JP3534711B2 - Audio editing device and audio editing program - Google Patents

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an audio editing apparatus and an audio editing program, and more particularly, to estimate whether a recording target voice or a non-recording target voice is associated with each voice portion included in a voice waveform to be edited. By displaying the results,
A technique for improving the efficiency of voice editing.

[0002]

2. Description of the Related Art Music, conversation, animation and dialogue in games are often recorded in a recording studio or the like, converted into digital data, and edited / processed by a voice editing program. In the voice editing program, based on digital format voice data (waveform data),
The waveform of the recorded voice is displayed on the computer display, and the editor specifies the waveform position (voice timing) or waveform range (voice section) on the display screen with a pointing device such as a mouse, and then the designation. The recorded voice after the waveform position or the recorded voice in the waveform range can be output as voice. And
While confirming the waveform contents (audio contents) appropriately, specify the arbitrary waveform position or waveform range with the mouse etc., and further specify the editing method such as cutting, copying, pasting, adding various sound effects, etc. It is designed to proceed with the editing work of.

[0003]

However, in the above-mentioned conventional audio editing program, the position or range of the waveform is specified on the display screen, and the audio content from the specified position or the audio content of the specified range is listened to. Without confirmation, there is a problem that it cannot be determined whether or not the sound is a recording target. That is, the recorded sound includes a sound to be recorded (recording target sound (editing target sound)) and other sounds (non-recording target sound (non-editing target sound)). If it is not possible to determine whether each voice portion included in the voice waveform is related to the recording target voice or the non-recording target voice, the editing efficiency is poor unless the above is confirmed. The recording target voice is, for example, a singing voice of a singer, a speech of a conference participant, a dialogue of a voice actor, or the like. On the other hand, the non-recording target voice is, for example, the soliloquy of the recording target person such as a singer, a conference participant, a voice actor, or the voice of the studio staff.

The present invention has been made in view of the above problems, and an object thereof is to provide a display for assisting determination of recording target voice or non-recording target voice with respect to each voice portion included in a voice waveform to be edited. By providing a voice editing apparatus and a voice editing program, it is possible to improve the efficiency of voice editing.

[0005]

In order to solve the above problems, a voice editing apparatus according to the present invention comprises a waveform display means for displaying a waveform of a voice and a voice portion determination for determining a voice portion included in the waveform. Means, estimating means for estimating whether each voice part included in the waveform relates to a recording target voice or a non-recording target voice, and the recording target in association with each voice part included in the waveform. Estimation result display means for displaying the estimation result by the voice estimation means.

According to the present invention, the waveform of the voice is displayed on the display means, and the voice portion is associated with each of the voices contained in the waveform and is associated with the voice to be recorded or the voice not to be recorded. The estimation result of the thing is displayed. In this way, the editor can easily determine whether each audio part relates to the recording target sound or the non-recording target sound based on the display content, and the editing efficiency can be improved. When displaying the estimation result, it may be possible to directly display whether each audio part is related to the recording target sound or the non-recording target sound. Alternatively, the display may be indirectly performed depending on the display mode. In addition, it indicates in stages whether each audio part is closer to the recording target sound or the non-recording target sound (for example, it is closer to the recording target sound, closer to the non-recording target sound, it can be said that neither). May be

Further, according to one aspect of the present invention, based on the peak value of each voice portion included in the waveform, the estimating means determines whether the voice portion is a recording target voice or a non-recording target voice. To estimate. The peak value is 1 of the voice part
One value may be used, or a plurality of values, for example, all values may be used. In this aspect, for example, by calculating a value based on the peak value such as sound pressure, effective sound pressure, average absolute value of peak value, peak value of peak value, etc., a voice portion whose value is larger than a predetermined threshold is recorded. It is possible to estimate that the target voice is related to the non-recorded target voice, and the target voice is equal to or less than the predetermined threshold.

Further, according to one aspect of the present invention, the estimating means determines, based on frequency information of each voice portion included in the waveform, whether the voice portion is a recording target voice or a non-recording target voice. To estimate. This makes it possible, for example, to estimate whether the audio part relates to the recording target sound or the non-recording target sound, based on the frequency characteristics of each audio part. The frequency information is, for example, the fundamental frequency (f ₀ ) information of the speaker, the number of zero crosses, the frequency distribution, and the like.

