JP2008275836A - Document processing method and device for reading aloud - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a document processing method and a device for reading aloud, in which a suitable modification means for a sound signal is attained in order to read letter information aloud in voice without redundancy. <P>SOLUTION: This invention relates to an information processing device comprising at least a data input and output device, a data recording device, and an arithmetic processing device, wherein a document which is a reading aloud indication letter string composed of a control code and a ground sentence, is read. Relation of the control code, an effect which is a content of modification of the ground sentence, and the number of effects, is read, and the document is separated into the control code and the ground sentence, and a combination of each control code and the ground sentence controlled by the control code, is obtained and a waveform data for reading the ground sentence aloud is generated. A ground sentence reading waveform data is modified according to the control code corresponding to the ground sentence, and the ground sentence reading waveform data and the waveform data in which the waveform is modified are combined and the obtained waveform data is output. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention provides a data input / output device, a data recording device, and an information processing device such as a personal computer equipped with an arithmetic processing device, which inputs and processes a text-to-speech document and outputs waveform data. The present invention relates to a document processing method and an apparatus for executing the document processing method.

In order to provide character information, there is a method of appealing to visual and auditory senses, and when character information is modified, there are visual modification and audio modification.
In the expression of visual modification, it is possible to modify the background sentence with various character decorations such as the size and color of the characters.
For example, when a web page written in general HTML is displayed in a general browser, if it is a part surrounded by anchor tags (<A> to </A>), Is expressed in different colors and underlines, etc., and it is easily understood visually that it is linked to other information. This is not mere character information, but indicates that other meanings are included by the modification of color and a plurality of pieces of information are attached.

On the other hand, it is possible to carry out a simple reading of the text in the ground using a conventionally known technique, but the processing of the modified character portion is often left as it is.
In visual expression, there are various appropriate and standard modification methods such as color, size, font type, etc., while in audio expression, modification is appropriate and standard for local sentences. This is because no method has been established. Otherwise, you have to take the redundant approach of reading both the local text and the control code in the document.

  Although it may seem unnecessary technology for healthy people, it would be anxious for visually impaired people to read out textual information in a state where redundancy is eliminated and information is sufficiently attached. is there. In addition, there are cases where information disparities occur in urban areas and rural areas, but it is also true that there is a disparity between healthy people and visually impaired persons in web-based information that has spread with visual precedence. Therefore, there is an urgent need for a technique for presenting document contents that include information beyond the local sentence by another method other than vision.

In recent years, voice guidance services in machine language have been expanded.
For example, depending on the bank, balance inquiry and transfer processing can be performed using telephone voice guidance, but a menu format based on voice guidance is often used. For example, the format is “Main menu. Press 1 to check balance, press 9 for others”. Everything is in the same tone, and the standard “Main Menu” is read out every time, and all the numbers corresponding to the menu are read out, so there are many people who are very redundant and frustrated.

Conventional techniques for reading out character information include Patent Documents 1-6.
JP 2004-198830 "Text reading system and method" Japanese Patent Laid-Open No. 2003-345378 “Text Reading System and Method” Japanese Patent Application Laid-Open No. 2003-223181 “Character-Voice Conversion Device and Portable Terminal Device Using the Same” Japanese Patent Laid-Open No. 2002-132282 “Electronic Text-to-Speech Device” Japanese Patent Laid-Open No. 2002-55925 “Speech-to-speech device and information processing device” Japanese Patent Laid-Open No. 2001-188777 “Method for associating speech with text, computer for associating speech with text, method for generating and reading a document with a computer, computer for generating and reading a document, method for performing speech playback of a text document with a computer, Computer for sound reproduction and method for editing and evaluating text in a document "

  However, none of the prior arts could read out the text information in an easy-to-understand manner with no redundancy.

  Therefore, an object of the present invention is to provide a document processing method and apparatus for reading that realizes appropriate modification means for a voice signal in order to read out character information in a voice in an easy-to-understand manner without redundancy.

