JP5022193B2 - Subtitle monitoring apparatus and subtitle monitoring program - Google Patents

Description

  The present invention relates to a caption monitoring apparatus that monitors whether captions correspond to the contents of a broadcast program, and a program thereof.

2. Description of the Related Art Conventionally, a technique for warning a failure that has occurred in a caption transmission device and a technique for monitoring that captions are not added to a broadcast program are known. For example, in Patent Document 1, special packet data indicating that captions are normally added to a broadcast program is added in advance, and whether or not captions are added to a broadcast program depending on the presence or absence of the special packet data. A teletext monitoring system to be monitored is described.
Japanese Patent Laid-Open No. 11-285032

  However, the invention described in Patent Document 1 has a problem that it cannot be monitored that subtitles different from the content of the broadcast program are displayed. In addition, it is not practical for humans to monitor all of the many broadcast programs, and broadcast programs whose contents do not support subtitles may be broadcast. This deficiency has been pointed out to viewers. Sometimes there was. For this reason, there has been a strong demand to monitor that subtitles correspond to the contents of broadcast programs without human intervention.

  An object of the present invention is to provide a caption monitoring device and a program thereof that can monitor not only the presence / absence of captions but also the correspondence of captions to the contents of broadcast programs without human intervention.

  In order to solve the above-described problem, the caption monitoring device according to claim 1 monitors that the caption corresponds to the content of the broadcast program being broadcast using the electronic program guide information corresponding to the broadcast program. The subtitle monitoring apparatus includes a broadcast program receiving unit, a voice recognition unit, a voice text morpheme analysis unit, a subtitle decoding unit, a subtitle text morpheme analysis unit, and a text comparison unit.

  According to such a configuration, the caption monitoring device receives the broadcast program signal by the broadcast program receiving unit, and separates the audio signal, the caption signal, and the electronic program guide information from the broadcast program signal. Here, the electronic program guide information (EPG: Electric Program Guide) is transmitted together with the broadcast program signal, and includes subtitle information related to the genre of the broadcast program and the subtitles of the broadcast program. Further, the caption monitoring device generates speech text by recognizing the speech signal separated by the broadcast program receiving means by the speech recognition means, and generating the relative start time of the speech text from the start time of the broadcast program. Voice time information including the time length of the voice text is generated. Here, the speech recognition means can perform speech recognition by a statistical method, and the line break of the speech text at the discontinuous portion can be known from the start time and the time length. Further, the caption monitoring device uses the speech text morpheme analysis unit to convert the speech text generated by the speech recognition unit into line-unit speech text in which the line start time based on the speech time information generated by the speech recognition unit is added to the morpheme. To divide.

  Further, the caption monitoring device generates caption text by decoding the caption signal separated by the broadcast program receiving means by the caption decoding means, and caption time information indicating a relative start time for each line break of the caption text Is generated. Here, since the caption signal includes a line break of the caption text, the caption decoding means sets the line break as the line start time of the caption text. Further, the caption monitoring device uses the caption text morpheme analyzing means to convert the caption text generated by the caption decoding means into line-by-line caption text in which the line start time is given based on the caption time information generated by the caption decoding means. To divide. In addition, the subtitle monitoring apparatus, when the subtitle information is included in the electronic program guide information separated by the broadcast program receiving means by the text comparison means, is line-by-line audio text generated by the audio text morphological analysis means in line units. The morpheme and line start time are compared with the morpheme and line start time of the line-by-line subtitle text generated by the subtitle text morpheme analyzing means, and the coincidence rate is calculated, and the calculated coincidence rate is preset below a preset threshold value A warning is output when the specified number of lines are continued. Here, when the electronic program guide information does not include subtitle information, the text comparing means determines that this broadcast program does not display subtitles, and does not calculate the coincidence rate and output a warning.

  The subtitle monitoring apparatus according to claim 2 is the subtitle monitoring apparatus according to claim 1, wherein the broadcast program receiving unit acquires a genre of the broadcast program from the electronic program guide information, and the voice recognition unit includes: A language model corresponding to the genre of the broadcast program acquired by the broadcast program receiving means is selected from a plurality of language models registered in advance, and the audio signal is recognized by using the selected language model. The speech text is generated.

  According to such a configuration, the caption monitoring device uses the language model corresponding to the genre of the broadcast program, so that the accuracy of voice recognition can be improved.

  Further, in the caption monitoring device according to claim 3, in the caption monitoring device according to claim 2, the text comparison unit sets the number of lines for each genre of the broadcast program included in the broadcast program transmission information. A warning is output when the calculated coincidence rate is equal to or less than the threshold value and continues for the number of lines set for each genre of the broadcast program.

