JP2004212799A - Dictation assisting device - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To facilitate hearing and dictation by applying processes such as proper division, speed reduction if necessary, and repetition to a program voice signal by using detected pauses. <P>SOLUTION: A dictation assisting device which assists a dictation of text data from a speech while the speech is heard is equipped with a pause detection part 1 which detects pauses of an input speech signal and a division part 11 which divides the input speech signal by a time or the number of characters enabling a dictator to memorize the divided signal for a short period on the basis of the detected pauses. <P>COPYRIGHT: (C)2004,JPO&NCIPI

Description

[0001]
TECHNICAL FIELD OF THE INVENTION
The present invention relates to a transcription support device that supports transcription of a program audio or the like using detection of a pause portion.
[0002]
[Summary of the Invention]
The present invention relates to an apparatus for supporting the transcription of a program sound or the like using the detection of a pause portion, and particularly, when listening to and transcribe a program sound, it is appropriate to use the detected pause to make an appropriate use. Processing is performed on the program audio signal by performing division, and if necessary, processing such as lowering the speed or repeating the processing, thereby assisting in easy listening and transcription.
[0003]
By performing this processing, it is possible to reduce the time required for the transcription and the feeling of tension and fatigue associated with the work. In particular, if this processing method is applied to the transcription of subtitle text, it will be possible to shorten the time for producing subtitle programs and expand the number of transcription workers by simplifying the work. Contribute.
[0004]
[Prior art]
When a receiver uses a television broadcast program with subtitles, it is important that the subtitles be easy to read and understand. For this reason, in the production of subtitles in the production of subtitle programs, skilled human resources are used to produce subtitles that are easy to read and understand with great effort and time.
[0005]
[Problems to be solved by the invention]
However, in the case of caption broadcasting, which is expanding in the field of applied programs and the number of programs in the future, this skilled manpower, such a subtitle program production system that requires a great deal of labor and time, poses a major bottleneck in subtitle program production. The improvement is urgently needed.
[0006]
Currently, the most common closed caption program production format is to use a program tape with time code superimposed on the video and, if necessary, a program script as a material, which can be used by someone with expertise in broadcasting or other specialized knowledge to 1) program He transcribes the summary of the speech, 2) creates a subtitle display image (according to the procedure for preparing subtitle manuscripts specified separately), and 3) enters the start and end time codes, and prepares a manuscript for subtitles. Based on the subtitle manuscript, the operator creates digitized subtitle data, and performs previewing and proofreading in the presence of the person in charge of the subtitle production, the manuscript creator, and the digitized operator to obtain the completed subtitles.
[0007]
The most time-consuming part of this work is to listen to the program speech in 1) and transcribe and create a subtitle manuscript. Most of this work depends on human intelligence and manual work.
[0008]
In a specific example of a program speech transcribing operation, a program tape is reproduced to listen to audio, and the tape is reproduced and listened to from the speech start point in the audio, and is transcribed by a word processor or writing. Actually, in order to confirm the transcription speed and the contents of the transcription operator, the heading and reproduction operations of the recording tape are repeated for one segment of the speech section, and the transcription operation is performed.
[0009]
Therefore, the transcripting work is a burdensome task that depends on human intelligence such as complicated tape operation such as repetition of tape cueing / playback operation, speech confirmation and transcription, and the necessary skills and labor are required. An appropriate support function that can be reduced is required, and its development is desired.
[0010]
In view of the above circumstances, the present invention performs appropriate division using the detected pose, performs processing such as slowing down or repetition as necessary on the program audio signal, so that it is easy to hear and transcribe. An object of the present invention is to provide a transcription support device that can support.
[0011]
[Means for Solving the Problems]
In order to achieve the above object, an invention according to claim 1 is a writing support device that supports a task of writing text data from a voice while listening to the voice, the pause supporting detecting a pause of an input voice signal. It is characterized by comprising a detecting unit, and a dividing unit that divides the input voice signal based on the detected pose based on the time or the number of characters in which a short time can be stored by the transcriber.