  In order to solve the above-described problems, a reading-out document processing apparatus of the present invention has the following configuration. That is, in an information processing apparatus including at least a data input / output device, a data recording device, and an arithmetic processing device, a document input means for reading a document that is a reading instruction character string composed of a control code and a local sentence; The control code, the input that reads the relationship between the effect that is the processing content of the ground sentence controlled by the control code, and the number of the effects, and the related input means that separates the document into the control code and the ground sentence, Document processing means for obtaining a set of a control code and a local sentence controlled by the control code, a local sentence reading waveform data generating means for generating waveform data for reading the local sentence, and a local sentence reading waveform data Waveform processing means for processing the sound according to the control code corresponding to the local sentence, Characterized in that it comprises a waveform data synthesizing means for synthesizing the waveform processing waveform data, the waveform data output means for outputting the waveform data obtained.

  The document processing method for reading out according to the present invention is a document which is a reading instruction character string composed of a control code and a local sentence in an information processing apparatus including at least a data input / output device, a data recording device, and an arithmetic processing unit. Is read, control code, the effect of the processing of the local sentence controlled by the control code, and the number of the effect, the relationship is read, the document is separated into the control code and the local sentence, each control A set of a code and a local sentence controlled by the control code is obtained to generate waveform data that reads the local sentence, and the local sentence reading waveform data is converted into a control code corresponding to the local sentence. Processing, and synthesizing the text-to-speech read-out waveform data and the waveform processed waveform data, and outputting the obtained waveform data. To.

  Here, when there are a plurality of effects, waveform processing corresponding to the plurality of control codes may be applied to the ground text-read waveform data so as to correspond to various effects.

  In addition, it is also possible to inspect the ground sentence reading waveform data in advance, and when there is no ground sentence to be read out, the output of the silent part may be deleted to reduce the redundancy.

  The processing content of the local sentence controlled by the control code may be expressed by a change in volume.

  Similarly, the processing content of the local sentence controlled by the control code may be expressed by a change in the amplitude of the specific frequency.

  The processing content of the local sentence controlled by the control code may be expressed by waveform deformation.

  The processing content of the local sentence controlled by the control code may be expressed by a change in reproduction speed.

  The processing content of the local sentence controlled by the control code may be expressed by a temporal change in the reproduction timing.

  The processing content of the local sentence controlled by the control code may be expressed by a pitch shift that changes the frequency.

  For the pitch shift, a change in accordance with classical rhythm is suitable.

  Similarly, a change in accordance with the equal temperament in which the pitch is divided by an equal frequency ratio is also suitable for the pitch shift.

  If the pitch shift is changed to halve or double the frequency, it will be easy to hear.

  Similarly, even if the pitch shift is turned to change the order of the chord constituent sounds, the sound can be easily heard.

  According to the present invention, a document composed of a control code and a ground sentence is read, and waveform data processed according to the control code is output, which contributes to reading out text information in an easy-to-understand manner without redundancy. . It can also be used as a standard for modifying audio signals.

Embodiments of the present invention will be described below with reference to the drawings.
There are many types of effects that change the sound. Those that change playback timing over time such as echoes, those that deform waveforms such as overdrive simulators, those that change the frequency such as voice changers, those that change the amplitude of a specific frequency such as an equalizer, those that change the volume or speed, etc. There is.

Considering the types of effects that can be applied to people's utterances and clearly understood even when others hear the utterances with those effects, pitch shift effects that change the frequency without losing clarity are most suitable. It is thought that there is.
Therefore, in the following description, pitch shift will be described, but other effects can be used as well.

In general, people are considered to be able to fully understand the utterances regardless of whether they are loud or low in voice, so the pitch of the voice that is being read may be slightly higher or lower. However, it should not have a significant impact on ease of listening.
When two or more people talk about different things at the same time, it is difficult for a person to understand the story. However, even if there are two or more speakers, It is possible for a person to understand the story. This is indicated, for example, when a choir consisting of a plurality of parts sings the same lyrics in various parts simultaneously, the audience can understand the lyrics.
This kind of recognition is not only based on the understanding of the original sound that is being uttered and the pitch-shifted sound created based on the original sound, but also on the simultaneous transmission of the original sound and the pitch-shifted sound, that is, the person understands even the chord. It means you can do it.

