JPH10308887A - Program transmitter - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a program transmitter by which real time checking of recited contents of an announcer and time progress management at a program transmission, and accuracy is improved by speech recognition as to a plurality of reading persons in charge. SOLUTION: A program transmitter A that generates a program original and transmits a program is provided with; a storage means M4 that stores a voice parameter of each person in charge for reciting; a KANA (Japanese syllabary of round form) file F2 for reciting of an original; a storage means M3 for information concerning persons in charge for reciting; a means 7 that uses a voice parameter selected based on information concerning persons in charge for reciting so as to recognize the voice of the persons in charge for reciting; and a comparison means M5 that discriminates coincidence between the speech recognition content of the persons in charge for reciting and contents of the KANA file F2.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a program transmitting apparatus in a broadcasting station, and more particularly to a program transmitting apparatus typified by a news program in which speakers speak.

[0002]

2. Description of the Related Art Conventionally, in programs, especially news programs,
News manuscripts are created and broadcast based on these manuscripts. In this case, the original is read to see if it matches the time in news units, and the expression is checked. In addition, computerization of the input of these originals is progressing, and the originals are often created by a word processor or a personal computer.

[0003]

When an announcer who is a speaker reads a manuscript in a news program or the like and speaks, a reading error or the like may occur. According to the prior art, it is possible to recognize the voice of an announcer reading a manuscript by voice recognition and automatically detect whether or not it matches the manuscript, but the text of these manuscripts is kanji Since it is a mixed sentence, it is necessary to convert the kanji in the manuscript into a form that can be used for speech recognition, such as hiragana and vowel consonants.

[0004] By the way, in the manuscript, it is difficult to determine how to read uniquely like a person's name, or as a familiar noun such as a local place name or a political party name as in current affairs news, or a technical term that is rarely used. Phrases or new words often occur. These terms are checked in a dictionary or the like when the manuscript is created, but when the announcer reads and speaks the manuscript, it is necessary to convert the kanji to hiragana again to recognize them. Thus, the similar work is repeatedly performed, thus causing waste and inefficiency. Furthermore, it was necessary to incorporate a dictionary for the speech recognition system.

[0005] Broadcasting generally organizes programs within a predetermined time frame. In particular, since a news program is a live broadcast, it is necessary to manage time progress in order to keep the time within a predetermined time. Therefore, it is monitored whether or not the announcer's reading speed of the original matches the expected speed or is within the allowable speed range. There was a problem that I went to if it was not efficient.

SUMMARY OF THE INVENTION The present invention has been made to solve the above-mentioned problems in the prior art, and enables a real-time inspection of the content read out by an announcer at the time of transmitting a program, and facilitates time progress management. It is an object of the present invention to provide a program transmission device that can improve the accuracy of a plurality of reading staff by voice recognition.

[0007]

In order to solve the above-mentioned problems in the prior art, the present invention stores a kana character as a file together with a manuscript when a news manuscript is created and inputted. Whether the reading of the manuscript of the announcer matches the manuscript is determined using a kana character file, and a real-time inspection is performed. That is, a kana character file is prepared for a read-out manuscript to be provided to an announcer, and a spoken term is checked using voice parameters corresponding to the announcer.

Further, it monitors whether or not the read speed of the original document and the actual read speed match, and displays the degree of the error. That is, the Japanese “ya”, “”
In this case, a silent state occurring after "" is detected to check whether or not the reading speed is appropriate.

Further, the name of the person in charge of the scheduled announcer is stored, and at the time of voice recognition, voice parameters are set for each announcer to improve the accuracy of voice recognition. This utilizes the fact that recognition accuracy is improved by so-called voice recognition by a specific speaker.

Hereinafter, the means according to the present invention will be described. A program sending device according to claim 1 of the present invention is a program sending device that creates a program manuscript and sends the program, and stores a voice parameter of each person who reads aloud,
Recognizing the voice of the person in charge of reading using a kana character file for reading a document, storage means for the information of the person in charge of reading, and voice parameters selected based on the information of the person in charge of reading. Means for comparing the contents of the kana character file with the voice recognition contents of the person in charge of reading out.

