JPH07191690A - Minutes generation device and multispot minutes generation system - Google Patents

Abstract

PURPOSE:To provide the minutes generation device which can generate accurate minutes easily, speedily, and automatically without requiring a dedicated recording person. CONSTITUTION:Speech signals of a 1st and a 2nd speakers which are inputted through microphones 109 and 110 have noise components removed by a noise processing part 102 and are converted by a speaker detection part 103 into speaker codes corresponding to the vocalizing speakers; and the speaker codes are stored as speech data in a speech data memory part 104 together with the speech signals. The speech data are converted by a speech recognition part 105 into character codes, which are further converted by a system controller 107 into speaker's names that the speaker codes in the speech data correspond to. The character codes and speaker's names are outputted to the outside through an output part 106 to generate minutes.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a minutes creating apparatus for creating minutes of ordinary meetings and a multipoint minutes creating system for creating minutes of meetings by a multipoint video conference system or the like.

[0002]

2. Description of the Related Art Conventionally, when a minutes is created, a professional recorder (clerk) records the speech of a speaker by handwriting at the same time as the conference progresses, or records it on a recording device such as a tape recorder, and the conference is recorded. After the completion, a method of creating a minutes by editing using a word processor was adopted.

[0003]

However, according to the above-mentioned conventional method, there is a problem that a specialized recorder having some experience is required to create the minutes. In addition, in the case of recording by handwriting, a lot of time and labor is spent for the recording, and in the case of recording by a recording device, it is necessary to reproduce the voice recorded after the end of the conference to create the minutes. However, there was a problem that it took a lot of time to create the minutes.

Further, when the minutes of a conference are created by a multipoint video conference system, it is difficult to judge who the voices of speakers at a plurality of points are, and it is often necessary to create the minutes. There was a problem of spending time.

The present invention has been made to solve the above problems, and its first object is to make it possible to accurately and easily record accurate minutes without requiring a professional recorder.
Moreover, it is to provide a minutes creating device which can be created automatically.

A second object of the present invention is to provide a minutes preparation apparatus capable of easily, quickly and automatically creating accurate minutes of a meeting by a multipoint video conference system. To provide.

[0007]

In order to achieve the above first object, the first invention (Claim 1) of the present invention is such that a voice signal of a speaker is inputted and the voice signal is converted into the voice signal. A voice signal input unit having a speaker identification unit for converting into a speaker code for identifying a corresponding speaker, and a speaker output from the speaker identification unit in a voice signal input from the voice signal input unit. Voice data storage means for storing voice data added with a code, voice recognition means for converting voice signals in the voice data stored in the voice data storage means into character codes, and storage in the voice data storage means The present invention is characterized by comprising an output means for outputting a speaker code in the generated voice data and a character code output from the voice recognition means.

In order to achieve the same object, a second invention of the present invention (claim 2) is that a voice signal of a speaker is input and the voice signal identifies a speaker corresponding to the voice signal. Voice signal input means having a speaker identification section for converting into a speaker code for use, a voice signal storage means for storing a voice signal input from the voice signal input means, and a speech output from the speaker identification section. Speaker code storage means for storing a person code, voice recognition means for converting a voice signal stored in the voice signal storage means into a character code, character code output from the voice recognition means and the speaker code conversion Output means for outputting the speaker code output from the means.

In order to achieve the same object, a third aspect of the present invention (claim 5) is that a voice signal of a speaker is input and the voice signal identifies the speaker corresponding to the voice signal. For inputting from a voice signal input means having a speaker identifying unit for converting into a speaker code, a clock means for measuring a generation time of a voice signal input from the voice signal input means, and the voice signal input means. A voice data storage unit for storing voice data obtained by adding a speaker code output from the speaker identification unit to the voice signal, and a voice signal in the voice data stored in the voice data storage unit as a character code. A voice recognition means for converting, a character code storage means for storing a character code output from the voice recognition means, a display means for displaying the character code stored in the character code storage means as a character, and the display hand. Editing means for editing the content displayed by, a speaker name registration means for previously registering the name of a speaker corresponding to the speaker code, and a speaker code stored in the speaker code storage means for the speaker code. A speaker name conversion means for converting the speaker name registered in the speaker name registration means, a speaker name output from the speaker name conversion means into a character code output from the voice recognition means, and the timekeeping. The present invention is characterized by further comprising output means for adding and outputting a generation time of an audio signal timed by the means and a device number for specifying the device.

In order to achieve the same object, a fourth invention of the present invention (claim 6) is to input a voice signal of a speaker and to identify the speaker corresponding to the voice signal. Voice signal input means having a speaker identification section for converting into a speaker code, time measuring means for measuring a generation time of a voice signal input from the voice signal input means, and voice input from the voice signal input means. A voice signal storage unit for storing a signal, a speaker code storage unit for storing a speaker code output from the speaker identification unit, and a voice for converting a voice signal stored in the voice signal storage unit into a character code. Recognition means, character code storage means for storing the character code output from the voice recognition means, display means for displaying the character code stored in the character code storage means as characters, and display by the display means Editing means for editing the stored content, speaker name registration means for registering the name of the speaker corresponding to the speaker code in advance, and the speaker code stored in the speaker code storage means for the speaker name. The speaker name conversion means for converting the name of the speaker registered in the registration means, the speaker name output from the speaker name conversion means into the character code output from the voice recognition means, and the timekeeping means. The present invention is characterized by comprising an output means for adding and outputting the time of generation of the timed audio signal and the device number for specifying the device.

In order to achieve the same purpose, a speaker name registration means for registering the name of the speaker corresponding to the speaker code in advance, and a speaker in the voice data stored in the voice data storage means. Speaker name conversion means for converting a code into a speaker name registered in the speaker name registration means, and the output means converts the speaker name conversion means into a character code output from the voice recognition means. It is desirable to add the speaker name output from (1) to the output (Claim 3).