Further, according to one aspect of the present invention, it further includes voice recognition means for performing voice recognition processing on the waveform, and the estimation means recognizes the recognition result in the voice recognition processing for each voice part included in the waveform. Based on the likelihood of, the audio part is estimated to be related to the recording target sound or the non-recording target sound. In this way, for example, it is estimated that the voice portion whose likelihood of the recognition result is higher than the predetermined threshold is related to the recording target sound, and that the voice portion below the predetermined threshold is related to the non-recording target sound. It becomes possible.

Further, the voice editing program according to the present invention includes a step of displaying a waveform of voice, a step of determining a voice portion included in the waveform, and each voice portion included in the waveform is a recording target voice. The computer is made to execute a step of estimating which one of the non-recorded target voices is involved and a step of displaying the estimation result in the estimating step in association with each voice part included in the waveform. is there.

According to the present invention, the waveform of the voice is displayed on the display means of the computer, and the voice portion is associated with each voice portion included in the waveform, and the voice portion is either the recording target voice or the non-recording target voice. Since the estimation result of whether or not the sound is related to is displayed, the editor can easily judge whether each audio part is related to the recording target sound or the non-recording target sound according to the display content, and the editing efficiency can be improved. Can be improved.

[0012]

BEST MODE FOR CARRYING OUT THE INVENTION Preferred embodiments of the present invention will now be described in detail with reference to the drawings.

FIG. 1 is a diagram showing the configuration of a computer system which operates as a voice editing apparatus according to an embodiment of the present invention. In the computer system 10 shown in the figure, a CPU (central processing unit) 14 and an image processing unit 16 are provided.
, Monitor 18, hard disk storage device 19, R
AM (random access memory) 20, ROM (read only memory) 22, input / output interface 24,
28 and 28 are connected to each other via the bus 12 so that data can be exchanged between them. A media reading device 26 is connected to the input / output interface 24, and an input device 30 is connected to the input / output interface 28.

The CPU 14 is a medium reading device 26.
The image processing unit 16 executes a program supplied from the computer and controls each unit of the computer system 10.
Generates image data under the control of the CPU 14,
It is converted into a video signal at a predetermined timing and the monitor 1
8 is output. The monitor 18 is a CRT or LC
It is a display device including D and the like. The hard disk storage device installs data such as programs read by the media reading device 26, and the CP.
It is a storage device that is used for work of U14. The RAM 20 is a storage device used for the work of the CPU 14. The ROM 22 has a BIOS
Data such as (Basic Input Output System) is stored.

The input / output interface 24 is the CPU 14
The data input / output interface 28 relays data exchange between the CPU 14 and the input device 30. The input / output interface 28 relays data exchange between the CPU 14 and the input device 30. The bus 12 is used for exchanging data and addresses between various parts of the system.

The media reading device 26 includes an FD (floppy (registered trademark) disk), an MO disk (magneto-optical disk), a CD (compact disk) -ROM, and a D.
It is a device for reading data such as programs from an information storage medium such as a VD (digital video disk). In addition,
Here, the program is supplied from the information storage medium, but a device for data communication may be connected to the computer system 10 and the program may be supplied via a communication network such as the Internet.

The input device 30 is, for example, a character input device such as a keyboard, a pointing device such as a mouse,
It includes a microphone and the like for inputting the voice that is the target of voice editing. The voice input from the microphone is digitized,
Hard disk storage device 19 as waveform data file
Memorized in. In addition, the voice to be edited is digitized by another device in advance and stored in the information storage medium as a waveform data file.
It may be read by and stored in the hard disk storage device 19. Alternatively, the waveform data file may be supplied to the computer system via a communication network such as the Internet and stored in the hard disk storage device 19.

A voice editing program is supplied from an information storage medium such as a CD-ROM or a DVD to the computer system 10 having the above-mentioned configuration, and the voice editing program is installed in the hard disk storage device 19 so that the computer system 10 can perform the voice editing. Functions as a device.

FIG. 2 shows an example of a voice editing (waveform editing) screen displayed on the monitor 18 when the computer system 10 activates the voice editing program. The voice editing screen shown in the figure is, for example, a GUI (Graphic User).
Interface) is adopted in the computer system 10, it is displayed on the monitor 18 as one window. As shown in the figure, on the voice editing screen,
The waveform 31 of the voice to be edited is displayed at the top of the screen. A waveform 31 shown in the figure shows the recorded voice with the horizontal axis as the time axis and the vertical axis as the amplitude (peak value).
One audio portion 38, 40, 42 is included.