[0012]
According to a second aspect of the present invention, in the transcript assisting apparatus according to the first aspect, the division unit converts the input audio signal by a most reliable pause among the pauses detected by the pause detection unit. The primary dividing means to be divided and, as a result of the division by the primary dividing means, a portion which is longer than a predetermined time or more than a predetermined number of characters that can be stored for a short time, is input with a next more reliable pause. And secondary division means for dividing the audio signal.
[0013]
According to a third aspect of the present invention, in the transcript assisting apparatus according to the first or second aspect, the division unit is configured to perform the division by the secondary division unit, and as a result, a predetermined time or more or a predetermined period that allows short-term storage is possible. For a portion that is longer than the number of characters, the input audio signal may be divided by the next next more reliable pause, or a tertiary division unit that mechanically divides the input audio signal at a predetermined time or number of characters may be included. Features.
[0014]
According to a fourth aspect of the present invention, in the transcription support apparatus according to any one of the first to third aspects, the short-term memory is available for about 2.5 seconds or the short-term memory is available. The number of characters is about 25 mora in number of mora.
[0015]
According to a fifth aspect of the present invention, in the transcript assisting apparatus according to any one of the first to fourth aspects, the division unit includes an overlap unit that overlaps a front and rear part of the divided voice by a predetermined time or the number of characters. It is characterized by:
[0016]
According to a sixth aspect of the present invention, in the transcript assisting apparatus according to any one of the first to fifth aspects, the pause detection unit determines a normal speed from a level signal indicating an amplitude level or a power level of the input audio signal. A specific frequency component extracting means for extracting a specific frequency component indicating a fluctuation characteristic of speech in the above, obtaining an envelope waveform of the extracted specific frequency component, setting a predetermined slice level with respect to the obtained envelope waveform signal, and performing pause. And a pause section detecting means for detecting a section.
[0017]
BEST MODE FOR CARRYING OUT THE INVENTION
<Background of the present invention>
As described above, the transcription of program audio is a burdensome task that depends on human intelligence, such as complicated tape operations such as cueing and repeating playback of tapes, confirmation of speech listening, and transcription. It is necessary to have appropriate support functions that can reduce the necessary skills and labor.
[0018]
For example, as for the "one-segment speech section" and its "repetition", if this is considered as one sentence, there is usually a pause portion before and after the sentence. Process the program audio so that it repeats. As a result, the speech section is easy to hear and transcribe.
[0019]
On the other hand, unnecessary non-speech sections that are transcribed can be automatically removed, so that the transcriber can concentrate on the transcription of the speech part, greatly contributing to the accurate and quick creation of subtitle text. can do.
[0020]
Further, since this processing also contributes to a reduction in necessary skills and labor, there is a possibility that the transcriber can be extended to general word processing workers.
[0021]
The detection of a pause has already been proposed in Japanese Patent Application No. 2002-160255, "Speech Speech / Pause Section Detecting Apparatus", and a pause detected by such a method (this method will be described as a first method) will be described in the present application. It can be applied to processing such as division of program audio.
[0022]
《Method of program audio division processing》
Considering further the significance of the "one-segment speech section", setting of an appropriate division unit for transcription and slowing down processing as necessary requires that the speech of this unit be generated within this division time. And the repetition of the division and the required number of times is the same.
[0023]
First, in terms of ease of listening, it is reasonable to divide the program sound using a place where there is no speech, or a break due to breathing or the like (a pause for writing).
[0024]
The length of the division will be described later, but if the required division length does not fit in this processing, it is necessary to consider additional division processing.
[0025]
For this reason, first, a detected relatively long pose (for example, 1 to 2 seconds) is used to divide a part, and a portion that is not sufficiently divided is divided using a shorter pose (for example, 0.5 to 1 second). (Short pause, predetermined time, etc.).
[0026]
《Consideration on length of division for transcription》
We think that the length of the division for transcription should be short enough to allow short-term memory, including unknown words and proper nouns, but it is reasonable from simple experiments.
[0027]
Although the specific image of the length that can be stored for a short period of time is hard to determine, the familiar examples include haiku, waka, and slogans. Since these characters are 17 characters (mora) and 31 characters (mora) in a hard way (mora), a study was conducted aiming at about 25 characters (mora). This is approximately 19 characters in a sentence mixed with kana-kanji characters in which the ratio of kanji is 1/3, and when one mora is set to 0.1 second, the time is equivalent to 2.5 seconds.