To achieve the same object, a character code storage means for storing a character code output from the voice recognition means, and a display means for displaying the character code stored in the character code storage means as a character. And editing means for editing the content displayed by the display means (claim 4).

Further, in order to achieve the second object, the fifth invention (Claim 7) of the present invention arranges the minutes preparing apparatus according to the third or fourth invention at multiple points, and
The multipoint editing means for compiling and editing the output data output from the output means of the minutes recording device at each of the points is provided, and the multipoint editing means measures the output data by the timekeeping means. It is characterized in that the minutes of the multipoint conference are created by rearranging the audio signals based on the generation time of the audio signals.

In order to achieve the same object, a sixth invention (Claim 8) of the present invention has the minutes preparing apparatus according to the third or fourth invention arranged at multiple points, No multipoint editing means for editing the output means output from the output means of the minutes creating device of one of the minutes creating devices of one of the minutes creating devices in one place. The minutes of the multipoint conference are created by rearranging the output data by the multipoint editing means based on the generation time of the audio signal timed by the timekeeping means.

[0015]

According to the minutes creating apparatus of the first invention of the present invention,
The voice recognition unit converts the voice signal in the voice data stored in the voice data storage unit into a character code, and the speaker identification unit converts the voice signal of the speaker input from the voice signal input unit into a speaker code. The speaker code in the voice data stored in the voice code storage means and the character code is output by the output means.

According to the minutes creating apparatus of the second aspect of the present invention, the voice signal stored in the voice signal storage means is converted into the character code by the voice recognition means, and the voice signal input means inputs the voice signal by the speaker identification section. The voice signal of the speaker thus converted is converted into a speaker code, and the character code and the speaker code stored in the speaker code storage means are output by the output means.

According to the minutes producing apparatus of the third invention of the present invention, the voice signal in the voice data stored in the voice data storage means is converted into the character code by the voice recognition means, and the voice signal is outputted by the speaker identifying section. The voice signal of the speaker input from the input means is converted into a speaker code, the speaker name corresponding to the speaker code is registered in advance by the speaker name registration means, and the speaker code is converted by the speaker name conversion means. The speaker code stored in the storage means is converted into the name of the speaker registered in the speaker name registration means, the character code stored in the character code storage means is displayed by the display means, and the display means is displayed by the editing means. The content displayed by is edited, and the output unit displays the name of the speaker output from the speaker name conversion unit to the edited character code, the generation time of the audio signal measured by the time measuring unit, and the device. It is output in a state where a constant to device number appended.

According to the minutes creating apparatus of the fourth aspect of the present invention, the voice signal stored in the voice signal storage means is converted into the character code by the voice recognition means, and input from the voice signal input means by the speaker identification section. The voice signal of the speaker thus converted is converted into a speaker code, the speaker name corresponding to the speaker code is registered in advance by the speaker name registration means, and stored in the speaker code storage means by the speaker name conversion means. The converted speaker code is converted into the name of the speaker registered in the speaker name registration means, the character code stored in the character code storage means is displayed by the display means and displayed by the editing means by the display means. The name of the speaker whose contents are edited and output from the speaker name conversion unit to the character code after being edited by the output unit, the generation time of the voice signal measured by the time counting unit, and the device number for identifying the device. There are output by the additional state.

According to the multipoint minutes preparation system of the fifth aspect of the present invention, the third or fourth points arranged at multiple points.
The output data output from the output means of the minutes creating apparatus relating to the invention of claim 1 is edited by the multipoint editing means in a state where the output data is collected in one place, and the multipoint editing means measures the output data by the timekeeping means. The minutes of the multipoint conference are created by rearranging them according to the generation time of the generated audio signal.

According to the minutes creating system of the sixth invention of the present invention, the output data output from the output means of the minutes creating apparatus according to the third or fourth invention arranged at multiple points is multipoint. One of the minutes creating devices is edited by the editing means of one of the minutes creating device in a state of being gathered in one place, and the output data is edited by the editing means, and the output time is generated by the time counting means. The minutes of the multipoint meeting are created by being sorted based on the.

[0021]

Embodiments of the present invention will be described below with reference to the drawings. (First Embodiment) First, a first embodiment of the present invention will be described with reference to FIGS.

FIG. 1 is a block diagram showing the schematic arrangement of a minutes creating apparatus according to this embodiment.

In the figure, reference numeral 101 is a voice signal input section for directly inputting the voice of the speaker as a voice signal. The voice signal input unit 101 detects a speaker that has input the voice signal from a plurality of speakers, and a noise processing unit 102 that removes noise included in the input voice signal. And a speaker detection unit 103 that outputs a speaker code corresponding to. The noise processing unit 102 is controlled by the system controller described later.

The output sides of the noise processing unit 102 and the speaker detection unit 103 are connected to the voice data memory unit 104,
Voice data obtained by adding the speaker code output from the speaker detecting unit 103 to the voice signal noise-processed by the noise processing unit 102 is stored in the voice data memory unit 104.

The output side of the voice data memory unit 104 is connected to the voice recognition unit 105 and the system controller 107. The voice recognition unit 105 recognizes a voice signal in the voice data sent from the voice data memory unit 104, converts it into a character code corresponding to the recognized voice signal, and outputs the character code. Further, in the voice recognition unit 105, the BUSY flag is set when the voice signal is being processed, and is cleared when the process is completed. Voice data memory unit 104
The transfer of data from the voice recognition unit 105 to the voice recognition unit 105 is controlled by the system controller, and only when the BUSY flag is cleared, the voice data memory unit 10
4 is permitted to transfer.

