during a call) or the call has been terminated and the line has been disconnected. (Step 902). If the line is disconnected, the control unit 111 stops the voice input / output processing (step 903), disconnects the line (step 904), and ends the processing (step 9).
08). As a result of checking the current line status in step 902, if a call is being made one cycle ago, the process is terminated as a connection being made (step 909).

In step 901, the line I / F one cycle before
If the telephone is not connected to the telephone line, the current line state is confirmed (step 905). If there is an incoming call, the line is connected to the device (step 906), and voice input / output is activated (step 907). . The line I / F unit connection and connection waiting confirmation processing is performed as described above.

Next, confirmation processing of speech recognition input / output, speech recognition processing, and memory processing will be described. FIG. 10 shows the detailed control for starting the buffering to the voice recognition unit or the memory according to the present embodiment. Here, the processing for the telephone line 1 will be described as an example.

First, it is checked whether the voice of the caller was being input via the telephone line when the voice input / output confirmation, recognition, and memory processing were performed one cycle before (step 100).
1). If the voice is not being input, it is checked whether or not the voice of the caller has been detected by the voice detection unit 204 of the current voice input / output unit 103 (step 1002). Is searched (step 1003). If no voice can be detected in step 1002, the process ends (step 1013). If the voice can be detected and the voice recognition unit can also be searched (step 1004), the input selection unit 105 is instructed by the control unit 111 and the voice input / output unit 103 and the selected voice recognition unit 10 are selected.
7 is connected (step 1005), the digital voice signal input from the line is transmitted to the voice recognition unit 107, and the start of recognition processing is notified to the voice recognition unit 107 in step 1007 (step 1007). If no vacant speech recognition unit is found in step 1004, there is no vacant speech recognition unit, so that the digital speech signal input from the line is temporarily stored in the memory 108 because there is no vacant speech recognition unit. Connects the voice input / output unit 103 and the memory 108 (step 1006), and instructs the start of buffering processing (step 1008).

If the voice was being input one cycle before in step 1001, it is checked whether or not the end of the input voice has been detected in the voice detection unit 204 of the voice input / output unit 103 (step 1009). . If the end of the voice has not been detected yet, the process ends (step 1013). When detected, it is next checked whether or not buffering is being performed in the memory (step 1010). If buffering is being performed, the end of the audio is detected, so buffering in the memory 108 is stopped.
The connection between the voice input / output unit 103 and the memory 108 is released (step 1011). If buffering to the memory has not been performed, the voice recognition unit 107 is stopped, and the connection between the voice input / output unit 103 and the voice recognition unit 107 is released (step 1012).

Next, the guidance output process will be described. FIG. 11 shows the detailed control of the guidance output of the voice recognition response device according to the present embodiment. First, check the status of the conversation on each telephone line (step 1101),
It is checked whether the guidance or the music on hold is being output (step 1102). If the guidance or the holding sound is not being output, it is next checked whether or not the recognition process is being performed (step 1103). If the recognition process is being performed, the process is terminated (step 1110). If the recognition processing is not being performed in step 1003, the dialogue state of the other line is confirmed (step 1104), the caller utterance timing of the own line is predicted, and the optimal guidance voice that provides a smooth dialogue is selected (step 1105). ). It is determined whether the optimal guidance sound has been obtained (step
1106), if selected, control is performed to output the selected guidance (step 1107). If the optimal guidance voice is not selected, a hold sound for temporarily stopping the dialogue is output while the recognition process takes a long time due to the overlapping of the recognition processes (step 1109). The output of the music on hold is stopped, and the process ends (step 1110).

On the other hand, if the guidance or hold sound is being output in step 1102, it is checked whether the hold sound is being output or the guidance is being output (step 1108). Step 1104 and subsequent steps are performed. If the guidance is being output instead of the holding sound, the process ends (step 1110).

Next, a description will be given of a process in the case where there is no vacant speech recognition device and the data stored in the memory is read and recognized. FIG. 12 is a diagram showing the detailed control of the recognition processing of the digital voice signal stored in the memory of the voice recognition response device. First, it is searched whether or not there is a free voice recognition unit (step 1201). Then, the memory storage state is searched (step 1202). It is checked whether or not there is memory storage data waiting to be processed in the memory and whether or not there is an empty voice recognition unit (step 1203). If there is, the memory and voice recognition are connected by input selection (step 1206). ), Start the recognition process (step 12)
07). Then, it is checked whether or not the voice recognition processing has been completed. If the processing has been completed, the voice recognition unit that has completed the processing is stopped (step 1208), and the connection between the memory and the voice recognition is released (step 1209). Control unit 111
Then, the recognition result is transmitted from the stopped speech recognition, and the process ends (step 1210).