According to the above configuration, the duplication of the conversion work between the special term and the kana character is eliminated, so that the discrimination between the read content of the manuscript such as news and the manuscript is made more efficient, and the misreading of the manuscript is performed. And monitoring for inappropriate terminology is automatically done. Further, by improving the recognition accuracy of speech recognition, the accuracy of determination for each speaker is improved, and thus, determination with less error is performed.

According to a second aspect of the present invention, there is provided a program transmitting apparatus comprising:
2. The program transmitting apparatus according to claim 1, further comprising: means for calculating a time length at the time of reading out the kana character file portion; and means for determining a silent state such as a period, and said comparing means for performing the match determination. It is characterized in that the progress of the verbal speech is determined for the time length set by the calculating means.

According to the above configuration, monitoring whether the reading speed of the document matches the actual reading is automatically executed. As a result, misreading of the original and use of inappropriate terms are automatically monitored, so that the program progress can be quickly and accurately managed.

According to a third aspect of the present invention, there is provided a program transmitting apparatus comprising:
3. The program transmitting apparatus according to claim 1, further comprising a word storage unit and a unit for comparing the content spoken orally, and when a word in the storage unit occurs, the fact is displayed.

According to the above arrangement, irregular sounds generated during real time are effectively removed at the time of discrimination.

[0016]

Preferred embodiments of the present invention will be described below in detail with reference to the accompanying drawings. The embodiment described below is a part of a preferred embodiment of the present invention,
Although various limitations that are preferable in terms of the technical configuration are given, the scope of the present invention is not limited to these embodiments unless otherwise specified in the following description.

FIG. 1 is a block diagram showing a main part of an embodiment of a program transmitting apparatus according to the present invention. In the figure, reference numeral F1 denotes a manuscript file containing kanji, and reference numeral F2.
Is a kana character file, code M1 is control means, code M2
Hana character file organization means, code M3 is person information storage means, code M4 is voice parameter storage means, code M5
Is a comparing means for determining a coincidence, M6 is a required time calculating means, M7 is a time measuring means, M8 is a silent state determining means, M9 is a reading speed comparing means, and M10 is a word storage means.

The original file (mixed with Chinese characters) F1 and the kana character file F2 are formed on a large-capacity recording medium such as a hard disk. Further, the person in charge information storage unit M3, the voice parameter storage unit M4, and the word storage unit M10 are also provided with a hard disk or a non-volatile memory device (for example, SR
AM, flash memory, etc.).

On the other hand, control means M1, kana character file organizing means M2, comparing means M5 for matching determination, required time calculating means M6, clocking means M7, silent state determining means M8,
Each of the reading speed comparison means M9 is stored in, for example, a ROM or the like in the form of a program executable by the CPU 1 (see FIG. 2 described later).
Is executed by

Reference numeral 40 denotes a microphone, and reference numeral 4 denotes a microphone.
5 is a keyboard, 2 is an output / display device. The output / display device 2 is configured by a monitor device such as a CRT or an LCD, or a printer device.

The information of the announcer in charge is input from the keyboard 45. In this case, there are two types of information.
The first type of information is the update / addition of the person-in-charge information to be stored. When this first information is input, the person-in-charge information storage means M3 is executed and the corresponding person-specific voice parameter ( And related data) are stored in the voice parameter storage means M4. In this case, the old data will be updated.

The second type of information specifies the name of the announcer in charge at the time of broadcasting, and when this information is input, the control means M1 determines the corresponding announcer from the voice parameter data stored in the voice parameter storage means M4. The data of the person in charge is taken out and a voice recognition device (voice recognition board) 7
To enter. As a result, the speech recognition device (speech recognition board) 7 performs parameter setting with the announcer as the specific speaker. Thereby, the voice recognition device 7 can accurately recognize the voice of the announcer.

Next, when an original is input from the keyboard 45, an original file (mixed with Chinese characters) F1 is created, and at the same time, a kana character file F2 is created by the kana character file organizing means M2. As described above, the preparation for broadcasting is completed.