For pitch shift, equal temperament can be used effectively.
The equal temperament is a temperament in which a pitch such as one octave is divided by an equal frequency ratio. In general, it often refers to the twelve equal temperament, which is familiar to modern people.
FIG. 1 is a table showing frequency ratios according to the twelve equal temperament.
The twelve equal temperament is a temperament obtained by dividing one octave into 12 equal parts, and the frequency ratio of adjacent sounds is the ratio of the 12th root of 2 to 1, which corresponds to a semitone of Western music.

  In the present invention, since the pitch shift is mechanically performed, it is not necessary to consider the difference from the classical rhythm peculiar to the twelve equal temperament. For the convenience of calculation, the classical rhythm may be preferable. Therefore, in the following, when describing a sound that is 5 degrees higher than the original sound, it means that the sound is 7 semitones higher than the original sound as a twelve average temperament. It is assumed to include the duality of 1.5 times the frequency of the original sound as the original classical rhythm. For convenience, a sound one semitone higher is a sound 100 cents above, a sound one semitone lower is a sound 100 cents lower, and one octave is represented as 1200 cents.

If you press C and G at the same time on the piano, you will hear a so-called chord between S and D in C major. Seo is 5 degrees higher than De. In other words, it is 7 semitones higher or 700 cents higher. In the classic classical tune, Seo should be 1.5 times the frequency of the de, but twelve equal-tempered instruments like the piano are well tuned because all errors except for the octave are included. However, it is not 1.5 times, but it is about 1.498307 times, which is the 12th root of 2 7.
In terms of the strict calculation amount, the calculation amount of 1.5 times in the classical rhythm is much smaller than the calculation of about 1.493307 times the sound of the twelve average temperament.

  The sound indicated by the pitch shift is a semi-tone do # (100 cents rise), a second higher second (200 cents rise), a third higher third # (300 cents rise), For example, there is a third-longer Mi (up 400 cents), a fourth-up pha (up 500 cents), a declining fifth-in-fa # (up 600 cents), and a fifth-up seo (up 700 cents).

When a person obtains a chord from the auditory sense, a sense different from that obtained visually is obtained. Therefore, it is desirable to selectively synthesize a chord using this characteristic. It is painful to continue to listen to the unsounding sound, so it is basically desirable to repeat the sound with a beautiful sound when actually using it.
In general, when two notes are at a simple integer frequency ratio, the chord will sound beautifully. Such a pitch is called a pure pitch, and the tuning method uses this to determine the scale.

  As for the equal temperament, only the twelve equal temperament that divides one octave into 12 is not available. As other equal temperaments, nineteen equal temperament, twenty-two equal temperament, thirty-one average temperament, fifty-three equal temperament, seventy-two equal temperament can be used.

  In addition, as the classical temperament, Pythagoras temperament, pure temperament (C major), medium whole temperament (Aron temperament, mean tone), werkmeister temperament I, II, III, kilnberger temperament I, II, III, Valotti Young temperament, Young temperament ii can also be used.

The implementation of pitch shift is as follows.
FIG. 2 shows an example of a document to be processed. FIG. 2 (a) is a character string that clearly indicates a control code, and FIG. 2 (b) is a character string that clearly indicates a visual expression.
The document is composed of control codes and ground sentences. In the example shown in the figure, the text is composed of a local sentence described in HTML and a tag as a control code. In line with HTML 1.0, B tags are in bold, I tags are in italics, and so on.
A general visual display after tag interpretation is an expression including bold characters as shown in FIG.