The system controller 107 determines the speaker code in the voice data stored in the voice data memory unit 104 based on the speaker name registered in the speaker name registration unit 108. Convert to name and output part 10
Go to 6. Further, the system controller 107 is connected to the voice input unit 101, the voice recognition unit 105, and the output unit 106, and controls their operations. The character code output from the voice recognition unit 105 is the output unit 106.
At, the speaker name recognized by the system controller 107 is added, and then output from the output unit 106 to a printing device such as a printer or a display connected to the outside.

FIG. 2 is a block diagram showing a schematic configuration of the noise processing unit 102 shown in FIG. In the figure, a noise processing unit 102 includes a microphone 111 that picks up ambient noise as a reference signal, microphones 109 and 111,
An analog for discretizing a signal input via 110-
A digital converter (hereinafter referred to as an A / D converter) 112,
113 and 114, and an adaptive filter 115 that filters out a noise signal from the voice signal input via the microphones 109 and 110. The adaptive filter 115 is connected to the output sides of the A / D converters 112 to 114, and based on a reference signal input from the A / D converter 113 as described later, the A / D converter 11
The noise of the voice signal input from 2, 114 is removed.

FIG. 3 is a block diagram showing a schematic configuration of the speaker detecting unit 103 shown in FIG. The speaker detection unit 103
Analog-digital converters (hereinafter referred to as A / D converters) 116 and 117 for discretizing an input signal, a level detection unit 118 for detecting the level of a voice signal, and a multiplexer 119 for distinguishing a speaker by switching. When,
The multiplexer 119 comprises a speaker code generator 120 for generating a speaker code corresponding to the speaker switch turned on. The level detection unit 118 stores in advance the input levels of the audio signals s1 and s2 input from the microphones 109 and 110 in order to detect whether or not the first speaker and the second speaker have spoken. Whether or not it is equal to or larger than the threshold value is determined. Further, the multiplexer 119 turns on the speaker switch corresponding to the speaker whose voice level has been detected by the level detector 118.

Next, the minutes creating process performed by the minutes creating apparatus having the above configuration will be described in detail with reference to FIG. FIG. 4 is a flowchart showing the minutes creating operation of the minutes creating apparatus according to the present embodiment.

Now, consider the case where the first speaker speaks.
The voice signal s1 of the first speaker is input to the noise processing unit 102 and the speaker detection unit 103 via the one microphone 109 (step S401). Next, the noise processing unit 102
Then, noise removal processing is performed on the input voice signal (step S402), and the speaker detection unit 103 detects the speaker based on the input voice signal (step S403).

The noise processing performed in step S402 is performed according to the flowchart shown in FIG.

The audio signal input to the A / D converter 112 is converted to a digital signal and then input to the adaptive filter 115 (step S501).

The audio signal s1 is, as shown in FIG.
It is a signal (s1 + n1) on which the noise component n1 is superimposed. In order to remove this noise component n1, the noise input to the microphone 111 is taken into the adaptive filter 115 as the reference signal n (step S502), and the filter coefficient is calculated based on the reference signal n and the previous output audio signal o1. (Steps S503 and S504). In the adaptive filter 115, the adaptive filter processing is performed on the audio signal (s1 + n1) using the calculated filter coefficient (step S505), and the new audio signal o1 from which the noise n1 is removed is sent to the audio data memory unit 104 in the subsequent stage. This is output and this processing operation is ended (step S50).
6). The new audio signal o1 is fed back to the adaptive filter 115 in FIG. 2 and used for the next filter coefficient calculation.

On the other hand, the speaker detection process performed in step S403 of FIG. 4 is performed according to the flowchart shown in FIG. First, the audio signal A / D converted by the A / D conversion units 116 and 117 is input to the level detection unit 118 (step S601). Here, the level detection of the voice signals of the first and second speakers is performed. Specifically, it is determined whether or not the input levels of the audio signals s1 and s2 input from the microphones 109 and 110 are equal to or higher than a threshold value stored in advance (step S602), and the answer is negative (NO). If it is, the above step S601
And step S602 are repeated to enter a standby state. When the answer in step S602 is affirmative (YES), it is confirmed that an utterance has been made. As mentioned above,
Here, since the case where only the first speaker utters is considered, it is determined that only the level of the voice signal of the first speaker is equal to or higher than the threshold value, and the utterance of the first speaker is confirmed. Next, the multiplexer 119 turns on the speaker switch of the speaker (first speaker) whose voice signal level has been detected (step S603), and the speaker code generator 120 of FIG.
An audio signal will be input to. As a result, the on-state of the speaker switch is detected by the speaker code generator 120, and the speaker code corresponding to the switch is output to the subsequent circuit (step S604). Here, as the speaker code, for example, the speaker code of the first speaker is set to "1" and the speaker code of the second speaker is set to "2".

Returning to FIG. 4, step S604 of FIG.
The speaker code output in step (1) is stored in the voice data memory unit 104 in the format shown in FIG. 7 as voice data together with the voice signal o1 that has been noise-processed by the noise processing unit 102 in FIG. 2 (step S404). One voice data is
It ends when an unvoiced sound portion is detected, and an end code (EOD) is added to the end of the voice data.

Here, the system controller 10 of FIG.
7, it is determined whether or not the BUSY flag of the voice recognition unit 106 is cleared (step S405). When the answer is negative (NO), the voice recognition unit 106 finishes the previous voice data recognition process. Not so
It becomes a standby state. If the answer to step S405 is affirmative (YES), the system controller 107 permits the audio data memory unit 104 to transfer audio data.