Here, the voice portion means a waveform portion relating to voice that is not noise, that is, a waveform portion where voice is recognized by the voice recognition processing. The start position (timing) and end position (timing) of each audio portion 38, 40, 42 may be indicated on the audio editing screen by marker display or the like.

From the voice portions 38, 40, 42, a character string representing the voice content is recognized by a known voice recognition process, and all or part of the character string is recognized by the recognition result frames 32, 34 ,. It is supposed to be displayed in 36. Here, of the frame lines of the recognition result display frames 32, 34, and 36, the left vertical line corresponds to the start position of the voice portion, and the right vertical line corresponds to the end position of the voice portion.
In this way, in association with each voice part 38, 40, 42,
Recognition result display frames 32, 34, 36 are displayed,
The character strings which are the voice recognition results of the voice portions 38, 40 and 42 are displayed in the respective frames.

When the recognition results are displayed in the recognition result display frames 32, 34 and 36, the estimation result of whether each voice part is related to the recording target sound or the non-recording target sound depending on the font size. Is displayed. That is, when it is estimated that it is the recording target voice, the recognition result is displayed in a large font, and when it is estimated that it is the non-recording target voice, the recognition result is displayed in a small font. It has become so. Further, if neither can be estimated, the recognition result is displayed in a medium font. In this way, the editor, from the font size of the recognition result displayed corresponding to each voice portion, whether the voice portion is a portion related to the recording target sound, or is a portion related to the non-recording target sound, You can grasp at a glance.

The recording target voice is, for example, a singing voice of a singer, a speech of a conference participant, a dialogue of a voice actor, or the like. On the other hand, the non-recording target voice is, for example, the soliloquy of the recording target person such as a singer, a conference participant, a voice actor, or the voice of the studio staff.

In the voice editing screen shown in the figure, although not shown, a scale indicating the amplitude and time of the waveform 31 and
Toolbars and menus for editors to select various tools for voice editing are also displayed.

FIG. 3 is a flow chart showing an editing screen display routine which is a part of the voice editing program. In addition to the edit screen display routine, the voice editing program also includes various edit routines for editing the waveform 31 displayed by the routine based on an instruction from the editor.

As shown in the figure, in the edit screen display routine, first, the waveform file name, display range, and scale are acquired (S101). The waveform file name is the name (and path if necessary) of the waveform file in which the voice to be edited is recorded. For example, the editor may specify the input device 30 such as a mouse on the menu screen, or the file name. The waveform file name is acquired by inputting the input from the input device 30 such as a keyboard. The display range is a range in which the waveform 31 is displayed on the edit screen of the voice to be edited. For example, the waveform 31 is once displayed on the monitor 18 and then horizontally scrolled by the input device 30 such as a mouse or the GUI. The display range is acquired by changing the window size in and allowing the editor to specify the display range. The reduced scale is a display reduced scale of the waveform 31, and the reduced scale is acquired by an enlargement instruction or a reduction process by the input device 30 such as a keyboard.

Next, the waveform file having the waveform file name acquired in S101 is stored in the hard disk storage device 19
Is read from (S102) and based on it, S101
The waveform 31 of the display range acquired in step S10 is drawn in the RAM 20 or the VRAM included in the image processing unit 16 (S10).
3). Subsequently, voice recognition processing is performed on the waveform file read in S102, each voice portion included in the waveform 31 is specified, and a character string that is the recognition result thereof is generated (S104). The number of phrases F, phrase position of the display range _(Ps i, _Pe i), the phrase number _y i (i = _1~F) is stored in the RAM 20 (S10
5) Here, the phrase number F is the number of voice portions in the display range, and the phrase position Ps _i is the start position (x coordinate (horizontal position)) of the i-th voice portion on the voice editing screen.
And the phrase position Pe _i is _i on the voice editing screen.
It is the end position (x coordinate (horizontal position)) of the th audio part. Further, the phrase character number y _i is the number of characters in the recognition result for the i-th speech portion. After that, the phrase position P
The phrase position Ps _i is subtracted from e _i , whereby the phrase length x _i is calculated (S106). Phrase length x
_i represents the length (number of pixels) in the horizontal direction (time axis direction) of the i-th audio portion on the audio editing screen.