FIG. 3 shows an example of a document that is a reading instruction character string. FIG. 3A is a character string that clearly indicates a control code, and FIG. 3B is a character string that is accompanied by a phonetic expression.
In the phonetic expression according to the present invention, first, in the portion where the tag exists, the sentence of the ground is divided into “all citizens”, “right to work”, “has a duty”, and “is obligated”.
Assuming that the B tag is an effect of pitch shift sound that is three times lower, the text of the ground that is not under the control of the B tag (such as “All the people are”) is synthesized and transmitted as it is, and the B tag is controlled. For the local text (such as “Right to Work”), first create sound information such as a PCM file, which is a speech synthesis of the local text as it is, and then 3 times lower pitch-shifted sound based on that sound information. Information is created, and finally, the sound information of the local sentence and the pitch shift sound information are synthesized and transmitted. For synthesis of sound, conventionally known techniques are appropriately used.
Then, all the local sentences are read out, but as shown in FIG. 3 (b), only the control part by the B tag is read out by the chord that the simultaneous reading of the reading sound is three times lower than the reading of the local sentence. . As a result, the listener can understand the presence or absence of the tag based on the difference between a single tone and a chord.

Similarly, FIG. 4 is an example showing a document to be processed. FIG. 4A is a character string that clearly indicates a control code, and FIG. 4B is a character string that clearly indicates a visual expression.
In this example, a plurality of tags are used, and different pitch shifts are applied to different portions.
As an example of the misappearance of multiple tags, I tags that are different tags are used.
A general visual display after tag interpretation is an expression including bold and italics as shown in FIG.

FIG. 5 shows an example of a document that is a reading instruction character string. FIG. 5A is a character string that clearly indicates a control code, and FIG. 5B is a character string that is accompanied by a phonetic expression.
Assuming that the I tag is an effect of a pitch shift sound that is five times higher, first divide it into the parts where the tag exists, such as “All citizens”, “Right to work”, “Has obligation”, and B Sentences that are not controlled by the tag or I tag ("ha", "has") are sent out as they are, and under the control of the B tag, the local sentence ("right of work" , “I have no obligation”), synthesizing the speech as it is, creating the sound information, then creating pitch-shifted sound information three times lower based on the synthesized sound information, and finally, Sound information and pitch-shifted sound information are synthesized and transmitted, and under the control of the I tag, first the local sentence (“all citizens”) is synthesized as it is, then the sound information is created, and then the synthesized sound Based on the information, 5 degree higher pitch shift sound information is made, and finally, the sound information and pitch shift of the local sentence The sound information is synthesized and outgoing.
Then, all the local sentences are read out, but as shown in Fig. 5 (b), only the B tag part is read out by a chord that is a simultaneous reading of a reading sound three times lower than the reading of the local sentence. Only the part is read out by a chord of simultaneous sending of a reading sound that is 5 degrees higher than the reading of the local sentence. Thereby, the user can understand the presence or absence of the tag by the difference between a single tone and a chord.

Similarly, FIG. 6 is an example showing a document to be processed. FIG. 6A is a character string that clearly indicates a control code, and FIG. 6B is a character string that clearly indicates a visual expression.
In this example, a plurality of tags of B tag and I tag are used, and there is a local sentence under the control of the plurality of tags.
A general visual display after tag interpretation is an expression including bold and italics as shown in FIG.
In other words, “all citizens” are unchanged, “work” under B tag control is bold, “right” under B tag and I tag control is bold italic, “has” under I tag control is “Obligations” under italics, B tag and I tag controls are in bold italics, and “O” under B tag controls are in bold. Thus, in visual expression, bold and italics are effects that can be applied without interfering with each other. This can also be done by voice.

FIG. 7 shows an example of a document that is a reading instruction character string. FIG. 7A is a character string that clearly indicates a control code, and FIG. 7B is a character string that is accompanied by a phonetic expression.
In other words, “all the people” are unchanged, “work” under B tag control is a chord with a pitch shift sound that is three times lower, and “right” under B tag and I tag control is a pitch shift sound that is three times lower. A chord with a pitch shift sound that is 5 degrees higher, “has” under I tag control is a chord with a pitch shift sound that is 5 degrees higher, and “obligation” under B tag and I tag control is a pitch shift sound that is 3 degrees lower. A chord with a pitch shift sound that is 5 degrees higher, and “Otofu” under B tag control, becomes a chord with a pitch shift sound that is 3 degrees lower.
Thus, even in the audio expression, a plurality of pitch-shifted sounds are effects that can be applied without interfering with each other.