If there is no data waiting to be processed in the memory at step 1203, a search is made for a voice recognition unit which is currently performing voice recognition processing on the data stored in the memory (step 1204). It is determined whether or not the processing has been completed (step 1205). If the processing has been completed, the processing from step 1208 on is performed. If the processing has not been completed, the processing ends (step 1210).

In the processing of the control unit 111 shown in FIGS. 9 to 12, the processing for obtaining the line state one cycle before in FIGS. 9 and 10, the processing for searching for an unused speech recognition unit in FIG. For checking the conversation status of the line of FIG.
The process of searching for an available voice recognition unit is determined by referring to a recognition unit management table and a line state management table as shown in FIG.

FIG. 13 is a diagram showing a configuration example of the recognition unit management table and the line state management table. FIG.
In the example shown in FIG. 3, the management table of the recognition unit in FIG. 13A is composed of the operation state of each recognition unit and the data of the connection line, and the management table of the line state in FIG. It consists of line status and conversation status data. The line state indicates the current processing state, and the conversation state indicates the degree of progress of the conversation with the caller. For example, taking the automatic exchange service described above as an example, the stage of outputting the guidance voice for a greeting such as "This is ..." immediately after an incoming call is the interactive procedure 0, the inquiry of the department name and the department. Dialogue procedure 1 refers to the step of performing name recognition processing, dialogue procedure 2 refers to the step of performing name inquiry and name recognition processing, and dialogue procedure 3 refers to the step of outputting connection guidance and performing connection processing. And express the stages of the dialogue. The operation state, the connection line, the line state, and the conversation state may be expressed by a predetermined code.

Next, a specific dialogue example of an automatic exchange service using the voice recognition response device of the present invention will be described with reference to FIGS. FIG. 14 is a diagram illustrating a specific example of a dialog between the caller and the voice recognition response device according to the present embodiment.
In FIG. 14, first, the voice recognition response device is the telephone line 1
After receiving the call, a guidance voice 1401 is output to the caller,
When there is an incoming call on the telephone line 2 as well, a guidance voice 1406 is output to the caller. At this time, the telephone line 2 predicts the utterance timing of the caller of the telephone line 1 and selects the guidance voice of the telephone line 2 so that the utterance timing does not overlap. This method is selected based on the procedure described in FIG. In this case, the guidance voice G1b shown in FIG. 7 is selected from the voice data unit 109 as the guidance voice 1401 of the telephone line 1, and the guidance voice G1c of FIG. 7 is selected as the guidance voice 1406 of the telephone line 2. If the caller utters as predicted, the voice “design part” 1404 of the telephone line 1 user and the voice “material part” 1408 of the telephone line 2 user do not overlap as shown in FIG. Processed in the recognition unit.

FIG. 15 shows another example of the dialogue of the voice recognition response apparatus according to the present embodiment. FIG. 15 shows an operation in the case where voices are simultaneously received from two telephone lines and guidances with different utterance time lengths are output, but the voice recognition processing is overlapped.

After the voice recognition response device receives the telephone line 1,
A guidance voice 1501 is output to the caller and telephone line 2
, A guidance voice 1506 is output to the caller. At this time, the telephone line 2 predicts the utterance timing of the caller of the telephone line 1 and selects the guidance voice of the telephone line 2 so that the utterance timing does not overlap. But,
As a result of the caller of the telephone line 2 interrupting the utterance only halfway through the guidance voice, the voice recognition unit 10
While 7 is recognizing the voice 1504 uttered by the caller of the telephone line 1, the caller of the telephone line 2 also makes a utterance 1510.
In this case, since the voice recognition unit is performing the recognition process 1 (1512), the control unit 111 controls the voice of the caller of the telephone line 2 to be temporarily stored in the memory 108 (151).
6). Then, the recognition processing 1 of the voice recognition unit 107 (151
After the end of 2), the data stored in the memory 108 is read out and the recognition process 2 (1513) is performed. At this time, a hold tone "Pipy ..." or the like (1507) is played to the caller of the telephone line 2 until the recognition processing is completed after the utterance. After the recognition process 2 (1513) ends, the process proceeds to the next dialogue. By this process, even if the speech recognition unit 107 is in use, the speech is temporarily stored in the memory, and after the recognition process is completed, the speech is read out from the memory and the process is performed, so that the conversation can proceed smoothly.

[0050]

According to the present invention, a single voice recognition unit can recognize voices received from a plurality of telephone lines, and a voice recognition response apparatus and a voice recognition response method that can efficiently use a small number of voice recognition units. In addition, a voice guidance method can be provided.