When the broadcaster starts and the announcer starts reading out the original, the voice of the announcer is input to the voice recognition device (voice recognition board) 7 via the microphone 40, and voice recognition is performed. The result of the voice recognition is input to the comparing means M5 for matching determination.

When the control means M1 extracts a corresponding kana character string from the kana character file F2 and inputs the character string to the comparing means M5, the comparing means M5 matches the input speech recognition result with the kana character string. Whether or not to perform the determination is determined by comparison, and the determination result is sent to the output / display device 2 for output / display. In the case of erroneous reading, the announcer can immediately deal with the error occurrence display with correction or the like.

If there is a part where the result of the speech recognition does not match the kana character string, the comparing means M5 stores the mismatched part in the word storage means M10, for example, "A", "A". If it matches an interjection such as, '' O '', it is displayed as irregular voice, though it is not an error, and the announcer can be alerted. In this way, the announcer can keep an even higher quality announcement.

On the other hand, the control means M1 also inputs the extracted kana character string to the required time calculation means M6. The required time calculating means M6 calculates the required reading time of the kana character string and sends it to the reading speed comparing means M9.

On the other hand, the silent state determining means M8 determines a silent state occurring at a punctuation mark or a reading point based on a voice signal from the microphone 40 or an output of the voice recognition device (voice recognition board) 7. At this point, a signal is sent to the timing means M7. The timer M7 counts the elapsed time from the previously generated silent state to the current silent state, and sends the elapsed time to the reading speed comparing means M9.

The reading speed comparing means M9 compares the required reading time input from the required time calculating means M6 with the elapsed time input from the time measuring means M7, and detects a difference between the reading speed and a predetermined reading speed. The result is sent to the output / display device 2 for output / display. If the reading speed is faster or slower than the predetermined speed, the announcer can immediately adjust the speed by watching this display.

FIG. 2 is an apparatus configuration diagram of an embodiment of a program transmission device according to the present invention. In the figure, reference numeral 1 denotes a CPU of a personal computer (personal computer) which is a main device of the program transmitting apparatus, and reference numeral 2 denotes an output /
A display device. The output / display device 2 is mainly a CRT
Monitor or LCD monitor. Reference numeral 3 denotes memories (RAM and ROM) in the personal computer, which store input / output displays and other various information. Reference numeral 4 denotes a kana-kanji conversion mechanism for converting kana characters input from a personal computer keyboard into kanji.

Reference numeral 5 denotes a general-purpose interface circuit (for example, a SCSI interface circuit) for controlling external devices, and reference numeral 6 denotes an RS-232C interface circuit connected to the control device 20 for transmitting a program.

Reference numeral 7 denotes a voice recognition means (specifically, a voice recognition board), which converts an external voice electric signal into code information that can be processed by a computer. Actually, an audio output from a switcher described later is connected. Reference numeral 8 denotes a hard disk (HD) device connected by the SCSI interface circuit 5, which stores various information for transmission, a news manuscript file generated inside the personal computer, and the like.

Reference numeral 20 denotes a control device for transmitting a program, reference numeral 21 denotes a CPU composed of a microprocessor for controlling the inside of the control device 20, and reference numeral 22 denotes information necessary for the operation of the CPU 21, in particular, notification from a personal computer. It consists of a working memory that holds program information to be executed. Reference numeral 23 denotes a general-purpose interface circuit for controlling an external device, specifically, an RS-232C interface circuit.

Similarly, reference numerals 24, 25, 26, and 27 denote interface circuits, specifically, an RS-422 interface circuit. Reference numeral 28 denotes an interface circuit, which is a time conversion circuit for converting an external time signal into data for a computer.

Reference numeral 30 denotes a switcher for generating images / sounds for transmission, which switches images / sounds from respective devices.
Alternatively, image synthesis and audio mixing are performed, and the interface circuit 24 controls the CPU 21.