FIG. 8 is a flowchart of a program for implementing the above-described processing, and FIG. 9 is a flowchart of a subroutine part of waveform processing in the flowchart.
The main part is processed in the following steps.
Initialization:
step 1;
A correspondence table T in which control code E-number of effects k-effect X (K) (K = 1 to k) is written is read.
The control code E corresponds to the tags B and I and the like, and the effect X (K) specifies the content of the effect to be given. For example, when the effect is limited to pitch shift, 100 means pitch shift 100 cents above the original sound, and -400 means pitch shift 400 cents below. The correspondence table T is a table describing the relationship between the control code E and the effect number. For example, if the control code E is B, the effect is -300, if the control code E is I, the effect is 700, and if it is any other control code The effect can be edited as appropriate by the user, such as 0. There is a control code E that does not give any effect, and different control codes E may give the same effect.

Step 2;
Read a document D to be read out consisting of a local sentence Z and a control code E.
Document D is a hybrid sentence composed of a local sentence Z and a control code E, and always includes one or more local sentences Z.

Step 3;
The integer variable i is initialized as 1 and used as a loop counter until all the local sentences Z are processed.

Document processing:
step 1;
Document D is divided into control code E and earth sentence Z, and earth sentence Z is divided at the position where control code E was present. The divided number is assigned to the integer variable N.

Step 2;
The local sentence Z is assigned to the character string array Z (n) (n = 1 to N).

Step 3;
The control code E applied to the local sentence Z is substituted into a character string array C (n) (n = 1 to N) composed of a combination of the delimiter z and the control code E.
The character string array C (n) can be described in a format of “delimiter character z control code E”, for example. When a plurality of control codes E are applied to the local sentence Z, it is assigned to the character string array C (n) as “separator character z control code first E1 separator character z control code second E2”. If the control code E is not applied to the local sentence Z, null is assigned to the character string array C (n).

Generation of read-out waveform data of local sentence:
Waveform data WZ (i) that reads out the local sentence Z (i) is created by using a conventionally known speech synthesis routine or the like.

Waveform data initialization for effects:
The effect waveform data WE is initialized as silence data having the same length as the read-out waveform data WZ (i). The effect waveform data WE is waveform data obtained by applying an effect to the read-out waveform data WZ (i).

Judgment of effect designation:
If the character string array C (i) is null, the effect processing is unnecessary and the process proceeds to the next processing.
If the effect means erasure other than null, the effect waveform data WE is substituted into the read-out waveform data WZ (i) and processed. In the case of effects other than erasure, the process is shifted to a waveform processing subroutine to apply the effect.

Waveform processing:
Processing described below is performed.

Combining ground text-to-speech waveform data and machining waveform data:
The waveform of the reading waveform data WZ (i) and the effect waveform data WE is synthesized and substituted into the reading waveform data WZ (i).
The waveform data is synthesized by using a conventionally known technique. For synthesis of waveform data, sox (SOund eXchange) can be used. For example, when combining inputfile1.wav and inputfile2.wav and assigning to outputfile1.wav, describe in the format `` soxmix inputfile1.wav inputfile2.wav outputfile1.wav ''. In this case, after combining processing, outputfile1.wav Rename wav to inputfile1.wav.

Judgment of end of sentence processing:
While the unprocessed local sentence Z remains (i <N), the counter is incremented and the loop process is repeated. If all the local sentences are processed (i == N), the next process Go to.

Resulting waveform data output:
If the read-out waveform data WZ (1) to WZ (N) are sequentially output, all the sentences Z in the document D to which the respective effects are applied are read out by the control code E.
The output of the read-out waveform data WZ (1) to WZ (N) may be a file output such as a wav file or an audio output from an attached speaker.