  According to this configuration, the caption monitoring device can set different numbers of lines for outputting warnings for each genre of broadcast program.

  In order to solve the above-described problem, the caption monitoring device according to claim 4 monitors that the caption corresponds to the content of the recorded broadcast program by using broadcast program transmission information corresponding to the broadcast program. A subtitle monitoring apparatus that includes an audio signal separation unit, a voice recognition unit, a voice text morpheme analysis unit, a subtitle decoding unit, a subtitle text morpheme analysis unit, and a text comparison unit.

  According to this configuration, the caption monitoring device receives the broadcast program signal and separates the audio signal from the broadcast program signal by the audio signal separation unit. Further, the caption monitoring device recognizes the voice signal separated by the voice signal separation means by the voice recognition means to generate voice text, and sets the relative start time of the voice text from the start time of the broadcast program. Voice time information including the time length of the voice text is generated. Here, the speech recognition means can perform speech recognition by a statistical method, and the line break of the speech text at the discontinuous portion can be known from the start time and the time length. Further, the caption monitoring device uses the speech text morpheme analysis unit to convert the speech text generated by the speech recognition unit into line-unit speech text in which the line start time based on the speech time information generated by the speech recognition unit is added to the morpheme. To divide.

  Further, the caption monitoring device receives the caption signal by the caption decoding means, decodes the caption signal, generates caption text, and displays caption time information indicating a relative start time for each line break of the caption text. Generate. Here, since the caption signal includes a line break of the caption text, the caption decoding means sets the line break as the line start time of the caption text. In addition, the caption monitoring device divides line-by-line subtitle text into which morphemes are provided by subtitle text morpheme analyzing means, to which the line start time based on the subtitle time information generated by the subtitle decoding means is added to the subtitle text generated by the subtitle decoding means. To do. In addition, the caption monitoring device receives broadcast program transmission information by the text comparison means and, when the broadcast program transmission information includes caption information, the line-by-line audio generated by the audio text morphological analysis means for each line. The match rate is calculated by comparing the morpheme and line start time of the text with the morpheme and line start time of the line-by-line subtitle text generated by the caption text morpheme analyzing means, and the calculated match rate is less than or equal to a preset threshold value. Output a warning when the set number of lines are continued. Here, the broadcast program transmission information is created together with the broadcast program and includes subtitle information regarding the genre of the broadcast program and the subtitles of the broadcast program. Therefore, when the broadcast program transmission information does not include caption information, the text comparison means determines that the broadcast program does not display captions, and does not calculate the coincidence rate and output a warning.

  Further, the caption monitoring device according to claim 5 is the caption monitoring device according to claim 4, wherein the voice recognition unit is configured to detect the broadcast program included in the broadcast program transmission information from a plurality of language models registered in advance. A language model corresponding to a genre is selected, and the speech text is generated by speech recognition using the selected language model.

  According to such a configuration, the caption monitoring device uses the language model corresponding to the genre of the broadcast program, so that the accuracy of voice recognition can be improved.

  The subtitle monitoring apparatus according to claim 6 is the subtitle monitoring apparatus according to claim 5, wherein the text comparison unit sets the number of lines for each genre of the broadcast program included in the broadcast program transmission information. A warning is output when the calculated coincidence rate is equal to or less than the threshold value and continues for the number of lines set for each genre of the broadcast program.

  According to this configuration, the caption monitoring device can set different numbers of lines for outputting warnings for each genre of broadcast program.

  In order to solve the above-mentioned problem, the caption monitoring program according to claim 7 uses the electronic program guide information corresponding to the broadcast program to indicate that the caption corresponds to the content of the broadcast program being broadcast. In order to monitor, the computer is caused to function as broadcast program receiving means, speech recognition means, speech text morpheme analysis means, caption decoding means, caption text morpheme analysis means, and text comparison means.

  In order to solve the above-described problem, the caption monitoring program according to claim 8 monitors that the caption corresponds to the content of the recorded broadcast program by using broadcast program transmission information corresponding to the broadcast program. Therefore, the computer is caused to function as an audio signal separating unit, a broadcast program receiving unit, an audio recognition unit, an audio text morpheme analyzing unit, a caption decoding unit, a caption text morpheme analyzing unit, and a text comparing unit.

  According to the caption monitoring device and the program thereof according to the present invention, the following excellent effects can be obtained. The inventions according to claims 1 and 7 use the caption information included in the electronic program guide information to calculate the coincidence ratio between the audio of the broadcast program currently being broadcast and the content of the caption, and this coincidence ratio is preset. A warning is output when a predetermined number of lines or more continue below the threshold. Therefore, according to the first and seventh aspects of the present invention, it is possible to monitor not only the presence / absence of captions but also the correspondence of captions to the contents of broadcast programs in real time without human intervention.