[Brief description of the drawings]

FIG. 1 is a diagram showing a configuration of a voice recognition response device of the present invention.

FIG. 2 is a diagram illustrating a configuration of a voice input / output unit of the voice recognition response device.

FIG. 3 is a diagram showing a configuration of an automatic extension switching service system using a voice recognition response device.

FIG. 4 is a diagram showing a configuration example of data mounted on a prediction data unit of the voice recognition response device.

FIG. 5 is a diagram showing a relationship between a guidance output state of another line and a prediction result of a prediction target line and a flowchart of a prediction process in the voice recognition response device of the present invention.

FIG. 6 is a diagram illustrating a configuration example of data mounted on a recognition data unit of the voice recognition response device.

FIG. 7 is a diagram illustrating a configuration example of guidance voice data mounted on a voice data unit of the voice recognition response device.

FIG. 8 is a flowchart illustrating a process of a control unit of the voice recognition response device.

FIG. 9 shows a line I / O in processing of the control unit of the voice recognition response device.
It is a flowchart figure which shows F confirmation processing.

FIG. 10 is a flowchart illustrating a voice recognition process and a buffering process in a memory among processes of a control unit of the voice recognition response device.

FIG. 11 is a flowchart showing a guidance output process among the processes of the control unit of the voice recognition response device.

FIG. 12 is a flowchart illustrating a voice recognition process of data stored in a memory among processes of a control unit of the voice recognition response device.

FIG. 13 is a diagram showing a configuration example of a recognition unit management table and a line state management table in the voice recognition response device of the present invention.

FIG. 14 is a diagram showing a specific example of a dialogue between a caller and a voice recognition response device according to an embodiment of the present invention.

FIG. 15 is a diagram showing a specific example of a dialog between the caller and the voice recognition response device according to an embodiment of the present invention.

[Explanation of symbols]

101,102… Line interface, 103,104… Audio input / output, 105… Input selection, 106… Output selection, 107…
Voice recognition, 108… Memory, 109… Speech data, 11
0: monitoring, 111: control, 112: prediction data, 113: recognition data, 114: memory control, 115: voice data control, 20
1 ... analog / digital conversion, 202 ... digital / analog conversion, 203 ... echo canceller, 204 ... voice detection, 301,302 ... telephone, 303 ... public network, 304 ... PB
X, 305, 306: extension telephone, 307: voice recognition response device, 30
8: control unit, 601, 602, 603: dictionary data.

 ──────────────────────────────────────────────────続 き Continued on the front page F term (reference) 5K015 AA06 AA07 AD02 AD05 GA04 GA07 5K024 AA76 BB01 BB03 BB04 CC04 DD01 DD02 EE09 FF06

Claims (19)