Further, a general HTML document includes a body in <BODY> to </ BODY> tags, and there are often other descriptions in <HEAD> to </ HEAD> tags. If the information inside the <HEAD> tag is read out as a local sentence every time, it may be redundant.
To avoid this, the effect of the <HEAD> control code can be set to silence (erase). That is, the local sentence is examined in advance, and if there is no local sentence to be read out, the output of the silent part may be deleted.

The subroutine processing of the waveform processing shown in FIG. 9 is performed in the following steps.
Initialization:
Step 1;
The number of control codes E in the character string array C (i) is counted and assigned to the integer variable M.

Step 2;
The following loop is repeated until all M control codes E in the character string array C (i) are processed. The integer variable j is initialized as 1 and used as a loop counter.

Determine effect from control code:
Get the jth control code E of the character string array C (i). Next, referring to the correspondence table T, the number k of effects to be applied and the effect X (K) corresponding to the control code E and the contents thereof are obtained.
This means that the effect number k and the effect X (K) are uniquely determined from the control code E. For example, in the table T, when the control code “B” is associated with the effect number 1 and the effect “pitch -300”, E = “B”, k = 1, X (1) = ”pitch -300 ".

Apply effects to temporary effect waveform data:
Temporary waveform data WT (K) (K = 1 to k) is created by applying the effect specified by the effect X (K) to the waveform data WZ (i) for reading the original sound. Temporary waveform data WT (K) is a variable for storing waveform data to which effects are temporarily applied.
For example, if effect X (1) is pitch shifted 300 cents below the original sound (X (1) = ”pitch -300”), the waveform data WZ (i) for reading the original sound is inputfile1.wav, and the effect is If the applied temporary waveform data WT (1) is outputfile1.wav and expressed using sox, it can be expressed in the format `` soxinputfile1.wav outputfile1.wav pitch -300 ''. can get.
When the number of effects k is plural, a plurality of effects are individually applied to the reading-out waveform data WZ (i), and the waveform data individually obtained from the plurality of effects becomes individual temporary waveform data WT (K).

Also, at the end of this process, the temporary waveform data WT (K) to which the effect X (K) is applied is inspected, and the main frequency in the data exceeds a predetermined threshold frequency such as 3400 Hz, for example, A process of halving the frequency (“pitch-1200”) is performed. If the frequency is below a predetermined threshold frequency such as 300 Hz, a process of doubling the frequency (“pitch 1200”) may be performed. As a result, when the pitch shift sound is too high, it is lowered by one octave, and when the pitch shift sound is too low, it is raised by one octave.
This has the same effect as turning to change the order of arrangement of the chord constituent sounds.

Since the frequency band of a general telephone is limited to about 300 Hz to 3400 Hz, it is possible to avoid generation of sound that cannot be reproduced by the above processing, which is effective when reading out via a general voice line. Also, the speaker attached to the telephone may be manufactured on the premise of this frequency band, so this process is effective even when reading out using a telephone using some application software without using a voice line. .
The frequency threshold value for determining whether or not to shift the pitch is preferably designed so that the user can adjust it to some extent to suit not only the characteristics of the speaker but also the audible range of the user.

Combining all obtained effects with effect waveform data:
The obtained waveform data WE to which the effect is applied and the waveform of the temporary waveform data WT (K) to which the effect is applied are synthesized and substituted into WE.

Determining the end of effect processing for a single sentence:
While the unprocessed effects remain (j <M), the counter is incremented and the loop processing is repeated. When all the effects are processed (j == M), the loop is terminated and this subroutine is executed. finish.

FIG. 10 shows an example of a document to be processed. FIG. 10 (a) shows a character string specifying a control code, FIG. 10 (b) shows a character string showing a visual expression, and FIG. A character string that clearly indicates another visual expression.
In the example of the B tag and the I tag described above, it is a means for processing a bold or italic expression by voice. For example, the A tag (anchor tag) may contain various information. That is, there are cases where the obtained information is spread by expressing it depending on the control code and its contents.