  According to the fourth and eighth aspects of the present invention, the subtitle information included in the broadcast program transmission information is used to calculate the coincidence ratio between the audio and subtitles of the broadcast program recorded on a recording medium such as a video tape or a server HDD. Then, a warning is output when the coincidence rate is equal to or less than a preset threshold and continues for a preset number of rows. Therefore, according to the inventions according to claims 4 and 8, before broadcasting a broadcast program, it is possible to monitor not only the presence / absence of subtitles but also the correspondence of subtitles to the contents of the broadcast program without manpower. it can.

  According to the inventions according to claims 2 and 5, since the accuracy of speech recognition is improved, speech text can be generated accurately, and the monitoring accuracy of the caption monitoring device can be improved. Further, according to the third and sixth aspects of the invention, even when proper nouns or the like for each broadcast program genre are difficult to recognize by voice, the warning is continuously output by adjusting the number of lines for outputting the warning. Since the situation can be prevented, the monitoring accuracy of the caption monitoring device can be improved.

  Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings as appropriate. In each embodiment, means having the same function and the same member are denoted by the same reference numerals, and description thereof is omitted.

(First embodiment)
[Configuration of caption monitoring device]
With reference to FIG. 1, the configuration of the caption monitoring device will be described. FIG. 1 is a block diagram of a caption monitoring device according to the first embodiment of the present invention. The caption monitoring device 1 includes, for example, a CPU (Central Processing Unit), a ROM (Read Only Memory), a RAM (Random Access Memory), and an HDD (Hard Disk Drive). Also, the caption monitoring device 1 includes a broadcast program receiving unit 10 and a voice recognition unit in order to realize a function of monitoring that the subtitles correspond to the contents of the broadcast program being broadcast without manpower. 20, a speech text morpheme analysis unit 30, a caption decoding unit 40, a caption text morpheme analysis unit 50, and a text comparison unit 60.

  The broadcast program receiving means 10 receives a broadcast program signal and separates an audio signal, a caption signal, and electronic program guide information from the broadcast program signal. Here, the broadcast program receiving means 10 decodes the video baseband signal, the audio baseband signal (audio signal), the subtitle signal, and the electronic program guide information from the broadcast program signal, and the broadcast program signal from the electronic program guide information. Get subtitle information and genre. For example, the broadcast program receiving means 10 can be realized by a terrestrial digital broadcast tuner or an expansion board having this function. The broadcast program receiving unit 10 outputs the audio signal, subtitle information, and genre to the audio recognition unit 20, outputs the subtitle information and subtitle signal to the subtitle decoding unit 40, and outputs the subtitle information to the text comparison unit 60. The broadcast program receiving means 10 may not output the caption information to the speech recognition means 20, the caption decoding means 40, and the text comparison means 60 if the electronic program guide information does not include caption information.

  The voice recognition means 20 recognizes the voice signal separated by the broadcast program receiving means 10 to generate voice text, and also the relative start time of the voice text from the start time of the broadcast program and the time length of the voice text. Is generated. Here, the speech recognizing means 20 knows the line break of the speech text at the discontinuous portion from the start time and the time length. Further, this discontinuous portion is a utterance break, and can be understood by detecting a silent portion of a speech signal. For example, the voice time information is generated by using a clock built in the computer, and a broadcast program disclosure time is set to 0: 0: 0, and a relative start time of the voice text is obtained. The speech recognition unit 20 outputs the speech text and the speech time information to the speech text morpheme analysis unit 30. Note that the speech recognition unit 20 does not have to generate speech text or speech time information when caption information is not output from the broadcast program reception unit 10.

  Here, various proper nouns are used for each genre of broadcast programs such as news programs, sports programs, variety programs, etc., but proper nouns and dialects such as characters and place names are difficult to recognize. For this reason, in order to improve the accuracy of speech recognition, a language model (dictionary database) is set for each broadcast program genre such as a language model for news programs, a language model for variety programs, a language model for sports programs, etc. , N-Gram and other statistical methods may be used and stored (registered) in advance in storage means (not shown). Then, the voice recognition unit 20 reads and selects a language model corresponding to the genre of the broadcast program output from the broadcast program reception unit 10 from a plurality of language models stored in the storage unit. Furthermore, the speech recognition means 20 generates speech text by speech recognition using the selected language model.

  Here, the storage means described above is, for example, an HDD, and stores, in addition to a plurality of language models used by the speech recognition means 20, a preset threshold value, a predetermined number of lines, and a predetermined time used by the text comparison means 60 described later. Is.