Reference numeral 31 denotes a video tape recorder (VT) controlled from the CPU 21 by the interface circuit 25.
R), reference numeral 32 denotes a still image generation device controlled by the CPU 21 by the interface circuit 26, and reference numeral 33 denotes a character generation device (CG) controlled by the CPU 21 by the interface circuit 27. Reference numeral 34 denotes a studio camera. Images / audios of these devices are used for transmission by the switcher 30. Reference numeral 35 denotes a time generator which outputs the time via the time conversion circuit 28 to the CPU 21.
Notify to

In such a configuration, the personal computer CPU 1 displays the news manuscript input by the operation of the
When a command to send a news program is issued, a screen as shown in FIG. 3 is displayed on the monitor 2, and the program is sent according to the command.

FIG. 3 is a diagram showing an example of a screen of a program table. This is called a play list, which is created on a creation screen (not shown) and stored in the hard disk HD 9 built in the personal computer.

In the structure of the play list shown in FIG. 3, in the item column, when the announcer or the like is different, the item number is designated to be different. For example, in FIG.
(Camera 1) is a material to be photographed, and shows that the announcer in charge of reading out in the ANA ID column is Mr. Sato. The check column is a mark indicating whether or not to check the content of speech and the reading speed, which will be described later.

If the mark is blank or '00', it indicates that the announcement content and the reading speed are not checked, '01' indicates that the announcement content is to be checked, and '10' indicates the reading speed. If "11", it indicates that both the announcement content and the reading speed are to be checked.

Normally, when the "start" button shown in FIG. 3 is pressed, the personal computer notifies the control device 20 of the play list for the control device to the CPU 21 via the RS-232C interface circuits 6 and 23. . CPU2
1 analyzes the playlist for the control device and controls each device such as the VTR 31 so as to output the designated image (including the combination in some cases) at the designated time. Specifically, immediately before the transmission of the corresponding material on the play list, an instruction for preparing the material is given to the corresponding machine via the general-purpose interface. At the transmission time, the material is played,
A switching command is issued to the switcher 30.

FIG. 4 conceptually shows a screen for broadcasting news. Broadcasting is performed by synthesizing subtitles (character strings) and still images with an announcer.

FIG. 5 is a diagram showing a display example of a news manuscript. This display is displayed on the terminal for creating the news manuscript or in FIG.
By clicking on the relevant news frame on the playlist screen in, display and printing are possible.

FIG. 6 is a diagram showing a configuration example of a manuscript file and a kana file. The manuscript file and the kana file are both files created on the hard disk HD, and the manuscript text is a sentence of the manuscript itself mixed with kanji. "N" at the beginning of the file name means News, followed by six characters indicating the date, an identifier within the same day ("M": morning "N": daytime "E": evening) and the structure. / Item number is indicated. Also, depending on the extension, the text of the manuscript (“TXT”), the kana character file (“KAN”)
Is shown.

The kana character file is used for a reading check, which will be described later. For example, “heavy rain” is “Oamewa” in the kana character file.

FIG. 7 shows a screen of a document being sent according to the present embodiment, in which the document itself, the chatter itself and the result of the speed check are displayed. In the figure, when the news part is read aloud by mistake during a broadcast, for example, a mark Mk1 indicates that the part "s" is erroneously read.

Further, by moving the speed mark Mk2 at the lower part of the screen from the center 0 to the plus side and displaying it, it indicates that the reading speed is slightly too high. further,"
Please be careful everyone ”is the cursor Mk3
Once the error has occurred, the operator indicates from which statement the next check is to be performed. The CPU 1 notifies the error once, but if the error release button Bt1 is clicked, the CPU 1 continues to the next statement. Character recognition is not performed (specifically, reading is performed as a sentence), and checking is performed from the next sentence. This is because, for example, if the announcer himself makes a word to the effect that it is wrong, it is treated as an error, unlike a news manuscript itself, and even if the correct manuscript is spoken after that as it is, it is treated as an error due to misalignment This is to avoid such inconveniences.

Furthermore, as a case where the manuscript does not match the reading (speaking), there is a case where a word or a sentence not present in the manuscript is uttered. In that case, a red line is displayed at the portion where the error occurred in FIG. For example, if the manuscript reads “News” but reads “News is”, in FIG. 7, a red line (“Fig.
(Not shown).