[0028]
First, use the actual program text to verify the validity of the position of the division, the number of mora as the division length raised here, and the time as a division method of the program audio that is easy to hear and transcribe. Was examined.
[0029]
《Examination of segmentation method using actual program text》
FIG. 2 shows a news program as an example, which is manually divided for transcription. Since the example shown in FIG. 2 is the program text, the punctuation marks that normally exist are first focused on, and the division is made in consideration of the phrases.
[0030]
FIG. 3 is an example of division for manual transcription using a documentary program as an example. Similarly, in the example shown in FIG. 3, first, the punctuation is focused on, and the punctuation is also taken into consideration to divide the phrase. Here, the terms used in both figures are defined as follows.
[0031]
Start: Start time of the corresponding program audio
End: End time of the corresponding program audio
Time: Elapsed time of the corresponding program audio
Kana: Number of kana in text
Kanji: Number of Kanji in text
Mora: Number of moras in text (Kanji etc. are calculated as 1.86 moras)
/ S: number of moras / elapsed time, which is the number of moras per second.
[0032]
In the case of FIG. 2, the elapsed time, that is, the time length is 2.45 seconds on average (minimum: 1.3 seconds, maximum: 3.4 seconds), and the number of mora is 24.4 on average (minimum: 12.3, maximum: 33.3 and 390/16). The total number of moras (total of moras) / total elapsed time (total of elapsed times) is 390 / 39.2 = 9.95, which is a value related to the speed of speech, which is rather fast.
[0033]
In the case of FIG. 3, the elapsed time, that is, the time length is 2.26 seconds on average (minimum: 0.9 seconds, maximum: 3.5 seconds), and the number of mora is 17.5 on average (minimum: 9.5, maximum: 25.2). The total number of moras (sum of moras) / total elapsed time (sum of elapsed times) is 7.76, which is a value related to the speed of speech, which is approximately the standard speed.
[0034]
For these two program examples, the text was manually divided as shown in FIGS. 2 and 3, and the program audio was divided so as to correspond to the divided text. Was done.
[0035]
As a result, except for a part of the long elapsed time portion, it was evaluated that it was fairly easy to hear, easy to transcribe, and fast, and the validity of division at an average of 25 mora and about 2.5 seconds was confirmed. I was able to do it.
[0036]
《Detection of speech / pause part in actual program audio》
However, the actual transcription target is the speech in the program audio, which does not include punctuation marks or phrases, and also has various background sounds as well as the target speech, and the situation is completely different.
[0037]
In the processing of the program audio, it is desirable to correctly determine the start and end times of each speech section, and to divide the speech section into an appropriate time length based on the start and end times according to the above description.
[0038]
However, in the actual program audio, not only the target speech but also various background sounds overlap, and it is generally difficult to correctly determine the start and end times of each speech section. On the other hand, in the case of the non-speech section, the silent section is certain, and even if there is a background sound, it may be possible to determine that the reliability is high depending on the level and the section length.
[0039]
Therefore, in the present invention, by analyzing a level signal indicating an amplitude level or a power level of an input audio signal, a specific frequency component indicating a fluctuation characteristic of speech at a normal speed is extracted, and an envelope waveform of the extracted specific frequency component is extracted. A first method of setting a predetermined slice level with respect to the obtained envelope waveform signal and detecting a pause section, or comparing a plurality of LCP cepstrum vectors in acoustic data with each other from a reference frame. The pause is detected by the second method (block cepstrum flux method) for more stably detecting the switching point of the sound data content.
[0040]
<Description of Embodiment>
FIG. 1 is a block diagram showing an embodiment of a transcription support device according to the present invention.
[0041]
The transcription support device shown in FIG. 1 includes a pause detection unit 1 that detects a pause of an input voice signal, and a time or number of characters that allows a transcription worker to perform short-term memory based on the detected pose. And a dividing unit 11 for dividing the audio signal.