The voice data stored in the voice data memory unit 104 in step S404 is transferred to the voice recognition unit 105 and the system controller 107 of FIG. 1 one by one, and converted into a character code and a speaker code. (Step S406). That is, the voice recognition unit 105 converts the input voice data into a corresponding character code.
Further, in the system controller 107, by referring to the speaker name registered in advance in the speaker name registration unit 108 of FIG.
It is converted into a speaker name corresponding to the speaker code.

When the speaker name is added to the character code by the system controller 107 and output from the output unit 106 of FIG. 1 to the outside (step S40).
7) The voice data stored in the memory unit 104 is released from the memory unit (step S408), and the minutes creating process for one voice data ends.

When the second speaker utters, the voice signal s2 becomes a signal (s2 + n2) on which the noise component n2 is superimposed, as shown in FIG.
After being input to the noise processing unit 102 from 0, the same processing as that performed when the first speaker speaks is performed, and then the character code and the speaker name are output from the output unit 106.

As described above, according to this embodiment,
Since the speaker's voice is recognized and the recognized voice signal is automatically converted into a character code, it is possible to easily and quickly, without handwriting or recording the conference contents with a recording device or a recording device. Moreover, minutes can be created automatically. Further, since the speaker is discriminated using the speaker code at the same time as the voice recognition, the speaker can be surely recognized. (Second Embodiment) Next, a second embodiment of the present invention will be described with reference to FIG. Since the basic configuration of the minutes creating apparatus in this embodiment is the same as that of FIG. 1 of the above-described first embodiment, the same drawing will be used for the description.

FIG. 8 is a block diagram showing a schematic configuration of the noise processing unit 102 used in the minutes creating apparatus according to this embodiment. In the figure, the same components as those in FIG. 2 of the first embodiment described above are designated by the same reference numerals.

8 is different from that shown in FIG. 2 in that the microphones 109 and 110 and the A / D converters 112, 112, except for the microphone 111 and the A / D converter 113 that take in ambient noise as a reference signal from the configuration of FIG. Amplifiers 121 and 1 for amplifying the audio signals s1 and s2 between
22 is provided and the adaptive filter is divided into first and second adaptive filters 115a and 115b. The other configuration is the same as that of FIG. 2, and the description thereof is omitted.

In the present embodiment, for example, when two people speak at the same time, as shown in FIG. 8, the signal i1 input from one of the microphones 109 is the voice signal s1 of the first speaker and the second speaker. Voice signal s2 is a signal superimposed as a noise signal n2, and a signal i2 input from the other microphone 110 is a voice signal s2 of the second speaker and a voice signal s1 of the first speaker superimposed as a noise signal n1. It is a signal that has been.

The input signals i1 and i2 are input to the amplifiers 121 and 12
2 is amplified and converted into digital signals by the A / D converters 112 and 114, and then the first adaptive filter 1
15a and the second adaptive filter 115b.

In order to remove the noise signal n2 from the digitalized input signal i1, first, the output signal o
1 is used to calculate and update the filter coefficient of the second adaptive filter 115b. The noise component e1 included in the input signal i1 is calculated based on the filter coefficient and the signal i2 input to the second adaptive filter 115b. Then, by removing the noise component e1 calculated from the input signal i1, the voice component of the first speaker in the input signal is extracted and output as the output signal o1.

Similarly, the filter coefficient of the first adaptive filter 115a is calculated and updated using the previous output signal o2, and the input is made based on the filter coefficient and the input signal i1 to the first adaptive filter 115a. The noise component e2 included in the signal i2 is calculated. Then, by removing the noise component e2 calculated from the input signal i2, the voice component of the second speaker in the input signal is extracted, and the output signal o2
Is output as.

The two output signals o1 and o2 are stored in the voice data memory unit 104 as voice data together with the speaker code output from the speaker detecting unit 103 in FIG. This voice data is processed in the same manner as in the first embodiment described above, converted into character data and a speaker name, and output from the output unit 106.

As described above, according to this embodiment,
Even when two people speak at the same time, it is possible to remove the voices of the other party who become noises with each other. Therefore, in addition to the effects of the first embodiment described above, it is possible to more reliably recognize the speaker. (Third Embodiment) Next, a third embodiment of the present invention will be described with reference to FIG. Since the basic configuration of the minutes creating apparatus in this embodiment is the same as that of FIG. 1 of the above-mentioned first embodiment, the description will be made by diverting this drawing.

FIG. 9 is a block diagram showing the schematic arrangement of the minutes creating apparatus according to this embodiment. In the figure, the same parts as those in FIG. 1 of the first embodiment described above are designated by the same reference numerals. 9 is different from FIG. 1 in that the voice data memory unit 104 is deleted from the configuration of FIG. 1 and a voice signal memory unit (voice signal storage means) 123a for storing a voice signal which is binary data is replaced with text data. And a speaker code memory unit (speaker code storage means) 123b for storing the speaker code. Since the other structure is the same as that of FIG. 1 of the first embodiment,
Detailed description thereof will be omitted.

In this embodiment, the voice signal from which the noise is removed by the noise processing unit 102 is stored in the voice signal memory unit 123a as it is. One voice signal ends when an unvoiced sound portion is detected, and a termination code is added to the end of the voice signal. On the other hand, the speaker code output from the speaker detection unit 103 is stored in the speaker code memory unit 123b. Similar to the voice signal, the termination code is added to the end of the speaker code at the same timing as in the voice signal.

The voice signal stored in the voice signal memory unit 123a is converted into a character code by the voice recognition unit 105, and the speaker code stored in the speaker code memory unit 123b is converted into an actual speaker by the system controller 107. Is converted to a name.

The character code and the speaker name are output by the output unit 1.
Output from the audio signal memory unit 1
The voice signal of 23a and the speaker code of the speaker code memory unit 123b are released from the memories 123a and 123b.