When an irregular sound such as "A" or "A" is emitted, an orange line (FIG.
(Not shown).

FIG. 8 is a diagram showing a file structure for reading speed check. FIG. 9 is a diagram showing a voice parameter file structure for each person in charge. These are H
This is a file stored on D9 for reading aloud to an announcer. The speed of reading aloud and the name of a parameter file for voice recognition are determined for each announcer.

The CPU 1 notifies the voice recognition board 7 of the parameters for voice recognition immediately before sending the item to be checked. The voice recognition board notifies the CPU 1 of a recognition result based on this information. The notification of the voice recognition parameter and the notification of the voice recognition result are performed via the API. At the time of transmission, the CPU 1 takes in the time information at the same time as taking out the recognition result, and determines the reading speed based on the reading speed in FIG.
Display that effect.

Note that the reading speed for each announcer shown in FIG. 8 is used in a screen for creating a news manuscript not described here, and the length of the item (DUR) relative to the number of characters input in the news manuscript is used. ) Is to indicate when inappropriate. In fact, it shows how much the DUR deviates from the standard speed. In the case of the Sato announcer, 300 characters / 1 minute is set as a standard, and an error is set to be displayed when it becomes 250 characters / 1 minute or less, or 350 characters / 1 minute or more.

FIG. 10 is an operation flowchart of an embodiment of the program transmitting apparatus according to the present invention. FIG.
FIG. 12 is an operation flowchart following FIG. 10, and FIG. 12 is an operation flowchart following FIG. The CPU 1 performs a check on an item in the case where there is a reading speed check instruction through the operations illustrated in FIGS. 10 to 12.

In the figure, when a news manuscript is input (step 96), a kana character file is organized (step 9).
7) Then, the person in charge is read out (step 9)
8), voice parameters are set (step 99). 15 seconds before the start (step 100), a news manuscript file corresponding to the item is fetched (step 10).
1).

The number of characters is calculated based on the fetched file (step 102) and converted into a frame (step 103). Next, the number of frames for each number of characters is calculated (step 104). Here, a frame is the minimum unit in an image signal. In the present embodiment, NTSC used in Japan and North America is adopted as a signal form, so
The second is 30 frames.

The reason for calculating the number of frames for each number of characters in step 104 is that it is sufficient to set the accuracy to frames since the image is in units of frames. Further, the number of frames for 10 or 100 characters other than one character is calculated in order to reduce the accumulated error caused by rounding the frame.

The item start time is reached (step 10
5) After passing through the step 106 for checking the end time, if a voice input occurs soon (step 109),
The voice information up to the period, that is, one sentence is captured (step 112). Immediately after this, a character string (sentence) to be compared next is specified in advance (step 11).
3).

Thereafter, in step 114, the character string in the voice recognition is compared with the character string on the document, and if they do not match, a message to that effect is displayed. Thereafter, the process returns to step 106.
Step 108 is for judging whether or not an error clear instruction has been issued when an error has occurred. If the error has been cleared, the character string from the voice recognition is skipped (Step 1).
10) When the clear button is no longer pressed, the sentence designated next is designated as a character string (sentence) to be compared next (step 111). Perform a comparison.

If an instruction to check a word has been given, the character string is sequentially checked in step 114 for any one that matches an invalid word in the HD 8. indicate.

On the other hand, if a voice input occurs in step 109, time counting is started (step 121), and the elapsed time until the occurrence of a silent state is timed (step 122). Thereby, the reading speed is compared and the result is displayed. As a result, the progress time is managed.

In the above-described embodiment, the identification of the reading is performed using the katakana character file. However, the same configuration can be used even if other characters (a hiragana file or a two-byte kanji code kana file) are used. be able to.

Although the above embodiment is applied to Japanese, it is also possible to apply to English and other languages.

Further, in the above-described embodiment, the reading speed check is performed based on punctuation marks. However, it is also possible to perform punctuation marks or individual characters.

In the above-described embodiment, the term check is performed based on the punctuation marks. However, it is also possible to perform the punctuation marks or the individual characters.