[0042]
The pause detection unit 1 includes a specific frequency component extraction unit 3 for extracting a specific frequency component indicating a fluctuation characteristic of speech at a normal speed from a level signal indicating an amplitude level or a power level of an input audio signal; And a pause section detecting means 5 for determining an envelope waveform of the component, setting a predetermined slice level with respect to the determined envelope waveform signal, and detecting a pause section below the level.
[0043]
The dividing unit 1 includes a primary dividing unit 13 that divides the input audio signal by the most reliable pause among the poses detected by the pause detecting unit 1. As a result of the division, for a portion having a predetermined time or more or a predetermined number of characters or more that can be stored for a short time, a secondary dividing means 15 for dividing the input voice signal by a next more reliable pause, As a result of the division by the next division means, for a portion that is longer than a predetermined time or a predetermined number of characters that can be stored for a short time, the input audio signal is divided by the next next more reliable pause or determined in advance. A tertiary dividing means 17 for mechanically dividing the divided voice or the number of characters, and an overlapping means 19 for overlapping the preceding and succeeding portions of the divided voice by a predetermined time or the number of characters. To have.
[0044]
Next, the operation of this embodiment will be described separately for the processing of the pause detecting section 1 and the processing of the dividing section 11.
[0045]
<< Process of Pose Detection Unit 1 >>
4 to 7 are explanatory diagrams showing the procedure of the pause detection processing.
[0046]
In the flowchart of FIG. 4, first, a program audio signal having an amplitude waveform as shown in FIG. 5A is fetched, and audio amplitude value standardization processing is executed (steps ST1 and ST3). In this process, only the low-frequency component of the envelope is extracted from the fetched program audio signal (step ST5), and the signal component is set to a predetermined level based on the amplitude value of the low-frequency component. . Generally, when the level of the low-frequency component is high, the level of the high-frequency component is also considered to be high. This is because if the level is different, the detection accuracy is affected, and it is necessary to standardize the level to some extent.
[0047]
The audio signal whose audio amplitude value has been standardized in this way is then subjected to absolute value processing, and the amplitude value waveform signal (Lch, Rch stereo audio signal) folded back to the + side as shown in FIG. If there is a signal, the speech part is expanded using the sum and difference signals) (step ST7). The waveform in FIG. 6A shows the time axis in the range indicated by the arrow in FIG.
[0048]
Next, digital high-pass filtering (step ST9) and digital band-pass filtering (step ST11) are executed from the amplitude-valued waveform signal that has been converted into an absolute value. In the digital high-pass filtering, a signal having a frequency of, for example, 2 Hz or more is extracted, and in the digital band-pass filtering, a signal having a frequency of, for example, 4 to 7 Hz is extracted. The waveform in FIG. 6B is a signal waveform of a 4 to 7 Hz component.
[0049]
It is known that when the fluctuation characteristics of speech in the time axis direction are compared with the phonetic symbol sequence of speech, the voice power corresponding to the vowel phonetic symbol tends to be larger than other parts. The prior application has shown that the fluctuation of the signal waveform in the time axis direction in the speech at the normal speed has a frequency of about 4 to 7 Hz (Japanese Patent Application No. 2002-160255). This embodiment focuses on the fluctuation of the waveform in the time axis direction, and detects a pause portion by capturing rough periodicity.
[0050]
A signal having a frequency of 4 to 7 Hz is extracted by the digital band filtering process, and when an absolute value process (step ST12) is performed, a signal having a waveform similar to speech is obtained. On the other hand, when the arithmetic processing (step ST13) is executed, components other than the main speech components are extracted as a difference between the extracted signals, and further, an absolute value process (step ST15), a digital low-pass filtering process (step ST15). Step ST17) is performed. In the level correction processing (step ST19), the waveform of the low-frequency component of 0.5 Hz or less generated by the low-frequency filtering processing mainly includes components other than the speech component. , The level of the speech-like waveform signal is corrected in the reverse direction, and as a result, the speech component is emphasized. FIG. 7 shows a speech approximate waveform after this processing.