Next, the variable i designating the voice portion is set to 1 (S107), and the font size z _i (in particular, one character is displayed here) when the voice recognition result for the i-th voice portion is displayed. The length (pixel) of one side of the square display area necessary for this is determined (S10).
8). In order to determine the font size z _i , it is first estimated whether the i-th voice portion represents a recording target voice or a non-recording voice. And
The font size is determined based on this estimation result. For example, a large font size is set to the font size z _i when it is estimated that the target voice is recorded, and a font size z is estimated when the target voice is not recorded.
Set a small font size for _i . If neither can be identified, a medium size font size is set. The font size determination process will be described in detail later with reference to FIGS. Above S10
5, RAM as a result of the processing in S106 and S108
The table shown in FIG. 5 is prepared for 20.

Next, the phrase length x _i is the font size z
_It is determined whether it is smaller than _i (S109). When the phrase length x _i is smaller than the font size z _i , the recognition result display frame cannot be completely accommodated in the display area immediately below the voice portion, and the exceptional character drawing process is executed (S111). The exceptional character drawing process is the process shown in FIG. On the other hand, when the phrase length x _i is equal to or _larger than the font size z _i , the recognition result display frame can be drawn in the display area immediately below the voice portion, and the normal character drawing process is executed (S110). The normal character drawing process is shown in FIG. The exceptional character drawing process and the normal character drawing process are both processes of drawing a recognition result display frame below the i-th voice portion and drawing a character string as a recognition result therein.

After the normal character drawing process (S110) or the exceptional character drawing process (S111) is executed, it is judged whether or not the variable i has reached the number F of phrases (S112). If not, 1 is set to the variable i. Is added, and the processing from S108 to S112 is executed again for the next voice portion. In this way, the variable i increases in order from 1 to F, and when the processing from S108 to S112 is executed for each variable i, the monitor 18 displays the voice edit screen drawn up to that point (S114). For example, when the voice edit screen is drawn in the RAM 20, the CPU 1
4 transfers it to the image processing unit 16 and outputs it to the monitor 18 at a predetermined timing, and a voice edit screen is displayed.

As described above, as the voice edit screen,
In addition to displaying the waveform 31, the recognition result display frame and the recognition result can be displayed below each voice portion.

FIG. 5 shows a font size determination process (S10).
It is a flowchart which shows an example of 8). In this example, whether each voice part is a part relating to a recording target voice or a part relating to a non-recording voice is estimated from the sound pressure of the voice part. In this process, first, the sound pressure P _i [dB] of the i-th voice part (phrase i) is calculated by the following equation (1) (S401). Here, the waveform data is 1
It is assumed to be 6-bit PCM data, and X is a value obtained by dividing the sum of absolute values of peak values per unit time by the number of samples. Also, 0x8000 is 8000H (hexa)
Means According to the equation (1), there is no sound at -96 dB and the maximum volume is at 0 dB.

[0033]

## EQU1 ## P _i = 20log ₁₀ (X / 0x8000) (1)

When the sound pressure P _i is calculated in this way, it is then determined whether the sound pressure P _i is larger than a predetermined threshold P1 (for example, -40 dB) (S402). Then, if it is larger than the predetermined threshold P1, "large (for example, 12 pt)" is set as the font size z _i (S404). On the other hand, if it is less than the predetermined threshold P1, it is next determined whether the sound pressure P _i is higher than the predetermined threshold P2 (for example, −60 dB) (S40).
3). Then, if it is larger than the predetermined threshold P2, "medium (for example, 10 pt)" is set as the font size z _i (S405). On the other hand, when the sound pressure P _i is less than or equal to the predetermined threshold P2, the font size z _i is “small (for example, 8 p
t) ”is set (S406).

In this way, the larger the sound pressure P _i of each voice part, the larger the font size z _i can be, and the editor can recognize the characters displayed in the recognition result display frames 32, 34, 36 and the like. By confirming the font size, it is possible to help determine whether the audio portion corresponding to the font size is related to the recording target sound or the non-recording target sound.

Generally, the recording target sound is recorded relatively firmly and the sound volume P _i is considered to be large, whereas the non-recording target sound is considered to have a small sound pressure P _i . It can be said that it is rational to estimate from the sound pressure P _i of each voice part whether or not the voice part relates to the recording target voice.