Although the above embodiment has been applied to a news program, it can be applied to, for example, dubbing of a drama, a foreign film, or an animation.

[0066]

As described in detail above, claim 1 of the present invention
The program sending device according to the present invention creates a program manuscript, and in the program sending device that sends out the program, a storage means for storing voice parameters of each person who performs reading, a kana character file for reading the manuscript, Means for recognizing the voice of the person in charge of reading using the voice parameter selected based on the information of the person in charge of reading, and means for storing the contents of the kana character file and the person in charge of the person in charge. Since it is configured to include a comparison means for determining the coincidence with the read-out content, there is no duplication of conversion work between a special term and a kana character as in the related art,
Therefore, it is possible to efficiently determine whether the read content of the original such as news matches the original. In addition, it is possible to automatically monitor misreading of a manuscript such as news and the use of inappropriate terms.

Further, since the recognition accuracy of voice recognition can be improved, the accuracy of discrimination for each speaker can be improved, and thus, discrimination with less error can be performed.

According to a second aspect of the present invention, there is provided a program transmitting apparatus comprising:
2. The configuration according to claim 1, further comprising a calculation unit for calculating a time length at the time of reading out a kana character file portion, and a determination unit for determining a silent state such as a period, and a comparison unit for performing the matching determination. It is designed to automatically determine the progress of the verbal speech for the time length set by, so it can automatically monitor whether the reading speed of the original such as news and the actual reading are the same. Therefore, there is an effect that time management of program progress can be performed quickly and with high accuracy.

According to a third aspect of the present invention, there is provided a program transmitting apparatus comprising:
The configuration according to claim 1 or 2, further comprising a storage unit for the word and a comparison unit for comparing the contents spoken orally, and when a word in the storage unit occurs, the fact is displayed. In addition, irregular sounds generated during real time can be effectively removed in the discrimination, so that the discrimination accuracy can be improved.

[Brief description of the drawings]

FIG. 1 is a block diagram of a main part of an embodiment of a program transmission device according to the present invention.

FIG. 2 is an apparatus configuration diagram of an embodiment of a program transmission device according to the present invention.

FIG. 3 is a diagram illustrating an example of a screen of a program guide.

FIG. 4 is a conceptual diagram of a news broadcast screen.

FIG. 5 is a diagram illustrating a display example of a news manuscript.

FIG. 6 is a diagram illustrating a configuration example of a manuscript file and a kana file.

FIG. 7 is a diagram illustrating an example of a screen of a document being sent.

FIG. 8 is a diagram showing a file structure for reading speed check.

FIG. 9 is a diagram showing a voice parameter file structure for each person in charge.

FIG. 10 is an operation flowchart of one embodiment of a program transmission device according to the present invention.

FIG. 11 is an operation flowchart following FIG. 10;

FIG. 12 is an operation flowchart following FIG. 11;

[Description of Signs] F1: manuscript file (mixed with kanji), F2: kana character file, M1: control means, M2: kana character file organization means, M3: staff information storage means, M4: voice parameter storage means, M5 ... Comparison means for determining coincidence, M6 ...
Required time calculation means, M7 ... time counting means, M8 ... silence state discriminating means, M9 ... reading speed comparison means, M10 ... word storage means, 2 ... output / display device, 7 ... voice recognition device (voice recognition board), 40 ... microphone.

Claims (3)

[Claims] 1. A program sending device for creating a program manuscript and sending the program, a storage means for storing voice parameters of each person who performs reading, a kana character file for reading the manuscript, Means for storing information of a person in charge of reading, means for recognizing the voice of the person in charge of reading using voice parameters selected based on the information of the person in charge of reading, and And a comparing means for judging coincidence with the voice recognition content of the user. 2. A computer according to claim 1, further comprising means for calculating the length of time for reading out the kana character file portion, and means for determining a silent state such as a punctuation mark. 2. The program transmitting apparatus according to claim 1, wherein the progress of the verbal conversation is determined with respect to the time length. 3. The program according to claim 1, further comprising a word storage means and a means for comparing the contents spoken orally, and when a word in the storage means is generated, the fact is displayed. Sending device.