[0051]
The envelope-corrected envelope waveform signal is displayed as a waveform on a display for verifying its suitability (step ST21). Then, as shown in FIG. 7, the slice is performed at a predetermined slice level (threshold) (step ST23).
[0052]
Next, a minute pause section removal process is executed (step ST25). In this processing, for example, from the sliced audio amplitude value signal, for example, a section of a slight vowel section variation or a degree of breathing is excluded from detection targets. For example, the detection time range is set to about “0.2 to 2 seconds”, and the subsequent time is excluded as a detection target and removed. As a result, pause detection with no meaning and low reliability can be effectively prevented.
[0053]
The detected pause section is displayed on the screen (step ST27). The speech section is actually measured for the purpose of optimizing pose detection accuracy and slice (step ST23) level setting, and compared with the actually measured speech section as shown in FIG. 7 (step ST29). As a result, the pause section derived from the actually measured speech section is compared with the detected pause section, and by performing the comparison, the pose detection accuracy can be checked or the slice level can be changed to be optimal. .
[0054]
<< Processing of Dividing Unit 11 >>
If a pause is detected as described above, then a relatively long pause can be said to have the highest reliability, as shown in the flowchart of FIG. To divide the input audio signal (step ST41). This processing is executed by the primary dividing means 13.
[0055]
As a result of this division, for a portion that is longer than a predetermined time or a predetermined number of characters that can be stored for a short period of time by an operator, the input voice signal is divided by the next long pause (about 0.5 to 1 second) (step ST43, ST45). This processing is executed by the secondary dividing means 15.
[0056]
As a result of this division, for a portion that is longer than a predetermined time or a predetermined number of characters that can be stored for a short term by the operator, the input voice signal is generated by the next next longer pause (for example, 0.2 to 0.5 seconds). Is divided or mechanically divided at a predetermined time or number of characters (steps ST47 and ST49). This processing is executed by the tertiary division means 17.
[0057]
When the division is performed by the division unit 15 or 17 which may have a reliability problem, the duplication unit 19 sets the preceding and succeeding portions of the divided voice to a predetermined time or number of characters. (Step ST51). For example, by providing an overlapping portion of about 1 second, it is possible to prevent the transcriber from being unable to hear or lose the transcript caused by the low reliability of the pause detection.
[0058]
《Pose detection by block cepstrum flux method》
A block cepstrum flux (Bloch Cepstrum Flux) method as a second technique for detecting a speech section will be briefly described.
[0059]
This method detects a switching point of audio data content more stably by comparing a plurality of LCP cepstrum vectors in audio data with each other from a reference frame.
[0060]
The graph of FIG. 9 shows the results of analyzing the sound of an actual television program considered to be an average background sound level by both of these methods.
[0061]
In FIG. 9, (A) is a processed value (cflx analysis value) by the block cepstrum flux method, (B) is a value extracted from the 4-7 Hz component of the audio envelope (by the first method), and (C) is a value of (B). This is data obtained by slicing the waveform at an appropriate level and binarizing it.
[0062]
In the waveform (C), the high-level range corresponds to speech, and the low-level range corresponds to a non-speech (pause) section. In this case, (C) agrees quite well with the actually measured speech section. I can say.
[0063]
Therefore, the speech section in the voice can be grasped to some extent accurately from the waveform of (C). In the waveform of (C), P1 is a short pause, and P3 is a relatively long pause. It can be said that P3 is the most reliable pose and P2 is the next most reliable pose.
[0064]
Note that the block cepstrum flux method (A) (A) was compared in a similar manner. The result is slightly worse than that of the first method, but can be applied in some cases.
[0065]
<Division experiment using pause detection of actual program audio>
A place where there is no speech in the actual program audio, or a break due to breathing, etc. is detected by appropriate processing of the program audio (a pause for transcription), and the speech to be transcribed based on this pause etc. The work was divided into easy-to-use units, and speed reduction was performed as necessary.
[0066]
As described above, it is desirable that the time of the speech that facilitates the writing operation is about 2.5 seconds. Since the target program has a considerably high level of background sound (music), the division method here first detected and divided a relatively long pause (for example, about 2 seconds or more).
[0067]
FIG. 10 illustrates the detection of a pause portion from the sound of a target program (there is a background sound (music) having a considerably high level).
