JP2012103654A - Voice synthesizer and program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a technology capable of easily assigning a note sequence to a character sequence.SOLUTION: A user touches a touch screen 31 to enter the character sequence into a text box 101. A voice synthesizer 100 displays the entered character sequence on the touch screen 31 as an entered character image 104. When the user traces the touch screen 31, a pitch curve 103 is displayed and the entered character sequence is pronounced from a speaker 111 in accordance with a shape of the pitch curve 103.

Description

  The present invention relates to a speech synthesizer and a program.

  When multiple pitches and pitches (hereinafter referred to as “note strings”) are input as data, continuous pitch changes and vibrato are applied to the input note strings in response to mouse and keyboard operations. Techniques for reflecting such acoustic effects are known (for example, Patent Documents 1 to 4).

Japanese Patent Laid-Open No. 10-143155 Japanese Patent No. 3781167 Japanese Patent No. 3620405 JP 2002-372972 A

By the way, the song is a character string that is associated with the character string of the lyrics and the above-described note string. However, the techniques described in Patent Documents 1 to 4 are intended to process only the note string. This is not a technology that takes into account the relationship between the character string and the note string.
It is an object of the present invention to provide a technique that can easily assign a note string to a character string.

  The present invention provides a character string acquisition unit that acquires a character string composed of a plurality of characters, a character string display unit that displays on the display unit each character that constitutes the acquired character string, and a first that represents time. When a figure in a coordinate system having an axis and a second axis representing a pitch is designated by the user, the figure is displayed on the display means in a state in which the figure is associated with each character constituting the character string. Means for assigning a pitch and a tone length to each character based on coordinate values of positions corresponding to each character constituting the displayed character string in the displayed graphic, and the character There is provided a speech synthesizer comprising speech synthesizer for synthesizing speech data that causes each character constituting a sequence to be pronounced at a pitch and a tone length assigned by the assigning means.

  In a preferred embodiment, a pronunciation length for storing a length of a pronunciation time of each character constituting the word or a ratio of a pronunciation time of each character constituting the word with respect to the pronunciation time when the word is pronounced for a plurality of words A dictionary storage means, wherein the assigning means is a sound length when the entire character string is pronounced and is designated by the user, and the pronunciation length dictionary for each character constituting the character string A sound length may be assigned to each character based on the length of the sound generation time or the ratio of the sound generation times stored in the storage means.

  In another preferred aspect, the display device includes indicator display means for displaying on the display means an indicator when a pitch or a tone length is assigned to each character constituting the character string, and the assigning means is displayed by the indicator display means. The coordinate value of the position corresponding to each character constituting the character string in the graphic is corrected according to the index, and the pitch and the sound length are assigned to each character based on the corrected coordinate value. Also good.

  In still another preferred embodiment, the graphic display means includes acoustic effect storage means for storing an acoustic effect applied when a character is pronounced in association with each of a plurality of graphic shapes, and the graphic display means includes the display means When a figure superimposed on the figure displayed on the screen is designated by the user, the superimposed figure is displayed on the display means, and the assigning means is stored in the acoustic effect storage means. Among the shapes of a plurality of figures, the shape of a figure whose similarity with the superimposed figure exceeds a threshold is specified, and the acoustic effect stored in association with the shape of the specified figure is You may make it allocate to the said character currently displayed on the position corresponding to the coordinate value of the superimposed figure.

  Further, the present invention provides a computer with a character string acquisition function for acquiring a character string composed of a plurality of characters, a character string display function for displaying each character constituting the acquired character string on a display means, When a graphic in a coordinate system having a first axis representing pitch and a second axis representing time is designated by the user, the display means is associated with each character constituting the character string. A graphic display function to be displayed on the screen, an assignment function for assigning a pitch and a pitch to each character based on the coordinate value of the graphic corresponding to each character constituting the displayed character string, and the character string There is provided a program for realizing a voice synthesis function for synthesizing voice data for causing each constituent character to be generated at a pitch and a tone length assigned by the assignment function.

  According to the present invention, it is possible to easily assign a note string to a character string.