  In the case of the document illustrated in FIG. 10 (a), according to a normal browser, as shown in FIG. 10 (b), the character part of “Patent Office” under the control of the A tag is discolored and underlined. Is displayed. As a result, the information that the A tag exists is attached to the text portion of “Patent Office”, and clicking there moves to another page. Furthermore, if there is a history of visiting the page, as shown in FIG. 10 (c), the text portion of “Patent Office” is further changed to another color, and information on the visit history is attached. This means that not only the A tag but also the contents of the control code included in the A tag and the history managed by the browser are combined.

It is possible to give an effect with respect to the contents of such a control code by voice expression.
FIG. 11 shows an example of a document that is a reading instruction character string. FIG. 11A is a character string that clearly indicates a control code, FIG. 11B is a character string that is accompanied by a phonetic expression, and FIG. C) is a character string to which another phonetic expression is added.
In the above-described embodiment, the original sound and the pitch-shifted chord are used. However, it is also possible to accumulate three or more chords, and even the same tag can be changed by a variable in the tag. For example, there is a general URI (Uniform Resource Identifier) indicated by HREF in the A tag. The algorithm for recognizing the contents of URI (specified in RFC2396) uses a conventionally known technique.
In this case, the description in the A tag is searched for HREF = U, the initialization in the main routine is performed so that the description U is disassembled and matched with the contents to select an effect. The contents of the control tag are also described in the correspondence table T in which the effect relation is described.

It is indicated by URI when assuming the effect of pitch shift sound that is unread site (long third), read site (short third), port 80 (genus sound), protocol https (short seventh) If the link destination is a page that uses port 80 on an already-read site and adopts https, the original sound is the main sound, and a chord of 3rd minor + genera + 7th minor is synthesized, similar to the so-called minor seventh code The sound of.
The major third and minor third are semitone differences, and if you send them simultaneously, it does not sound beautiful, but since the site is an alternative of whether it is unread or read, you do not actually need to listen to this chord, Different speech expressions can be made as shown in FIG.
In addition to making it depend on the URI indicated by HREF in the A tag, it is also possible to change the chord depending on the accesskey in the same A tag.

Example:
As described above, if the pitch-shifted sounds are selectively stacked based on the twelve equal temperament, etc., it is possible to present a plurality of information with a sound having a beautiful sound that is unlikely to be uncomfortable.
An example of such a case is a telephone voice guidance service. For example, depending on the bank, operations such as balance inquiry and transfer processing can be performed using telephone voice guidance, but a menu format based on voice guidance is often used.
FIG. 12 shows a voice guidance menu, FIG. 12 (a) shows the words spoken by the conventional voice guidance, and FIG. 12 (b) shows the words spoken by the voice guidance according to the present invention. is there.
Conventionally, all words were read out as shown in FIG. In order to avoid this problem, the present invention omits reading out the numbers by making the user know in advance the voice corresponding to the numbers such as 1, 2, 9 and the like. In addition, the “main menu” “is”, “other transactions” “transactions”, “please press the button” “no button”, etc. It is also easy to simplify by skipping part utterances.

  According to the present invention, character information that has been visually modified or accompanied by information can be expressed by voice modification. Therefore, it is possible for a visually impaired person to obtain sufficient visual information by voice, and it is possible to provide a voice guidance service that is easy to understand without being redundant.