[0068]
Although the original audio waveform of the target program is shown at the lower end, the frequency range component of 4 to 7 Hz is extracted from the envelope of this waveform by using the above-described first technique, and the amplitude value is referred to as “processed audio waveform”. ". The “sound waveform after processing” was compared in magnitude at an appropriate threshold level, and defined as “detection pause”. In order to evaluate the suitability of the “detected pose”, data of measured speech is also displayed for reference. It can be said that at least a relatively long pause (for example, about 2 seconds or more) is correctly detected.
[0069]
Using the result of the pause detection, the first division was performed when the pause was 1 to 2 seconds, and the second division was performed when the pause was 0.5 to 1 second. In this division, a portion that takes about 2.5 seconds or more is divided into about 2.5 seconds by any of the following methods.
[0070]
(1) A shorter pause (for example, 0.2 to 0.5 seconds or more) is detected, and further divided using the detected pause.
(2) Divide this part equally so that it takes about 3 seconds.
[0071]
In addition, at least a portion to which the processing of (1) or (2) is applied is provided with an overlapping portion of about 1 second to avoid omission or inaudibility. Further, in order to indicate an overlapping portion, an overlapping display sound CH is inserted at the beginning of the overlapping portion so that the overlapping portion can be recognized.
[0072]
<Example of division / duplication processing>
An example of the actual processing is shown below in text. The underlined portion indicates the overlap processing. In addition, PP: a detection pause of 1 to 2 seconds or more, SP: a detection pause of 0.5 to 1 second or more, and CH: overlapping display sound (underlined portions are overlapping portions).
[0073]
Processing example of (1) 1 second overlap 18 divisions
PP nature, there is always,Creature
CHCreatureThere is mystery and impression that we weave. SP
Come, travel to the earth in the middle of nature. SP
In search of new encounters,Special
CHSpecialIt is the beginning of the journey of the earth. PP
Today's stage is mediumIn Costa Rica
CHIn Costa RicaIt is a spreading tropical forest. SP
In this forest, manyCreatures
CHCreatures, I live with elaboration. SP
Above all, unique wisdomLiving with
CHLiving withThere is a capuchin capuchin monkey. PP
The wisdom of white-faced capuchin monkeys varies. SP
Tree nuts on stoneTighten, medium
CHTighten, mediumTake out the seeds. SP
In addition, with different types of monkeysLive in peace
CHLive in peaceThere is even the wisdom of bargaining. PP
White-faced capuchin monkey in a forest lifeshow, SP wisdom
CHShow, WisdomStared at a number of. PP
[0074]
(2) Processing example Equal division between PP, 1 second overlap 16 division
PP nature, there is always,Creatures
CreaturesThere is mystery and impression that weave.here we go
here we go, Into the middle of natureLand of living things
Land of living thingsBall travel. New outIn search of a meeting,
In search of a meeting,It is the beginning of a special earth journey. PP
Today's stage is Central America CostaSpread over mosquitoes
Spread over mosquitoesIt is a tropical forest.In this forest,
In this forest,Many creatures are creativeLiving hard
Living hardYou. Above all, GermanyWith special wisdom
With special wisdomThe giant capuchin monkey lives. PP
The wisdom of white-faced capuchin monkeys varies.wood
woodHit the stones on the stone, insideTake out the seeds
Take out the seedsYou. In addition,Different monkeys and flats
Different monkeys and flatsThere is even the wisdom of bargaining to live in harmony. PP
White-faced capuchin monkey living in the forestShow in
Show in, Staring at the wisdom. PP
[0075]
《Writing experiment using split program audio》
The experimental program was 78 seconds in total, and the speech part was 48 seconds (62%).
[0076]
An experiment was conducted by using the program audio according to the above-described division method. In the experiment, as shown in FIG. 11, the divided voice processed as described above was put into MIDIPLAYER (shareware software), and the operation function of this software was used. More specifically, each divided voice is repeatedly played back, and when transcription of the hatched portion is completed, the operation proceeds to the next divided voice by key operation. FIG. 11 shows an example of the process (1).