Even if the memory unit for storing the voice signal and the memory unit for storing the speaker code are separately provided as in this embodiment, they are easy and quick as in the first and second embodiments. In addition, it is possible to automatically create accurate minutes.

In this embodiment, the speaker code output from the speaker detecting unit 103 is stored in the speaker code memory unit 123b.
However, the speaker code may be directly sent to the system controller 107 without providing the speaker code memory unit 123b. (Fourth Embodiment) Next, a fourth embodiment of the present invention will be described with reference to FIGS.

FIG. 10 is a block diagram showing the schematic arrangement of a minutes creating apparatus according to this embodiment. In the figure, the same parts as those in FIG. 1 of the first embodiment described above are designated by the same reference numerals.

10 is different from FIG. 1 in that a character code memory unit 124 (character code storage means) is provided between the voice recognition unit 105 and the output unit 106 of FIG. 1, and the character code memory unit 124 has a CRT. That is, the editing unit 125 having the display unit 126 (display unit) such as is connected. The rest of the configuration is the same as that of FIG. 1, and therefore its explanation is omitted. Here, the character code memory unit 124 is connected to the system controller 107, and its operation is controlled by the system controller 107.

The minutes creating process of the minutes creating apparatus having the above configuration is performed according to the flowchart shown in FIG.

In the figure, the processing from step S1101 to step S1106 is the same as the processing from step S401 to step S406 shown in FIG. 4 of the above-mentioned first embodiment, and therefore its explanation is omitted. As described above, one piece of voice data ends when an unvoiced sound portion is detected, and a termination code is added to the end of the voice data.

The character code output from the voice recognition unit 105 of FIG. 10 in step S1106 is added with the speaker name by the system controller 107 and stored in the character code memory unit 124 (step S1107). Since the stored character code may have been erroneously recognized, the character code stored in the character code memory unit 124 is displayed on the display unit 126 to edit the character code erroneously recognized by the operator. (Step S1108). The edited character code is stored again in the character code memory unit 124 and output from the output unit 106 (step S1).
109).

Such a minutes creating apparatus will be described with a specific example.

Now, assume that the following conversation is performed between the first speaker and the second speaker.

First speaker: I am A. Are you B?

Second speaker: No.

First speaker: Are you C?

Second speaker: Yes, that's right.

At this time, the voice signal whose noise has been reduced by the noise processing unit 102 of FIG. 10 is stored as voice data in the voice data memory unit 104 of FIG. 10 together with the speaker code in the format shown in FIG. . This voice signal is 1
The data is transferred to the voice recognition unit 105 one by one, recognized in phoneme units, converted into a character code, and output.

On the other hand, the speaker code in the voice data is converted by the system controller 107 into the speaker name registered in advance in the speaker name registration unit 108 of FIG. 10 in the format shown in FIG. 13, for example. The character code is added to the speaker name, transferred to the character code memory unit 124, and stored therein. The stored character codes are sequentially displayed on the display unit 126 which is, for example, the display of the editing unit 125 of the computer. An example of the display format at this time is shown in FIG.
4 shows.

Here, the editor corrects a recognition error recognized in the conversation displayed on the display unit 126 and converts it into a kanji or alphabet. In the case of FIG. 14, since "ha" is mistakenly recognized as "wa", it is changed to "ha". Furthermore, the kana characters "e", "bii", and "shii" are the alphabetical characters "A". , "B",
Converted to "C". In this way, the editing unit 125 performs the process of converting or adding a punctuation mark to make it easier to read by the editor. Then, this result is output by the output unit 106.
Is output to the outside in the format shown in FIG.

As described above, according to this embodiment, the editing unit 125 having the display unit 126 is provided in the configuration of the above-described first to third embodiments, and the editor makes a voice recognition error or conversion into Kanji or alphabet. As described above, it is possible to create an accurate minutes easily, quickly and automatically as in the first to third embodiments, and also to create an even more accurate and readable minutes. Become.

In this embodiment, the display unit 126 and the editing unit 125 are provided in the minutes creating apparatus shown in FIG. 1 of the above-mentioned first embodiment, and the editor recognizes a character code recognition error, a kanji character or an alphabet. Although the conversion can be performed, the editing unit 12 is added to the configurations of the second and third embodiments described above.
It goes without saying that the number 5 and the display unit 126 may be added. (Fifth Embodiment) Next, a fifth embodiment of the present invention will be described with reference to FIG. FIG. 16 is a block diagram showing the configuration of the minutes creating apparatus according to the present embodiment. In FIG. 16, the same parts as those in FIG. 10 of the above-described fourth embodiment are designated by the same reference numerals. 16 is different from FIG. 10 in that the speaker detection unit 103 and the noise processing unit 10 of the voice signal input unit 101 are different.
2 is connected in series.

In this case, the input audio signal is as shown in FIG.
A / D converters 116 and 11 in the speaker detector 103 shown in FIG.
Since it has been converted to a digital signal by the A.D. 7, the A / D conversion units 112 and 114 in the noise processing unit 102 shown in FIG.
Can be deleted. (Sixth Embodiment) Next, a sixth embodiment of the present invention will be described with reference to FIGS. The present embodiment is a multipoint minutes recording system in which the minutes recording devices shown in the above-described first to fifth embodiments are arranged at multiple points to create the minutes of a conference by a multipoint video conference system. It is a thing.

FIG. 17 is a conceptual diagram for explaining the multipoint video conference. In this embodiment, the main venue, A
Video conferences are held at three locations, venue B and venue B, and there are two attendees at each venue. For the sake of convenience, the speaker names are "a" and "b" for the two people in the main venue, and "ha" for the two people in the A venue.
Let's say "H" and "H" for the two people in "D" and B venue. Also,
The device numbers of the minutes creating devices installed in the main venue are "1", and the device numbers of the minutes creating devices installed in A venue and B venue are "2" and "3", respectively. Since the configuration of the minutes creating device of each venue (point) used in this embodiment is the same as that in FIG. 16 of the fifth embodiment described above,
The same drawing will be used for description.