Table showing the frequency ratio in the twelve equal temperament It is an example showing a document to be processed, (a) is a character string that clearly specifies the control code, (b) is an explanatory diagram of the character string that clearly indicates the visual expression It is an example which shows the document used as the reading-out instruction | indication character string corresponding to FIG. 2, (A) is a character string which specified control code, (B) is explanatory drawing of the character string which added audio | voice expression. It is an example showing a document to be processed, (a) is a character string that clearly specifies the control code, (b) is an explanatory diagram of the character string that clearly indicates the visual expression It is an example which shows the document used as the reading instruction | indication character string corresponding to FIG. 4, (A) is a character string which specified control code, (B) is explanatory drawing of the character string which added audio | voice expression. It is an example showing a document to be processed, (a) is a character string that clearly specifies the control code, (b) is an explanatory diagram of the character string that clearly indicates the visual expression It is an example which shows the document used as the reading instruction | indication character string corresponding to FIG. 6, (A) is a character string which specified the control code, (B) is explanatory drawing of the character string which added audio | voice expression. Flow chart of main part of program for processing implementation Flow chart of subroutine part of waveform processing It is an example showing a document to be processed, (A) is a character string that clearly specifies a control code, (B) is a character string that clearly specifies a visual expression, and (C) is an explanatory diagram of a character string that specifies another visual expression FIG. 11 is an example showing a document that is a reading instruction character string corresponding to FIG. 10, (A) is a character string that clearly indicates a control code, (B) is a character string that is accompanied by a phonetic expression, and (C) is another voice expression. Explanatory drawing of the character string with the The voice guidance menu is shown, (a) is a wording uttered by the conventional voice guidance, (b) is an explanatory diagram of the words uttered by the voice guidance according to the present invention.

Claims (14)

In an information processing apparatus including at least a data input / output device, a data recording device, and an arithmetic processing device,
Document input means for reading a document which is a reading instruction character string composed of a control code and a ground sentence,
Related input means for reading the relationship between the control code, the effect that is the processing content of the local sentence controlled by the control code, and the number of the effect,
A document processing means for separating a document into control codes and ground sentences, and obtaining a pair of each control code and a ground sentence controlled by the control code;
A local text-to-speech waveform generating means for generating waveform data to read a local text;
Waveform processing means for processing the local text-to-speech waveform data according to the control code corresponding to the local text;
Waveform data synthesizing means for synthesizing the text-to-speech reading waveform data and the waveform processed waveform data;
And a waveform data output means for outputting the obtained waveform data. In an information processing apparatus including at least a data input / output device, a data recording device, and an arithmetic processing device,
Read a document that is a text string that is composed of a control code and a local sentence.
Read the relationship between the control code, the effect that is the processing content of the local sentence controlled by the control code, and the number of the effect,
Separating the document into control codes and local sentences, and getting each control code and a pair of local sentences controlled by the control code,
Generate waveform data that reads the local sentence, and process the waveform data that reads the local sentence according to the control code corresponding to the local sentence.
Synthesize the text-to-speech reading waveform data and the waveform processed waveform data,
A document processing method for reading, characterized by outputting the obtained waveform data. The reading document processing method according to claim 2, wherein when there are a plurality of effects, waveform processing according to the plurality of control codes is performed on the local sentence reading waveform data. The document processing method for reading out according to claim 2 or 3, wherein the reading data of the ground sentence reading waveform data is checked in advance, and if there is no ground sentence to be read out, the output of the silent part is deleted. 5. The reading document processing method according to claim 2, wherein the processing content of the local sentence controlled by the control code is a change in volume. 5. The document processing method for reading out according to claim 2, wherein the processing content of the local sentence controlled by the control code is a change in the amplitude of the specific frequency. 5. The reading document processing method according to claim 2, wherein the processing content of the local sentence controlled by the control code is a waveform deformation. 5. The reading document processing method according to claim 2, wherein the processing content of the local sentence controlled by the control code is a change in reproduction speed. 5. The reading document processing method according to claim 2, wherein the processing content of the local sentence controlled by the control code is a temporal change in reproduction timing. 5. The document processing method for reading aloud according to claim 2, wherein the processing content of the local sentence controlled by the control code is a pitch shift for changing a frequency. The document processing method for reading out according to claim 10, wherein the pitch shift is a change in accordance with a classical rhythm. The document processing method for reading according to claim 10, wherein the pitch shift is a change in accordance with an equal temperament obtained by dividing a pitch by an equal frequency ratio. The document processing method for reading out according to claim 11 or 12, wherein the pitch shift is a change that reduces or doubles the frequency. The document processing method for reading out according to claim 11 or 12, wherein the pitch shift is a turn to change a sequence of chord constituent sounds.

Images