  The speech text morpheme analyzing unit 30 divides the line unit speech text in which the line start time is given to the speech text generated by the speech recognition unit 20 into morphemes. Here, the speech text morpheme analyzing means 30 divides a line unit speech text into a plurality of morphemes by performing a morphological analysis by a statistical method such as HMM. Further, the speech text morpheme analyzing unit 30 outputs the line unit speech text to the text comparing unit 60.

  The subtitle decoding means 40 decodes the subtitle signal separated by the broadcast program receiving means 10 to generate subtitle text and generates subtitle time information indicating the relative start time for each line break of the subtitle text. is there. For example, the caption time information can be generated from a clock built in the computer, and is generated by adding a relative time from the start time of the broadcast program to the start portion of the line break of the caption text included in the caption information. Also, the caption decoding means 40 decodes the caption signal encoded in the NAB format or ARIB format, and outputs the generated caption text and caption time information to the caption text morpheme analyzing means 50. Note that the caption decoding means 40 may not generate the caption text and the caption time information when the caption information is not output from the broadcast program receiving means 10.

  The subtitle text morpheme analyzing unit 50 divides the line unit subtitle text in which the line start time is added to the subtitle text generated by the subtitle decoding unit 40 into morphemes. Here, the caption text morpheme analyzing means 50 divides the line unit caption text into a plurality of morphemes by performing morphological analysis by a statistical method such as HMM. The subtitle text morpheme analyzing unit 50 outputs the generated line unit subtitle text to the text comparing unit 60.

  When the electronic program guide information separated by the broadcast program receiving means 10 includes subtitle information, the text comparison means 60 morphemes and lines of the line-by-line speech text generated by the speech text morphological analysis means 30 in line units. The match rate is calculated by comparing the start time with the morpheme of the line-by-line subtitle text generated by the caption text morpheme analysis means 40 and the line start time, and the calculated match rate is equal to or less than a preset threshold value. A warning is output when several or more consecutive.

  At this time, subtitles that summarize the voices of announcers and actors in broadcast programs are often displayed so that the screen display is not hindered. If only one line is compared, the contents of the audio text and subtitle text are exactly the same. If not, it can be considered. For this reason, a warning may be output when the coincidence rate becomes 0 continuously for five or more lines. Further, the text comparison means 60 sets the number of lines to output a warning for each genre of broadcast programs such as news programs, sports programs, variety programs, etc., and the calculated coincidence rate is equal to or less than a threshold value and when the number of lines continues for this number of lines. A warning may be output. Note that the number of rows is not limited to five, and may be one or more.

  In addition, since the line start time of the line unit audio text and the line start time of the line unit subtitle text rarely coincide with each other, the text comparison unit 60 adds the line start time to the line unit time. If the line start time of the line unit audio text is included within a predetermined time range (for example, within 2 seconds) from the line start time of the caption text, it may be determined that the lines are the same. The text comparison unit 60 is not limited to the case where the matching rate is equal to or lower than the threshold value, and may output a warning when the matching rate is equal to or higher than the threshold value, less than the threshold value, or higher than the threshold value.

  Further, the text comparison means 60 includes a display device that displays a warning and a comparison result on the screen, an alarm that sounds a warning sound, or an e-mail server that generates a warning mail and transmits the warning mail to a pre-registered person. May be. Here, the text comparison means 60 of FIG. 1 is provided with a display device capable of displaying the warning and the comparison result. Hereinafter, referring to FIG. 2, morpheme comparison and comparison result display in the text comparison unit 60 will be described (see FIG. 1 as appropriate).

  FIG. 2 is a screen display example of the comparison result of the caption monitoring device of FIG. As shown in FIG. 2, the text comparison means 60 displays a broadcast program information display area 61, a broadcast program display area 62, and a data comparison display area 63 separately. The text comparison means 60 acquires the electronic program guide information from the broadcast program receiving means 10, and the broadcast program information display area 61 displays the broadcast channel, program name, subtitle, broadcast start time, broadcast end time included in the electronic program guide information. , Event ID, presence / absence of captions (presence / absence of caption information), and broadcast program genre may be displayed. The text comparison means 60 may acquire the broadcast program signal from the broadcast program reception means 10 and display the broadcast program and its subtitles in the broadcast program display area 62. In this broadcast program, along with the yacht video, the subtitle “Today, the wind is blowing” is displayed at 12:00:25 (1st line), and at 12:00:45 The subtitle "It's a great yacht day" is displayed (second line).