Further, the voice data generation time added to the character code generated by each minutes preparation device is recorded up to hour / minute / second.
The numbers 1, 2, 3, ... Are recorded from the earliest generation time.

As shown in FIG. 18, the above-mentioned three minutes preparation devices are provided with a multipoint editing device (multipoint editing means) 21 via a public line ISDN (Integrated Services Digital Network).
It is connected to 0. FIG. 19 shows the multipoint editing apparatus 2
10 is a block diagram showing a schematic configuration of a public line IS.
It is composed of a data buffer 211 which is an input unit of data generated by a plurality of minutes preparation devices connected to the DN, a system controller 212, and an output unit 213.

In the above configuration, the minutes of the multipoint video conference are performed according to the flowchart shown in FIG.

It is assumed that each attendee introduces himself as follows.

First speaker: I am Lee.

Third speaker: I'm Ha.

Fifth speaker: I'm ho.

Second speaker: I'm Ro.

Fourth speaker: I'm Ni.

Speaker 6: I'm F.

In step S2001 to step S2008 in the figure, the same processing operation as step S1101 to step S1108 of FIG. 11 of the fourth embodiment described above is performed by the minutes creating device installed in each venue. .

That is, the voice of each speaker is converted into the noise processing unit 102 and the speaker detection unit 103 of the minutes creating devices 1 to 3 of each venue.
(Step S2201). Next, the noise processing unit 102 performs noise processing on the input voice signal (step S2202), and the speaker detection unit 103 outputs a speaker code based on the input voice signal (step S2203). . The noise-processed voice signal is
The speaker code is added and stored as voice data in the voice data memory unit 104 (step S2204).
When the system controller 107 determines that the BUSY flag of the voice recognition unit 106 is cleared (step S2205), the voice data is transferred from the voice data memory unit 104 to the voice recognition unit 105 and the system controller 107. , Respectively, the character code and the speaker name are output (step S2206). The output character code is added with the speaker name and the generation time of the voice signal and stored in the character code memory unit 124 (step S2207). The character code stored in the character code memory unit 124 is displayed on the display unit 126, and the editing unit 12
Editing processing is performed in step 5 (step S220).
8).

Here, the minutes data (output data) obtained by performing the editing process in each minutes creation device is:
From the output unit 106, via the public line ISDN, is sent to the multipoint editing device 210 installed in the main venue,
The data is stored in the data buffer 211 in the multipoint editing apparatus 210 (step S2009). Then, the data stored in the data buffer 211 is rearranged in order from the earliest generation time based on the audio signal generation time of the header added to the data by the system controller 212 (step S2010). The data edited in the correct order is output as the minutes from the output unit 213 to a printer or the like connected to the outside to generate the minutes (step S2011).

As described above, according to the multipoint minutes preparation system of this embodiment, the minutes preparation devices of the above-mentioned first to fifth embodiments are arranged at multiple points, and the public lines are connected to the respective minutes preparation devices. By connecting the multipoint editing apparatus 210 via ISDN, it becomes possible to easily and quickly create an accurate minutes, even when a conference is held at a plurality of points. . (Seventh Embodiment) Next, a seventh embodiment of the present invention will be described with reference to FIGS. In the sixth embodiment described above, the character code is created by the minutes creating device installed in each venue and the multipoint editing device 2 installed in the main venue.
The minutes are edited and output according to 10. However, in this embodiment, the minutes are stored in the voice data memory unit 104 of FIG. The sixth embodiment is different from the sixth embodiment in that the processing up to the addition of the header to the voice data is performed, and the character encoding, editing, and output of this data are performed by the minutes creating device of the main venue.

21 and 22 are block diagrams showing a schematic configuration of the minutes creating apparatus according to the present embodiment. In both figures, the same parts as those of FIG. 16 of the sixth embodiment described above are designated by the same reference numerals. It is attached. FIG. 21 shows a minutes creating device installed in the A and B venues (hereinafter referred to as sub-venues), and FIG. 22 shows a minutes creating device installed in the main venue.

In FIG. 21, a minutes creating apparatus (hereinafter referred to as a subsystem) installed in a sub-venue has a configuration in which the speech recognition unit 105 and the subsequent parts shown in FIG. 16 showing the fifth embodiment described above are omitted. Audio data memory unit 10
4 is directly connected to the public line ISDN. Further, in FIG. 22, the minutes recording device (hereinafter referred to as the main system) installed in the main venue connects the public line ISDN to the voice data memory unit 104 shown in FIG. It is configured to temporarily store the voice data sent from it. In addition,
The audio data memory unit 10 of the main system shown in FIG.
It is preferable that the memory capacity of 4 is large so that the data sent from the subsystem of FIG. 21 can be stored. 21 and 22, the configuration other than the above is the same as the configuration shown in FIG. 16, and thus detailed description thereof will be omitted.

FIG. 23 is a block diagram for explaining a multipoint minutes preparation system composed of the main system of FIG. 21 and the subsystem of FIG. In the figure, the multipoint minutes preparation system is configured by connecting the main system and the subsystem via a public line ISDN.

In the above structure, the minutes of a multipoint video conference or the like are created according to the flowchart shown in FIG.