The sound pressure is calculated by the above equation (1) and is information based on the peak value of each voice part, but other information based on the peak value of each voice part (for example, the peak value of the peak value). It is also rational to judge whether or not the sound part is related to the sound to be recorded, based on the peak value, effective sound pressure, sum of absolute values of peak values, etc.). In any case, whether or not each voice portion is the recording target voice can be appropriately estimated based on the crest value (one or more) of each voice portion.

FIG. 6 shows a font size determination process (S10).
It is a flowchart which shows the other example of 8). In this example, whether each voice part is a part relating to a recording target voice or a part relating to a non-recording voice is estimated from the frequency distribution of the voice part. In this process, first, the frequency distribution of the i-th voice part (phrase i) is calculated (S501). The frequency distribution is information indicating which frequency band and to what extent each voice part contains.

After the frequency distribution is calculated, it is next checked whether or not the frequency distribution is a dialogue type frequency distribution (S50).
2). As the speech-type frequency distribution, for example, an average frequency distribution when many voice actors read various scripts can be adopted. In the processing of S502, for example, S501
It is only necessary to calculate the degree of coincidence between the frequency distribution calculated in step 1 and the previously prepared dialogue type frequency distribution, and determine whether or not it is the dialogue type depending on whether or not it exceeds a predetermined threshold value. When calculated frequency distribution S501 is dialogue type, set the "large (eg 12pt)" in the font size _{z i,} if not speech type, the "small (e.g. 8pt)" to the font size _{z i} Set.

In this way, the font size z _i can be increased when the frequency distribution of each voice portion is close to the speech-type frequency distribution, and the editor can recognize the recognition result display frames 32, 3.
Check the font size of the characters displayed in 4, 36, etc., and determine whether the corresponding voice part is related to the recording target voice (here, dialogue), or the non-recording target voice (here, not dialogue) It can be used to judge whether or not it is related to.

Although the frequency distribution of each voice part is judged to be speech type or non-speech type here, it may be judged whether it is a singing type or a non-singing type. It may be judged whether it is conversational. In any case, each voice part is recorded by preparing an average frequency distribution of the target voice (speech, song, conversation, etc.) in advance and checking if the frequency characteristics of each voice part are close to it. It can be estimated whether or not the target voice is concerned. Although an example using only the frequency distribution is shown here for simplification of the description, it is also very effective to use the dynamic characteristic parameter (how to change) of the frequency distribution in addition to this.

In general, the voice to be recorded is often pronounced directly toward the microphone, whereas the voice to be recorded is not emitted from a distance from the microphone, or is in a direction deviating from the direction of the microphone. Since it is thought that it is often pronounced, there is a difference in both frequency distributions, and it is possible to estimate from the frequency distribution of each voice part whether or not that voice part relates to the recording target voice. Can be said to be rational. Similarly, it can be said that it is rational to estimate whether or not the voice portion is related to the recording target voice based on other frequency information such as the fundamental frequency (f0) of each voice portion and the number of zero crossings. .

An average frequency distribution when a male voice actor reads various scripts and an average frequency distribution when a female voice actor reads various scripts are prepared in advance,
By checking which one of the frequency distributions calculated in S501 is closer, it is determined whether each voice portion is a script read by a male voice actor or a woman voice actor. It is possible to estimate and change the font size and font color accordingly. By doing this, it is possible to use the font size and font color to judge whether each voice part is due to a female voice actor or a male voice actor.

FIG. 7 shows the font size determination process (S10).
It is a flowchart which shows another example of 8). In this example, it is estimated whether each voice part is a part related to a recording target voice or a part related to a non-recorded voice based on the likelihood (likelihood) of the recognition result obtained in the voice recognition process for the voice part. I am trying to do it. That is, in the voice recognition process, in the voice recognition process of S104 (FIG. 3), the recognition result for each voice part is generated.
Calculate the likelihood as a recognition result for many character strings,
The character string with the highest likelihood is used as the actual recognition result. Here, based on the likelihood calculated in S104 for the recognition result for each voice part, it is estimated whether or not the voice part relates to the recording target voice.

In this process, first, the likelihood α _i of the i-th voice part (phrase i) is acquired (S601). In addition,
The likelihood α _i is stored in association with each voice part in the processing of S104.