  Further, in the data comparison display area 63, the time, the caption data (line unit caption text), the speech recognition result (line unit speech text), and the match are displayed as comparison results in order from the left side. Here, the line start time of the line-by-line subtitle text is displayed as the above time. For the above-mentioned adaptation, “OK” is displayed when the matching rate exceeds a threshold (for example, 0), and “NG” is displayed when the matching rate is equal to or less than the threshold. Furthermore, in the data comparison display area 63, the comparison results of a plurality of lines may be sorted and displayed in chronological order. In this case, the text comparison means 60 displays the time information of the broadcast program displayed in the broadcast program display area 62. (Time code) may be synchronized with the time of the comparison result displayed at the top of the data comparison display area 63.

  For example, the text comparison unit 60 calculates the ratio of the number of matching morphemes as the matching rate with respect to the total number of morphemes of the line unit speech text or the line unit caption text. First, the text comparison means 60 uses a line-by-line subtitle text “Today is a good wind | I am blowing” on the first line and a line-by-line speech text “Kyoto | Are compared with each other (“|” indicates a morpheme break). Here, it is assumed that the line start times of the line unit subtitle text and the line unit audio text are substantially equal. As mentioned above, “Today” and “Kyoto” do not match, “Good wind” and “Good wind” do not match, and “Wow” and “Wipe” do not match. As described above, since all morphemes do not match, the text comparison unit 60 calculates these matching rates as 0, and determines that they do not match because the matching rate is equal to or less than a threshold value (for example, 0). To display “NG”. If the matching rate is equal to or lower than the threshold value and the number of lines (for example, 5 lines) for which a warning is output continues for a number of times, the text comparison unit 60 outputs a warning. Note that the text comparison unit 60 may determine that these lines match if the morphemes of each line to be compared match a predetermined number (for example, two) or more.

  Next, the text comparing means 60 uses each morpheme for the line-by-line subtitle text “Great | Yacht | From the Day” and the line-by-line speech text “Great | Yacht | Compare whether or not. Here, the morphemes “perfect” and “yacht” match, and the morphemes “from day” and “from beauty” do not match. Therefore, since the two morphemes match among the three morphemes, the text comparison unit 60 calculates the match rate as 2/3, and the match rate exceeds a threshold value (for example, 0). Then, “OK” is displayed. As described above, the caption monitoring device 1 can calculate and compare the match rate between the content word and the function word included in the line unit speech text and the line unit caption text, and output a warning according to the match rate. it can. Note that the caption monitoring device 1 may calculate the matching rate of only the content words, or calculate the matching rate including the functional words, and output a warning according to the matching rate.

  Note that the voice recognition means 20 cannot perform voice recognition while symbols are displayed as subtitles in a broadcast program, for example, while note marks are displayed as subtitles in a music program. Here, the caption monitoring device 1 registers in advance a symbol that cannot be recognized in the voice in the setting file of the storage means. Then, the text comparing means 60 reads the symbol registered in the setting file at the time of comparison, and if this symbol is included in the line unit subtitle text, the line is not compared. Thereby, the caption monitoring device 1 can improve the accuracy of the comparison result.

[Operation of caption monitoring device]
The operation of the caption monitoring device will be described with reference to FIG. 3 (see FIG. 1 as appropriate). FIG. 3 is a flowchart showing the operation of the caption monitoring device of FIG. First, the caption monitoring device 1 receives the broadcast program signal by the broadcast program receiving means 10 and separates the audio signal, the caption signal, and the electronic program guide information from the broadcast program signal (step S1).

  Subsequent to step S1, the caption monitoring device 1 uses the voice recognition unit 20 to recognize the voice signal separated by the broadcast program receiving unit 10 to generate voice text, and to generate the voice text from the start time of the broadcast program. Voice time information including the relative start time of the voice and the time length of the voice text is generated (step S2). Also, the caption monitoring device 1 reads the language model corresponding to the genre of the broadcast program output from the broadcast program receiving unit 10 by the voice recognition unit 20 from the plurality of language models stored in the storage unit, and selects the selected language model. Speech text is generated by speech recognition using.

  Subsequent to step S2, the caption monitoring device 1 uses the speech text morphological analysis unit 30 to divide the line-unit speech text in which the line start time is added to the speech text generated by the speech recognition unit 20 into morphemes (step S3). .

  Subsequent to step S3, the subtitle monitoring apparatus 1 generates subtitle text by decoding the subtitle signal separated by the broadcast program receiving unit 10 using the subtitle decoding unit 40, and relative to each line break of the subtitle text. Subtitle time information indicating the start time is generated (step S4). Further, the caption monitoring device 1 divides the line unit caption text in which the line start time is given to the caption text generated by the caption decoding means 40 by the caption text morpheme analyzing means 50 (step S5).