The conversation conducted at the sub-venue is input to the subsystems (apparatus numbers 2 and 3) installed in each hall, and the noise signal processed by the noise processing unit 102 is converted into the speaker detecting unit 103. The voice signal memory unit 1 adds voice signal generation time measured by the speaker controller and the system controller, as voice data.
04 is temporarily stored (step S2401). The voice data stored in the voice data memory unit 104 is transferred to the voice data memory unit 104 of the main system (device number 1) via the public line ISDM (step S240).
2). On the other hand, the conversation conducted at the main venue is input to the main system (device number 1), the same process as the process performed by the subsystems (device numbers 2 and 3) is performed, and the voice data memory unit 104 is processed. Is stored.

When the system controller 107 confirms that the BUSY flag of the voice recognition unit 105 is cleared (step S2403), the voice data stored in the voice data memory unit 104 is sent from another subsystem. It is sent to the voice recognition unit 105 and the system controller 107 together with the voice data and converted into a character code and a speaker name (step S2404). The converted character code is stored in the character code memory unit 124 (step S2405). Then, the system controller 107 rearranges the audio data stored in the audio data memory unit 104 in order of earliest generation time with reference to the audio signal generation time of the header added to the audio data (step (S2406), the minutes data edited in the correct order is output from the output unit 106 (step S2407), and the minutes are generated.

As described above, it is possible to increase the memory capacity of the minutes creating device installed in the main venue even if complete minutes data is not created and filed by the minutes creating device installed in the sub-venue. As a result, like the sixth embodiment described above, it is possible to easily, quickly, and automatically create an accurate minutes.

[0094]

As described above, according to the minutes creating apparatus of claims 1 and 2, the speaker code for converting the voice signal of the speaker into the character code and identifying the speaker is used as the speaker. Since the character code and the speaker name are output after the name is converted into the name, there is an effect that an accurate minutes can be created easily, quickly, and automatically.

According to the minutes producing apparatus of claim 3, the speaker code is converted into the speaker name based on the speaker name registered in advance and is output together with the character code. In addition to the effects 1 and 2, there is an effect that the speaker can be surely recognized.

Further, according to the minutes creating apparatus of claim 4, the display means and the editing means can edit the voice recognition error and the conversion into kanji / alphabet. In addition to the effect of the minutes creating device, it is possible to create accurate and easy-to-read minutes by eliminating typographical errors, omissions, and conversion errors.

Further, according to the minutes producing apparatus of claims 5 and 6, the voice signal generation time is added to the character code, and the character code can be edited based on the voice signal generation time. In addition to the effects of claim 1 to claim 6, even when a plurality of speakers are having a conversation, it is possible to create an easy-to-understand minutes that accurately corresponds to the elapsed time of the meeting. Play.

According to the multipoint minutes preparation system of claim 7, the character codes generated by the minutes preparation device installed at the multipoints are collected in the multipoint editing means, and the characters are written by the multipoint editing means. Since the character code can be edited based on the generation time added to the code, accurate minutes of the multipoint video conference where meetings are held in multiple places can be performed.
The effect is that it can be created easily and quickly, and automatically.

Further, according to the multipoint minutes preparation system of claim 8, the output data output from the output means of the minutes preparation device installed at the multipoint is within the multipoint minutes preparation device. In addition to the effect of the multipoint minutes preparation system of claim 7 described above, the multipoint editing means can be added to the editing means of one of the minutes preparation apparatus and edited by the editing means to create the minutes. Since there is no need to provide the special, there is an effect that the configuration is simplified.

[Brief description of drawings]

FIG. 1 is a block diagram showing a schematic configuration of a minutes creating apparatus according to a first embodiment of the present invention.

FIG. 2 is a block diagram showing a schematic configuration of a noise processing unit of the minutes creating apparatus shown in FIG.

FIG. 3 is a block diagram showing a schematic configuration of a speaker detection unit of the minutes creating apparatus shown in FIG.

FIG. 4 is a flowchart showing a procedure of creating minutes of the minutes creating apparatus shown in FIG.

5 is a flowchart showing a noise processing procedure of a noise processing unit in the minutes creating apparatus shown in FIG.

6 is a flowchart showing a speaker detecting procedure of a speaker detecting unit in the minutes creating apparatus shown in FIG.

7 is an explanatory diagram showing a format of audio data in the minutes creating apparatus shown in FIG.

FIG. 8 is a block diagram showing a schematic configuration of a noise processing unit in the minutes creating apparatus according to the second embodiment of the present invention.

FIG. 9 is a block diagram showing a schematic configuration of a minutes creating apparatus according to a third embodiment of the present invention.

FIG. 10 is a block diagram showing a schematic configuration of a minutes creating apparatus according to a fourth embodiment of the present invention.

11 is a flowchart showing a procedure of creating minutes of the minutes creating apparatus shown in FIG.

12 is an explanatory diagram showing a storage format of an audio signal stored in an audio data memory unit of the minutes creating apparatus shown in FIG.

13 is an explanatory diagram showing a speaker name registration format of a speaker name registration unit of the minutes creating apparatus shown in FIG.

14 is an explanatory diagram showing a storage format of an audio signal stored in a character code memory unit of the minutes creating apparatus shown in FIG.

15 is an explanatory diagram showing an example of the output of the minutes created by the minutes creating apparatus shown in FIG.

FIG. 16 is a block diagram showing a schematic configuration of a minutes creating device according to a fifth embodiment of the present invention.

FIG. 17 is an explanatory diagram illustrating a multipoint video conference to which the multipoint minutes recording system according to the sixth embodiment of the present invention is applied.

FIG. 18 is a block diagram showing a schematic configuration of a minutes creating device in the multipoint minutes creating system according to the embodiment.

FIG. 19 is a block diagram showing a schematic configuration of a minutes creating device in the multipoint minutes creating system according to the embodiment.

FIG. 20 is a flowchart showing a multipoint minutes preparation process procedure of the multipoint minutes preparation system according to the embodiment.

FIG. 21 is a block diagram showing a schematic configuration of a minutes creating device (subsystem) in a multipoint minutes creating system according to a seventh embodiment of the present invention.