After the likelihood α _i , it is then determined whether the likelihood α _i sound is larger than a predetermined threshold α1 (S602). If it is larger than the predetermined threshold value α1, the font size z
“Large (for example, 12 pt)” is set as _i (S60
4). On the other hand, if it is less than the predetermined threshold α1, then the likelihood α _i
Is greater than a predetermined threshold value α2 (S60)
3). If it is larger than the predetermined threshold value α2, "medium (for example, 10 pt)" is set as the font size z _i (S605). On the other hand, if the likelihood α _i is less than or equal to the predetermined threshold α 2, the font size z _i is “small (for example, 8 p
t) ”is set (S606).

By doing this, the font size z _i can be increased as the α _{i of} each voice part is higher, and the editor can change the font size of the characters displayed in the recognition result display frames 32, 34, 36 and the like. Can be used to judge whether the audio part corresponding to the audio data is related to the recording target sound or the non-recording target sound.

Generally, the recording target voice is recorded relatively clearly, and the likelihood α _{i of the} recognition result is considered to be large, whereas the non-recording target voice is considered to have a small likelihood α _i. Therefore, it can be said that it is reasonable to estimate from the likelihood α _i of the recognition result for each voice part whether or not the voice part is related to the recording target voice.

Next, FIG. 8 is a diagram showing a normal character drawing routine. As shown in the figure, in the normal character drawing routine, first, the number of lines n that satisfies the following expression (2) is determined (S201).

[0050]

## EQU00002 ## x _i × (n-1) <y _i × z _i ≤x _i × n (2)

Next, it is determined whether the number of rows n is less than or equal to the maximum number of rows N (S202). When the number of lines n is less than or equal to the maximum number of lines N, a recognition result display frame of vertical z _i × n dots and horizontal x _i dots is drawn below the i-th voice portion (phrase) (S
203). At this time, the drawing is performed so that the x coordinate (horizontal coordinate) of the upper left corner of the recognition result display frame is the phrase (start) position Ps _{i of the} i-th voice portion. Then, the recognition result for the i-th voice portion obtained in S104 is drawn in the recognition result display frame with the font size z _i (S20).
4).

On the other hand, when the number of rows n is larger than the maximum number of rows N, vertical z _i × N dots and horizontal x _i are located below the i-th voice portion.
A dot recognition result display frame is drawn (S205). At this time, the drawing is performed so that the x coordinate (horizontal coordinate) of the upper left corner of the recognition result display frame becomes the phrase (start) position Ps _{i of the} i-th voice portion. Then, among the character strings that are the recognition results for the i-th voice portion obtained in S104, the characters whose display on the voice editing screen is omitted are determined (S20).
6). Specifically, the number of abbreviated characters a _i represented by the following equation (3) is calculated, and in the character string as the recognition result, _ai consecutive characters are excluded as abbreviated characters except for the first character and the last character. Pick out. Here, int () is a function that converts the numerical value in parentheses into an integer. The second term represents the number of characters that can be displayed in the recognition result display frame, and the third term is provided to prepare one character number for filling a symbol (for example, "...") in which an abbreviated character exists. There is.

[0053]

[Equation 3]

A _i = y _i −int (x _i / z _i ) × N + 1 (3)

Then, the character not selected as the abbreviated character is drawn with the font size z _i in the recognition result display frame drawn in S205 (S207). At this time, it is preferable to display a symbol indicating that the characters are omitted, such as "..." or "~", in the portion where the omitted characters originally existed.

With the above arrangement, the recognition result display frame can be displayed immediately below the voice portion, and a part or the whole of the character string as the recognition result can be displayed therein. At this time, even when the character string as the recognition result is partially omitted and displayed on the voice editing screen, the first character and the last character are not omitted, so that the editor can easily judge the content of the voice portion. become. Furthermore, the font size of the character string displayed in each recognition result display frame is determined according to the estimation result of whether or not the corresponding voice portion is related to the recording target voice. Therefore, the editor can use the font size to judge whether or not each audio portion is related to the recording target audio.

Next, FIG. 7 is a diagram showing an exceptional character drawing routine. As shown in the figure, in the exceptional character drawing routine, it is first determined whether the following expression (4) is satisfied (S).
301).