  Subsequent to step S5, when the subtitle information is included in the electronic program guide information separated by the broadcast program receiving unit 10 by the text comparison unit 60, the subtitle monitoring apparatus 1 performs voice text morphological analysis unit on a line basis. 30 compares the morpheme and line start time of the line unit speech text generated by 30 with the morpheme and line start time of the line unit subtitle text generated by the subtitle text morpheme analysis means 40, and calculates the match rate. When is below a preset threshold value, a warning is output (step S6).

(Second Embodiment)
[Configuration of caption monitoring device]
The configuration of the caption monitoring device will be described with reference to FIG. FIG. 4 is a block diagram of a caption monitoring device according to the second embodiment of the present invention. The caption monitoring device 1B is an audio signal for realizing a function for monitoring that the caption corresponds to the content of a broadcast program recorded on a recording medium such as a video tape or an HDD of a server without human intervention. Separation means 70, speech recognition means 20B, speech text morpheme analysis means 30, subtitle decoding means 40B, caption text morpheme analysis means 50, and text comparison means 60B are provided.

  The audio signal separating means 70 receives the broadcast program signal and separates the audio signal from the broadcast program signal. Here, the caption monitoring device 1B includes an interface (not shown) that can be connected to a video deck or a server, and the audio signal separation means 70 receives a broadcast program signal recorded on the recording medium. The audio signal separation means 70 outputs this audio signal to the voice recognition means 20B.

  The voice recognition means 20B recognizes the voice signal separated by the voice signal separation means 70 to generate voice text, and also calculates the relative start time of the voice text from the start time of the broadcast program and the time of the voice text. The voice time information including the length is generated. Here, the broadcast program transmission information is recorded on the recording medium together with the broadcast program, and the voice recognition means 20B receives the broadcast program transmission information recorded on the recording medium, and the voice recognition means 20 of FIG. In the same manner as described above, speech text and speech time information are generated. Note that if the broadcast program transmission information does not include caption information, the speech recognition unit 20B determines that the broadcast program does not display captions and does not have to generate speech text or speech time information.

  Further, similarly to the voice recognition unit 20 in FIG. 1, a language model may be created for each broadcast program genre and stored (registered) in advance in a storage unit (not shown). Then, the voice recognition unit 20B reads and selects a language model corresponding to the genre of the broadcast program included in the broadcast program transmission information from the plurality of language models stored in the storage unit. Further, the speech recognition means 20B generates speech text by speech recognition using the selected language model.

  The subtitle decoding means 40B receives a subtitle signal, decodes the subtitle signal to generate subtitle text, and generates subtitle time information indicating a relative start time for each line break of the subtitle text. Here, the caption decoding means 40B receives the broadcast program transmission information recorded on the recording medium, and generates the caption text and the caption time information in the same manner as the caption decoding means 40 of FIG. Note that, when the broadcast program transmission information does not include the caption information, the caption decoding unit 40B determines that the broadcast program does not display the caption, and does not need to generate the audio text or the audio time information.

  The text comparison unit 60B receives the broadcast program transmission information and, when the broadcast program transmission information includes subtitle information, the morphemes and lines of the line unit speech text generated by the speech text morpheme analysis unit 30 for each line. The coincidence rate is calculated by comparing the start time with the morpheme and line start time of the line-by-line subtitle text generated by the subtitle text morpheme analysis means 50, and a warning is output when the calculated coincidence rate is equal to or less than a preset threshold value. Is. Here, the text comparison means 60B calculates the match rate and outputs a comparison result and a warning, like the text comparison means 60 of FIG.

[Operation of caption monitoring device]
The operation of the caption monitoring device will be described with reference to FIG. 5 (see FIG. 4 as appropriate). FIG. 5 is a flowchart showing the operation of the caption monitoring device of FIG. First, the caption monitoring device 1B receives the broadcast program signal by the audio signal separation means 70 and separates the audio signal from the broadcast program signal (step S11).

  Subsequent to step S11, the caption monitoring device 1B recognizes the voice signal separated by the voice signal separation unit 70 by the voice recognition unit 20B to generate voice text, and also uses the voice text from the start time of the broadcast program. The voice time information including the relative start time and the time length of the voice text is generated (step S12). In addition, the caption monitoring device 1 reads out the language model corresponding to the genre of the broadcast program included in the received broadcast program transmission information from the plurality of language models stored in the storage unit by the voice recognition unit 20B, and selects the selected language. Speech text is generated by speech recognition using the model.

  Subsequent to step S12, the subtitle monitoring apparatus 1B uses the speech text morphological analysis unit 30 to divide the line unit speech text in which the line start time is added to the speech text generated by the speech recognition unit 20B into morphemes (step S13). .