FIG. 22 is a block diagram showing a schematic configuration of a minutes creating device (main system) in the multipoint minutes creating system according to the embodiment.

FIG. 23 is a block diagram showing a schematic configuration of a multipoint minutes preparation system according to the embodiment.

FIG. 24 is a flowchart showing a multipoint minutes preparation process procedure of the multipoint minutes preparation system according to the embodiment.

[Explanation of symbols]

101 voice signal input unit (voice signal input unit) 103 speaker detection unit (speaker identification unit) 104 voice data memory unit (voice data storage unit) 105 voice recognition unit (voice recognition unit) 106 output unit (output unit) 107 System controller (speaker code conversion means, clocking means) 108 speaker name registration unit (speaker name registration means) 109 microphone 110 microphone 123a voice signal memory unit (voice signal storage unit) 123b speaker code memory unit (speaker code) Storage unit) 124 Character code memory unit (Character code storage unit) 125 Editing unit (Editing unit) 126 Display unit (Displaying unit) 210 Multipoint editing device (Multipoint editing unit)

Claims (8)

[Claims] 1. A voice signal input means having a speaker identification section for receiving a voice signal of a speaker and converting the voice signal into a speaker code for identifying a speaker corresponding to the voice signal. A voice data storage unit for storing voice data obtained by adding a speaker code output from the speaker identification unit to a voice signal input from the voice signal input unit, and a voice stored in the voice data storage unit. A voice recognition means for converting a voice signal in the data into a character code; and an output means for outputting the speaker code in the voice data stored in the voice data storage means and the character code output from the voice recognition means. A minutes preparation device characterized by being equipped with. 2. A voice signal input means having a speaker identification section for receiving a voice signal of a speaker and converting the voice signal into a speaker code for identifying a speaker corresponding to the voice signal. A voice signal storage unit for storing a voice signal input from the voice signal input unit, a speaker code storage unit for storing a speaker code output from the speaker identification unit, and a voice signal storage unit for storing the voice signal. A voice recognition means for converting the voice signal into a character code; and an output means for outputting the character code output from the voice recognition means and the speaker code output from the speaker code conversion means. Characteristic minutes making device. 3. A speaker name registration means for registering the name of a speaker corresponding to the speaker code in advance, and a speaker code in the voice data stored in the voice data storage means for the speaker name registration means. And a speaker name conversion means for converting the speaker name into a name of a speaker output from the speaker name conversion means into a character code output from the voice recognition means. 3. The minutes preparing apparatus according to claim 1, wherein the minutes are added and output. 4. A character code storage unit for storing a character code output from the voice recognition unit, a display unit for displaying the character code stored in the character code storage unit as a character, and a display unit for displaying the character code. 4. The minutes creating apparatus according to claim 1, further comprising an editing means for editing the contents. 5. A voice signal input means having a speaker identification unit for receiving a voice signal of a speaker and converting the voice signal into a speaker code for identifying a speaker corresponding to the voice signal. A time measuring means for measuring a generation time of a voice signal input from the voice signal input means, and a voice code input from the voice signal input means, and a speaker code output from the speaker identifying unit. A voice data storage unit that stores voice data, a voice recognition unit that converts a voice signal in the voice data stored in the voice data storage unit into a character code, and a character code that is output from the voice recognition unit are stored. Character code storage means, display means for displaying the character code stored in the character code storage means as characters, and editing means for editing the contents displayed by the display means,
Speaker name registration means for registering the speaker name corresponding to the speaker code in advance, and the speaker code stored in the speaker name storage means for the speaker code stored in the speaker code storage means. A speaker name converting means for converting into a character code output from the voice recognizing means, a speaker name output from the speaker name converting means, a generation time of a voice signal timed by the time counting means, and An apparatus for preparing minutes, comprising: an output means for adding and outputting a device number for identifying a device. 6. A voice signal input means having a speaker identification section for receiving a voice signal of a speaker and converting the voice signal into a speaker code for identifying a speaker corresponding to the voice signal. Outputting from the speaker identifying unit, and a clocking unit that counts a generation time of a voice signal input from the voice signal input unit, a voice signal storage unit that stores a voice signal input from the voice signal input unit, and a speaker identification unit. A speaker code storage unit for storing a speaker code, a voice recognition unit for converting a voice signal stored in the voice signal storage unit into a character code, and a character code for storing a character code output from the voice recognition unit. Storage means, display means for displaying the character code stored in the character code storage means as a character, editing means for editing the content displayed by the display means, and corresponding to the speaker code in advance. Speaker name registration means for registering a speaker name, and speaker name conversion means for converting a speaker code stored in the speaker code storage means into a speaker name registered in the speaker name registration means. And a name of the speaker output from the speaker name conversion means, a generation time of the voice signal timed by the timekeeping means, and a device number for specifying the device in the character code output from the voice recognition means. A minutes creating device, comprising: an output means for additionally outputting. 7. The minutes preparing apparatus according to claim 5 or 6 is arranged at multiple points, and the output data output from the output means of the minutes preparing apparatus at each of the points is collected in one place. A multipoint editing means is provided, and the minutes of the multipoint conference are created by rearranging the output data by the multipoint editing means based on the generation time of the audio signal timed by the timekeeping means. A multi-point minutes recording system featuring. 8. The minutes creating device according to claim 5 or 6 is arranged at multiple points, and the editing means of one of the multiple points minutes creating device is used to edit the minutes of each point. No multi-point editing means for editing the output data output from the output means of the recording device in one place,
A multipoint minutes preparation is made by rearranging the output data by the multipoint editing means based on a generation time of an audio signal timed by the timekeeping means to create a minutes of a multipoint meeting. system.