[0058]

## EQU00004 ## Ps _i + z _i > Ps _{i + 1} (4)

When the above expression (4) is satisfied, when a character is drawn on the lower side of the voice portion with the font size z _i , the character reaches right below the voice portion on the right side, and there is recognition for the voice portion. Since you can not draw the result,
Interrupt the exception character drawing routine and its parent process, the edit screen display routine, for example, font size z _i ,
Display a message prompting the editor to change the display range and scale.

On the other hand, if the above equation (4) is not satisfied, then it is determined whether the number of rows n is the maximum number of rows N or less (S3).
02). When the number of lines n is equal to or less than the maximum number of lines N, a recognition result display frame of vertical z _i × n dots and horizontal z _i dots is drawn below the i-th voice portion (phrase) (S303). At this time, the x coordinate (horizontal coordinate) of the upper left corner of the recognition result display frame is i.
It is drawn so as to be the phrase (start) position Ps _{i of the} th voice part. After that, S104 is displayed in the recognition result display frame.
The recognition result for the i-th speech portion obtained in step S3 is drawn with the font size z _i (S304). Here, the recognition result is displayed vertically.

On the other hand, when the number of rows n is larger than the maximum number of rows N, vertical z _i × N dots and horizontal z _i are located below the i-th voice portion.
A dot recognition result display frame is drawn (S305). At this time, the drawing is performed so that the x coordinate (horizontal coordinate) of the upper left corner of the recognition result display frame becomes the phrase (start) position Ps _{i of the} i-th voice portion. Then, among the character strings that are the recognition results for the i-th voice portion obtained in S104, the characters to be omitted from the display on the voice edit screen are determined (S30).
6). Specifically, the number of abbreviated characters a _i shown in the above equation (3) is calculated, and in the character string that is the recognition result, _ai consecutive characters are excluded as abbreviated characters except for the first character and the last character. Pick out.

Then, the character not selected as the abbreviated character is drawn with the font size z _i in the recognition result display frame drawn in S305 (S307). At this time, it is preferable to display a symbol indicating that the characters are omitted, such as "..." or "~", in the portion where the omitted characters originally existed.

With the above arrangement, the recognition result display frame can be displayed below the voice part, and a part or all of the character string as the recognition result can be displayed therein. At this time, if the right side frame line (vertical line) of the recognition result display frame enters the part of the speech part on the right side where the recognition result display frame should be displayed, the process is interrupted. .

According to the voice editing apparatus (voice editing program) described above, the waveform 31 is displayed on the voice editing screen, and the voice portions 38, 40, 42 are displayed in the recognition result display frames 32, 34, 36. Since the line (vertical frame line) is divided from the other part (non-voice part) to be represented, the editor can grasp the voice part included in the waveform 31 at a glance. Further, the recognition result display frames 32, 34, 36 are displayed in association with the voice parts 38, 40, 42, respectively, and all or part of the recognition results of the voice parts 38, 40, 42 are displayed therein. Therefore, the editor can judge what the contents of each audio part included in the waveform 31 are, at a glance. Furthermore, since the font size at that time is determined according to the estimation result of whether or not each voice part is related to the recorded voice, the editor looks at the font size and Can be used to determine whether or not the voice is related to the recording target voice, and the efficiency of voice editing can be significantly improved.

The present invention is not limited to the above embodiment.

For example, the voice editing screen is not limited to that shown in FIG. 2, and various patterns can be adopted. For example, as shown in FIG. 10, each voice part 58, 60, 62
In association with the recognition result display frame 52,
54 and 56 are displayed, respectively, and markers 66, 68, and 70 are displayed on the upper side.
6, 68, 70 of the size of each voice portion 58, 60, 62
May be determined based on the estimation result of whether or not is related to the recording target voice. In addition, the voice part 58,
The colors of 60 and 62 themselves or the background color thereof may be changed according to the same estimation result. In any case, by displaying whether or not each voice portion is related to the recording target voice in association with the voice portion, the audio editing efficiency can be significantly improved.

[0067]

As described above, according to the present invention,
While displaying the waveform of the voice, by judging the voice portion included in the waveform, it is estimated whether each voice portion relates to the recording target voice or the non-recording target voice, and it is estimated in association with each voice portion. Since the result is displayed, the editor can easily judge whether each audio part relates to the recording target sound or the non-recording target sound according to the display content, and the editing efficiency can be improved. it can.