  Subsequent to step S13, the caption monitoring device 1B receives the caption signal by the caption decoding unit 40B, decodes the caption signal, generates caption text, and sets a relative start time for each line break of the caption text. The subtitle time information shown is generated (step S14). Moreover, the subtitle monitoring apparatus 1B divides the line unit subtitle text in which the line start time is added to the subtitle text generated by the subtitle decoding means 40B by the subtitle text morpheme analyzing means 50 (step S15).

  Subsequent to step S15, the caption monitoring device 1B receives the broadcast program transmission information by the text comparison unit 60B. When the broadcast program transmission information includes the caption information, the caption monitoring device 1B performs the speech text morphological analysis unit line by line. 30 compares the morpheme and line start time of the line unit speech text generated by 30 with the morpheme and line start time of the line unit subtitle text generated by the subtitle text morpheme analysis means 50 to calculate the match rate, and calculates the match rate When is below a preset threshold value, a warning is output (step S16).

  As described above, the subtitle monitoring apparatus according to the present invention can monitor that the subtitles correspond to the contents of the currently broadcast program or the pre-recorded broadcast program without manpower. it can. In each embodiment, the caption monitoring device according to the present invention has been described as an independent device. However, the caption monitoring device according to the present invention may be incorporated in a broadcast program transmission device, and a general computer may be configured as the above-described means. It can also be operated by a program that functions as: This program may be distributed via a communication line, or may be distributed by writing in a recording medium such as a CD-ROM or a flash memory.

1 is a block diagram of a caption monitoring device according to a first embodiment of the present invention. It is a screen display example of the comparison result of the caption monitoring device of FIG. It is a flowchart which shows operation | movement of the caption monitoring apparatus of FIG. It is a block diagram of the caption monitoring device concerning a 2nd embodiment of the present invention. It is a flowchart which shows operation | movement of the caption monitoring apparatus of FIG.

Explanation of symbols

1,1B Subtitle monitoring apparatus 10 Broadcast program receiving means 20, 20B Speech recognition means 30 Speech text morpheme analysis means 40, 40B Subtitle decoding means 50 Subtitle text morpheme analysis means 60, 60B Text comparison means 70 Audio signal separation means

Claims (8)