The block diagram showing the hardware constitutions of the speech synthesizer concerning the embodiment of the present invention. A diagram showing the contents of the pronunciation dictionary DB A diagram showing the contents of the shortest pronunciation time DB The figure which shows the contents of voice DB The figure showing the contents of the sound effect DB Block diagram showing functional configuration of speech synthesizer The figure showing the appearance and display contents of the speech synthesizer Processing flow diagram of speech synthesizer Schematic diagram for explaining trajectory analysis processing and character spacing control processing Schematic diagram for explaining speech synthesis processing Schematic diagram for explaining speech synthesis processing Schematic diagram for explaining speech synthesis processing The figure which shows the contents of voice DB Schematic diagram for explaining the modification 10 Schematic diagram explaining the correction function for pitch Schematic diagram explaining the correction function for sound length The figure which shows the contents of initial value pronunciation dictionary DB The figure showing the display content of the speech synthesizer based on modification 13

Hereinafter, an embodiment of the present invention will be described.
<Embodiment>
<Configuration>
FIG. 1 is a block diagram showing a hardware configuration of a speech synthesizer 100 according to an embodiment of the present invention. The speech synthesizer 100 includes a control unit 10, a storage unit 20, a UI (User Interface) unit 30, and a speech output unit 40, which are connected via a bus. The voice synthesizer 100 is a device that synthesizes voice data based on a character string and voice information that is information related to pronunciation including a note string, and outputs a voice based on the synthesized voice data. In the present embodiment, the speech synthesizer 100 is a smartphone. The control unit 10 includes a CPU (Central Processing Unit), a ROM (Read Only Memory), a RAM (Random Access Memory), and the like. The CPU reads out a control program stored in the ROM or the storage unit 20, loads it into the RAM, and executes it to control each unit of the speech synthesizer 100 via the bus. The RAM functions as a work area when the CPU processes data.

  The storage unit 20 stores an application program for causing a computer to function as a speech synthesizer (hereinafter, this program is referred to as “speech synthesis application”). When the control unit 10 executes this speech synthesis application, the function shown in FIG. 6 to be described later is realized in the speech synthesis device 100. The storage unit 20 includes a pronunciation dictionary DB (Database) 21, a shortest pronunciation time DB 22, a voice DB 23, and an acoustic effect DB 24. The pronunciation dictionary DB 21 is composed of a plurality of pronunciation records, which are data used as a reference when a pronunciation time is assigned to a character string input by a user. The shortest pronunciation time DB 22 is obtained by assigning, for each character such as “A” and “I”, a minimum length of time required when the character is pronounced, from a plurality of shortest pronunciation time records. Become.

  The voice DB 23 is composed of a plurality of voice records, which are data related to voice synthesized according to the content input by the user. In this voice record, a note string and a sound effect are associated with a character string (referred to as an input character string) input by a user for each character constituting the character string. This voice record is generated according to the input content by the user and registered in the voice DB 23. The acoustic effect DB 24 is composed of a plurality of acoustic effect records in which predetermined graphic shapes (referred to as acoustic effect graphics) and acoustic effect types are associated with each other. For example, an acoustic effect called vibrato is associated with a waveform figure having a peak value at an interval of a certain range.

  FIG. 2 is a diagram showing the contents of the pronunciation dictionary DB 21. Each pronunciation record included in the pronunciation dictionary DB 21 includes a plurality of items such as an identification ID, a character string, the number of characters, and a reference assigned sound length. The identification ID is an ID for uniquely identifying each pronunciation record, and is composed of, for example, a six-digit number. The character string is predetermined as a word to be pronounced. The number of characters is the number of characters constituting the character string. The reference assigned sound length is the length of the pronunciation time assigned in order from the first character of the character string for each character constituting the character string. In the reference assigned sound length, the sound length of each character is determined in advance based on the length of time taken when the corresponding character string is pronounced with natural inflection. For example, the pronunciation record with the identification ID “000001” is “0.2” for the characters “o”, “ha”, and “yo” for the character string consisting of “four” characters “good morning”. "Second" is pronounced, and "u" is pronounced "0.1 second". These numerical values of 0.2 seconds or 0.1 seconds mean a ratio of pronunciation time of each character constituting the character string, so that the pronunciation dictionary DB 21 has a value corresponding to the pronunciation time when the word is pronounced. It is an example of the pronunciation dictionary memory | storage means which memorize | stores the pronunciation time ratio of each character which comprises a word about several words.

  FIG. 3 is a diagram showing the contents of the shortest pronunciation time DB 22. Each shortest pronunciation time record included in the shortest pronunciation time DB 22 includes a plurality of items such as an identification ID, characters, and the shortest pronunciation time. The identification ID is an ID for uniquely identifying each shortest pronunciation time record, and is composed of, for example, a 4-digit number. For example, in the case of Hiragana, the characters are characters from “a” to “n”. Here, the characters are not limited to hiragana, but may be kanji, numbers, alphabets, or the like. The shortest pronunciation time is a minimum length of time required when a corresponding character is pronounced. For example, the shortest pronunciation time is determined in advance based on the shortest time that the character can be heard experimentally. Yes. For example, in FIG. 3, the letters “A” indicate that a pronunciation time of at least “0.05 seconds” is required. The shortest pronunciation time DB 22 is shortest pronunciation time storage means for storing the shortest sound length when a character is pronounced for a plurality of characters. Depending on the character (for example, “A”, “I”, “U”, “E” and “O” which are vowels and “N” which is a repellent sound), the shortest pronunciation time record is registered in the shortest pronunciation time DB 22. It does not have to be.