[Claims] An interactive voice for outputting voice guidance to a plurality of calling terminals connected via a communication line and recognizing voice input from each calling terminal in response to the voice guidance. A recognition response device, comprising: control means for controlling an output time of voice guidance to be output to each of the calling terminals according to a dialogue state with another calling terminal; and receiving from the plurality of calling terminals. And a voice recognition means for performing time-division recognition processing of voice. 2. The control means controls the output time of the voice guidance so as to minimize the overlapping time of the voices received from the calling terminals by changing the output time length of the voice guidance. Claim 1 characterized by the following:
A speech recognition response device according to claim 1. 3. The voice guidance comprises a plurality of steps for collecting information from an operator of each calling terminal through a dialog, and a plurality of voice guidances having different output times are prepared as the above-mentioned steps. The control means predicts the utterance period of the voice input from the other calling terminal from the required output time of the voice guidance being output to the other calling terminal, The voice recognition response device according to claim 1, wherein an optimum voice guidance is selected and output from the voice guidance. 4. A voice recognition / response apparatus for outputting voice guidance to a plurality of calling terminals connected via a communication line and recognizing voice input from each calling terminal in response to the voice guidance. And voice recognition means for recognizing and processing the voice received from the calling terminal, and from the another calling terminal when the voice recognition means receives voice from another calling terminal during the recognition processing. A voice recognition response, comprising: a memory for storing a received voice; and a memory reading means for reading out the voice stored in the memory and transmitting the voice to the voice recognition means after completion of the recognition processing in the voice recognition means. apparatus. 5. An interactive voice for outputting voice guidance to a plurality of calling terminals connected via a communication line and receiving voice input from each calling terminal in response to the voice guidance. An answering device, comprising: a control unit that controls an output time of the audio guidance to be output to each calling terminal according to a dialogue state with another calling terminal. 6. The control means controls the output time of the voice guidance by changing the output time length of the voice guidance so that the overlap time of the voices received from each of the calling terminals is minimized. 6. The method according to claim 5, wherein
The voice response device according to claim 1. 7. The voice guidance comprises a plurality of steps for collecting information from an operator of each calling terminal through a dialogue, and a plurality of voice guidances having different output times are prepared as the above-mentioned steps. The control means predicts the utterance period of the voice input from the calling terminal from the required output time of the voice guidance being output to another calling terminal, and performs the above-described steps for the calling terminal. The voice response device according to claim 5, wherein an optimum voice guidance is selected from a plurality of voice guidances and output. 8. A voice response device which outputs voice guidance to a plurality of calling terminals connected via a communication line and receives voice input from each calling terminal in response to the voice guidance. And receiving processing means for performing a receiving process of the voice received from the calling terminal, and from the another calling terminal when the receiving processing means receives a voice from another calling terminal during the receiving process. A voice response device comprising: a memory for storing the received voice; and a memory reading unit for reading the voice stored in the memory and transmitting the voice to the reception processing unit after the reception processing in the reception processing unit is completed. . 9. An interactive speech recognition for outputting voice guidance to a plurality of callers connected via a communication line and recognizing voices from the respective callers input in response to the voice guidance. In a response method, when outputting voice guidance to a first caller, voice output from the second caller is performed based on a required time for outputting voice guidance being output to a second caller. A voice recognition response method comprising: estimating a time at which the first caller is to be received; and selecting a voice guidance to be output to the first caller based on the prediction result. 10. The voice guidance output to the first caller, wherein the overlap time of the voice received from the first caller and the voice received from the second caller is minimized. 10. The voice recognition response method according to claim 9, wherein the voice guidance is selected. 11. The voice guidance comprises a plurality of steps for collecting information through dialogue with each caller, and a plurality of voice guidances having different required output times are prepared as the guidance for each step. The voice guidance to be output to the first caller by estimating the reception period of the voice input from the second caller from the required output time of the voice guidance being output to the second caller 10. The voice recognition response method according to claim 9, wherein an optimal voice guidance is selected and output from a plurality of voice guidances of each step. 12. A voice recognition response method for outputting voice guidance to a plurality of callers connected via a communication line and recognizing voice input from each caller in response to the voice guidance. And when a voice is received from the second caller during the process of recognizing the voice received from the first caller, the voice received from the second caller is temporarily stored, After the voice recognition processing of the first caller is completed,
A voice recognition response method, comprising reading out the stored voice of the second caller and performing a recognition process. 13. An interactive voice response for outputting voice guidance to a plurality of callers connected via a communication line and receiving voice input from each of the callers in response to the voice guidance. When outputting voice guidance to a first caller, the method outputs voice from the second caller based on a required output time of voice guidance being output to a second caller. A voice response method comprising predicting a reception time and selecting voice guidance to be output to the first caller based on the prediction result. 14. The voice guidance output to the first caller, wherein the overlap time of the voice received from the first caller and the voice received from the second caller is minimized. 14. The voice response method according to claim 13, wherein such voice guidance is selected. 15. The voice guidance comprises a plurality of steps for collecting information through dialogue with each caller, and a plurality of voice guidances having different required output times are prepared as the guidance for each step. The voice guidance to be output to the first caller by estimating the reception period of the voice input from the second caller from the required output time of the voice guidance being output to the second caller 14. The voice response method according to claim 13, wherein an optimal voice guidance is selected and output from the plurality of voice guidances in each of the steps. 16. A voice response method for outputting voice guidance to a plurality of callers connected via a communication line and receiving voice input from each caller in response to the voice guidance. If the voice received from the second caller is received during the process of receiving the voice received from the first caller, the voice received from the second caller is temporarily stored,
A voice response method characterized by reading out the stored voice of the second caller and performing a reception process after the reception process of the voice of the first caller is completed. 17. A voice guidance method for outputting voice guidance to a caller connected via a communication line, wherein a plurality of voice guidances having different required output times are prepared for the same guidance content. A voice guidance method characterized by selecting and outputting a voice guidance having an optimum output time from a plurality of voice guidances to a caller. 18. The voice guidance method according to claim 17, wherein the plurality of voice guidances having different required output times are voice guidances expressing the same guidance content in different terms. Guidance method. 19. The voice guidance method according to claim 17, wherein, when another caller is connected via a communication line, said other caller. A voice guidance method characterized by selecting a guidance voice to be output based on a guidance output status to the user.