A subtitle monitoring apparatus that monitors, using electronic program guide information corresponding to the broadcast program, that the subtitle corresponds to the content of the broadcast program being broadcast,
A broadcast program receiving means for receiving a broadcast program signal and separating the audio signal, the caption signal and the electronic program guide information from the broadcast program signal;
The speech signal separated by the broadcast program receiving means is recognized by speech to generate speech text, and the relative start time of the speech text from the start time of the broadcast program and the time length of the speech text Voice recognition means for generating voice time information including;
A speech text morpheme analysis unit that divides a line unit speech text in which a line start time is given based on the speech time information generated by the speech recognition unit to the text generated by the speech recognition unit;
Subtitle decoding means for decoding the subtitle signal separated by the broadcast program receiving means to generate subtitle text and generating subtitle time information indicating a relative start time for each line break of the subtitle text;
Subtitle text morpheme analyzing means for dividing the subtitle text generated by the subtitle decoding means into morphemes, the line unit subtitle text provided with the line start time based on the subtitle time information generated by the subtitle decoding means;
When subtitle information is included in the electronic program guide information separated by the broadcast program receiving means, the morpheme and the line start time of the line-by-line speech text generated by the speech text morpheme analyzing means in line units And the morpheme of the line unit subtitle text generated by the subtitle text morpheme analyzing means and the line start time are calculated to calculate a match rate, and the calculated match rate is preset below a preset threshold value. Text comparison means that outputs a warning when the number of consecutive lines exceeds,
A subtitle monitoring apparatus comprising: The broadcast program receiving means acquires a genre of the broadcast program from the electronic program guide information,
The voice recognition means selects a language model corresponding to the genre of the broadcast program acquired by the broadcast program reception means from a plurality of language models registered in advance, and uses the selected language model to select the voice model. The caption monitoring apparatus according to claim 1, wherein the speech text is generated by recognizing a signal.   The text comparison means sets the number of lines for each genre of the broadcast program included in the electronic program guide information, and sets the line number set for each genre of the broadcast program with the calculated matching rate being equal to or less than the threshold. The subtitle monitoring apparatus according to claim 2, wherein a warning is output when the number continues for a number of times. A subtitle monitoring device that monitors, using broadcast program transmission information corresponding to the broadcast program, that the subtitle corresponds to the content of the recorded broadcast program,
Audio signal separating means for receiving a broadcast program signal and separating an audio signal from the broadcast program signal;
The speech signal separated by the speech signal separation means is recognized by speech to generate speech text, and the relative start time of the speech text from the start time of the broadcast program and the time length of the speech text Voice recognition means for generating voice time information including;
A speech text morpheme analysis unit that divides a line unit speech text in which a line start time is given based on the speech time information generated by the speech recognition unit to the text generated by the speech recognition unit;
Subtitle decoding means for receiving a subtitle signal, decoding the subtitle signal to generate subtitle text, and generating subtitle time information indicating a relative start time for each line break of the subtitle text;
Subtitle text morpheme analyzing means for dividing the subtitle text generated by the subtitle decoding means into morphemes, the line unit subtitle text provided with the line start time based on the subtitle time information generated by the subtitle decoding means;
When the broadcast program transmission information is received and subtitle information is included in the broadcast program transmission information, the morpheme and the line start of the line unit speech text generated by the speech text morpheme analysis unit are generated in units of lines. The matching rate is calculated by comparing the time with the morpheme and the line start time of the line-by-line subtitle text generated by the caption text morpheme analysis means, and the calculated matching rate is preset below a preset threshold value A text comparison means for outputting a warning when the number of consecutive lines is exceeded,
A subtitle monitoring apparatus comprising:   The voice recognition means selects a language model corresponding to a genre of the broadcast program included in the broadcast program transmission information from a plurality of language models registered in advance, and uses the selected language model to select the voice model. The caption monitoring device according to claim 4, wherein the speech text is generated by recognizing a signal.   The text comparison means sets the number of lines for each genre of the broadcast program included in the broadcast program transmission information, and sets the line number set for each genre of the broadcast program with the calculated matching rate equal to or less than the threshold. The subtitle monitoring apparatus according to claim 5, wherein a warning is output when the number continues for a number of times. In order to monitor that the subtitles correspond to the content of the broadcast program being broadcast using the electronic program guide information corresponding to the broadcast program,
A broadcast program receiving means for receiving the broadcast program signal and separating the audio signal, the subtitle signal and the electronic program guide information from the broadcast program signal;
The speech signal separated by the broadcast program receiving means is recognized by speech to generate speech text, and the relative start time of the speech text from the start time of the broadcast program and the time length of the speech text Voice recognition means for generating voice time information including,
A speech text morpheme analysis unit that divides a line unit speech text in which a line start time is given based on the speech time information generated by the speech recognition unit into the text generated by the speech recognition unit;
Subtitle decoding means for generating subtitle text by decoding the subtitle signal separated by the broadcast program receiving means, and generating subtitle time information indicating a relative start time for each line break of the subtitle text;
Subtitle text morpheme analysis means for dividing the line-by-line subtitle text in which the line start time is given based on the subtitle time information generated by the subtitle decoding means to the subtitle text generated by the subtitle decoding means;
When subtitle information is included in the electronic program guide information separated by the broadcast program receiving means, the morpheme and the line start time of the line-by-line speech text generated by the speech text morpheme analyzing means in line units And the morpheme of the line unit subtitle text generated by the subtitle text morpheme analyzing means and the line start time are calculated to calculate a match rate, and the calculated match rate is preset below a preset threshold value. Text comparison means to output a warning when the number of consecutive lines exceeds
A subtitle monitoring program characterized by functioning as In order to monitor that the subtitles correspond to the contents of the recorded broadcast program using the broadcast program transmission information corresponding to the broadcast program,
An audio signal separating means for receiving the broadcast program signal and separating the audio signal from the broadcast program signal;
The speech signal separated by the speech signal separation means is recognized by speech to generate speech text, and the relative start time of the speech text from the start time of the broadcast program and the time length of the speech text Voice recognition means for generating voice time information including,
A speech text morpheme analysis unit that divides a line unit speech text in which a line start time is given based on the speech time information generated by the speech recognition unit into the text generated by the speech recognition unit;
Subtitle decoding means for receiving a subtitle signal, decoding the subtitle signal to generate subtitle text, and generating subtitle time information indicating a relative start time for each line break of the subtitle text;
Subtitle text morpheme analysis means for dividing the line-by-line subtitle text in which the line start time is given based on the subtitle time information generated by the subtitle decoding means to the subtitle text generated by the subtitle decoding means;
When the broadcast program transmission information is received and subtitle information is included in the broadcast program transmission information, the morpheme and the line start of the line unit speech text generated by the speech text morpheme analysis unit are generated in units of lines. The matching rate is calculated by comparing the time with the morpheme and the line start time of the line-by-line subtitle text generated by the caption text morpheme analysis means, and the calculated matching rate is preset below a preset threshold value Text comparison means that outputs a warning when the specified number of lines continues
A subtitle monitoring program characterized by functioning as

Images