[Brief description of drawings]

FIG. 1 is a diagram showing a configuration of a computer system that functions as a voice editing device according to an embodiment of the present invention.

FIG. 2 is a diagram showing an example of a voice edit screen.

FIG. 3 is a flowchart illustrating a voice edit screen display routine.

FIG. 4 is a diagram showing a table generated in a voice edit screen display routine.

FIG. 5 is a flowchart illustrating an example of a font size determination routine.

FIG. 6 is a flowchart illustrating another example of a font size determination routine.

FIG. 7 is a flowchart illustrating still another example of the font size determination routine.

FIG. 8 is a flowchart illustrating a normal character drawing routine.

FIG. 9 is a flowchart illustrating an exceptional character drawing routine.

FIG. 10 is a diagram showing a modification of the voice editing screen.

[Explanation of symbols]

10 computer systems, 12 buses, 14 CP
U, 16 image processing unit, 18 monitor, 19 hard disk storage device, 20 RAM, 22 ROM, 24,
28 input / output interface, 26 media reading device, 30 input device, 31, 64 waveform, 38, 4
0, 42, 58, 60, 62 voice part, 32, 34,
36, 52, 54, 56 Recognition result display frame, 66, 6
8,70 (Estimation result display) Marker.

─────────────────────────────────────────────────── ─── Continuation of front page (51) Int.Cl. ⁷ Identification code FI G10L 17/00 G10L 3/00 551B (56) References JP-A-8-63186 (JP, A) JP-A-7-49695 ( JP, A) JP 10-222187 (JP, A) JP 11-109988 (JP, A) JP 2002-297188 (JP, A) (58) Fields investigated (Int.Cl. ⁷ , DB) Name) G10L 15/00-17/00 G06F 3/16

Claims (6)

(57) [Claims] 1. A waveform display unit for displaying a waveform of a voice, a voice portion determination unit for determining a voice portion included in the waveform, and each voice portion included in the waveform includes a recording target voice and a non-recording target voice. An estimating means for estimating which one of the waveforms is related, a voice recognizing means for performing a voice recognizing process on the waveform, and a voice recognizing process for each voice part included in the waveform.
The font size of the character string that is the recognition result in
The audio part was estimated to be recorded by the estimation means.
If the size is estimated to be non-captured, in the first size
Determines a second size smaller than the first size
Font size determining means and the font determined by the font size determining means
Depending on the size, for each audio part included in the waveform
All the character strings that are the recognition results in the voice recognition process
Or a character string display means for displaying a part of the character string in association with each of the sound parts. 2. The voice editing apparatus according to claim 1, wherein the estimating unit determines whether the voice portion is a recording target voice or a non-recording target voice, based on a peak value of each voice portion included in the waveform. An audio editing apparatus, which estimates whether or not it is related to. 3. The audio editing apparatus according to claim 1, wherein the estimating unit includes a recording target voice and a non-recording target voice based on frequency information of each voice unit included in the waveform. An audio editing apparatus, which estimates which one of the above. In audio editing apparatus according to any one of claims 4] claims 1 to 3, before Symbol estimating means, on the basis of the likelihood of the recognition result of the speech recognition processing for each speech portion included in said waveform,
An audio editing apparatus, which estimates whether the audio part relates to a recording target sound or a non-recording target sound. 5.The voice according to any one of claims 1 to 4.
In the editing device, In the font size determining means, the audio part is
It is not possible to distinguish between recorded and non-recorded objects by the fixed means.
If not, the voice recognition for the voice part For processing
The font size of the character string which is the recognition result
Smaller than one size and larger than the second size
A voice editing device characterized by deciding to a third size
Place 6. A step of displaying a waveform of a voice, a step of determining a voice portion included in the waveform, wherein each voice portion included in the waveform relates to a recording target voice and a non-recording target voice. Whether or not the voice recognition processing is performed on the waveform, and the voice recognition processing for each voice portion included in the waveform.
For the character string that is the recognition result in the processing step
The size of the speech is estimated by the step of estimating the speech part.
If it is estimated that
If it is estimated to be recorded, it is smaller than the first size
The second size and the font size.
Depending on the font size, each audio part included in the waveform
Sentence which is the recognition result in the voice recognition process for the minute
A voice editing program for causing a computer to execute the step of displaying all or part of a character string in association with each voice portion.