[0077]
As a result of an experiment using the software shown in FIG.
Transcription time by the method of (1): about 135 seconds The number of times each divided voice is repeated About 2.8 times
Transcription time by the method of (2): about 152 seconds Number of repetitions of each divided voice About 3.2 times
Therefore, it was found that a dividing method that utilizes the detected pose as much as possible is desirable.
[0078]
【The invention's effect】
As described above, according to the present invention, by utilizing the detected pause, the program audio signal is subjected to processing such as appropriate division of the program audio, if necessary, slowing down, or repetition, so that the program audio signal is easily heard and transcribed. It will be possible to help.
[0079]
In addition, since the preceding and succeeding portions of the divided specific voice are overlapped by a predetermined time or number of characters by an overlapping means, it is possible to effectively prevent a transcriber from being unable to hear or lack a transcript. Can be.
[Brief description of the drawings]
FIG. 1 is a block diagram showing an example of a transcription support device according to the present invention.
FIG. 2 is an explanatory diagram showing an example of a division experiment for writing a news program.
FIG. 3 is an explanatory diagram showing an example of division verified by writing a documentary number.
FIG. 4 is a flowchart illustrating a procedure of pause detection.
FIG. 5 is an explanatory diagram showing a program audio signal waveform and an amplitude level signal or a power level signal corresponding to the waveform.
6 is an explanatory diagram showing the amplitude level signal or the power level signal shown in FIG. 5 with the time axis enlarged, and showing extracted component values in a characteristic frequency range.
7 is an explanatory diagram showing an envelope waveform and an actual measurement pause (speech) portion of an amplitude level signal or a power level signal in a specific frequency range shown in FIG. 6;
FIG. 8 is an explanatory diagram showing a processing procedure for dividing program audio by a pause.
FIG. 9 is an explanatory diagram of speech approximation data obtained by specially processing an audio signal.
FIG. 10 is an explanatory diagram of pause portion detection.
FIG. 11 is an explanatory diagram illustrating an example of a computer screen used when reproducing the divided program audio using audio reproduction software.
[Explanation of symbols]
1 Pause detector
3 Specific frequency component extraction means
5 Pause section detection means
11 Division
13 Primary dividing means
15 Secondary dividing means
17 Third division means
19 Overlapping means

Claims (6)

A transcription support device that supports a task of transcribing text data from the voice while listening to the voice,
A pause detector that detects a pause of the input audio signal;
Based on the detected pose, a dividing unit that divides the input audio signal by a time or the number of characters that allows short-term memory of the transcriber,
A transcription support device comprising: The transcription support device according to claim 1,
The dividing unit includes:
Primary division means for dividing the input audio signal by the most reliable pose among the poses detected by the pose detection unit;
As a result of the division by the primary division means, for a portion that is longer than a predetermined time or a predetermined number of characters that can be stored for a short period of time, a second division in which the input voice signal is divided by the next more reliable pause Means,
A transcription support device, comprising: The transcription support device according to claim 1 or 2,
The dividing unit includes:
As a result of the division by the secondary dividing means, for a portion that is longer than a predetermined time or a predetermined number of characters that can be stored for a short time, the input audio signal is divided by the next next more reliable pause, Or tertiary division means for mechanically dividing at a predetermined time or number of characters,
A transcription support device comprising: The transcription support device according to any one of claims 1 to 3,
The writing support device, wherein the short-term memory is available for about 2.5 seconds, or the short-term memory is available for about 25 mora characters. The transcription support device according to any one of claims 1 to 4,
The dividing unit includes:
Including overlapping means for overlapping the front and rear parts of the divided voice by a predetermined time or number of characters,
A transcription support device, characterized in that: The transcription support device according to any one of claims 1 to 5,
The pose detection unit,
From a level signal indicating the amplitude level or power level of the input audio signal, a specific frequency component extracting unit that extracts a specific frequency component indicating a fluctuation characteristic of speech at a normal speed,
Pause section detection means for obtaining an envelope waveform of the extracted specific frequency component, setting a predetermined slice level for the obtained envelope waveform signal and detecting a pause section,
A transcription support device, comprising:

Images