  FIG. 4 is a diagram showing the contents of the voice DB 23. The voice record included in the voice DB 23 includes a plurality of items such as a character order ID, input characters, pitches, pitches, and sound effects. The character order ID is an ID for uniquely identifying each voice record and indicating the order of the input characters, and is composed of, for example, a four-digit number. An input character is each character which comprises the character string input as a lyric by the user. The pitch is the pitch when a character is pronounced and is represented by a frequency. The sound length represents the time taken when a character is pronounced. The sound effect represents the type of sound effect applied to the character. In the following, when an acoustic effect is applied to a character, it may be said that the acoustic effect is “applied”. For example, when a character string corresponding to the voice record shown in FIG. 4 is pronounced, the pronunciation is in the order of “ko”, “n”, “ni”, “chi”, “ha” according to the character order ID. Made. In FIG. 4, for example, the character “ko” represents that the pronunciation is made with “vibrato” applied for “0.3 seconds” at a pitch of “406 Hz”. In FIG. 4, although only shown audio record of the character string "Hello", in fact, voice record for each character that is included in all of the character string other than the "Hello" is this voice DB23 Included.

FIG. 5 is a diagram showing the contents of the acoustic effect DB 24.
The sound effect record included in the sound effect DB 24 includes a plurality of items such as an identification ID, a sound effect graphic, and a sound effect. The identification ID is a number that uniquely identifies each sound effect record, and includes, for example, a three-digit number. The sound effect graphic is data representing the shape of the graphic, and the shape is different for each sound effect record. The acoustic effect is a type of acoustic effect that can be applied to characters during pronunciation. For example, as shown in FIG. 5, “Vibrato” is included in the sound effect record whose identification ID is “001” and the sound effect graphic is a wave-shaped graphic that takes a peak value at a certain range interval. Are associated with each other. The acoustic effect DB 24 is an acoustic effect storage unit that associates and stores acoustic effects applied when a character is pronounced in association with each of a plurality of graphic shapes.

  Returning again to FIG. The UI unit 30 includes buttons (not shown) and a touch screen 31. When the user operates the button or the touch screen 31, the UI unit 30 supplies a signal corresponding to the operation to the control unit 10. The control unit 10 controls the entire speech synthesizer 100 based on the received signal. The touch screen 31 has a structure in which a light transmissive touch sensor is stacked on a screen of a display device. The user gives instructions to the speech synthesizer 100 by performing an operation such as touching the touch screen 31 or tracing the touch screen 31 with a finger while viewing the image displayed on the display device. Enter.

  The audio output unit 40 includes a DAC (Digital Analog Converter), an amplifier, and a speaker. The audio output unit 40 converts the digital audio data supplied from the control unit 10 into analog audio data by the DAC, further amplifies it with an amplifier, and outputs audio corresponding to the amplified analog audio signal from the speaker. Is output.

  In this speech synthesizer 100, a user inputs a character string to be pronounced (that is, lyrics) via the touch screen 31, and further inputs speech information representing how to pronounce this character string as a figure. To do. This voice information represents a character string sound length, a pitch, a sound length, and a sound effect with respect to the input character string. The character string sound length is the time taken when the entire character string is pronounced, and corresponds to the total time of the sound lengths assigned to each character. The speech synthesizer 100 synthesizes speech data based on these character strings and speech information, and outputs speech based on the synthesized speech data.

  FIG. 6 is a block diagram illustrating a functional configuration of the speech synthesizer 100. The character string acquisition unit 11 acquires a character string composed of a plurality of characters input by the user via the touch screen 31, and stores it in the RAM. The reference sound length specifying means 12 searches the pronunciation dictionary DB 21 using the input character string stored in the RAM, and when the corresponding pronunciation record is specified, stores the specified pronunciation record in the RAM.

  The display control unit 13 controls the content displayed on the touch screen 31 in accordance with an operation performed by the user through the UI unit 30. For example, the display control unit 13 causes the touch screen 31 to display an image (referred to as an input character image) representing each character constituting the input character string for the input character string stored in the RAM. Thus, the display control means 13 functions as the character string display means 13A that displays each character constituting the acquired character string on the touch screen 31 that is a display means. In addition, the display control unit 13 causes the touch screen 31 to display a locus of the position of the fingertip when the user traces the touch screen 31 with a finger. As will be described later with reference to FIG. 7, the touch screen 31 includes a display area in which a coordinate system including a first axis representing time and a second axis representing pitch is set. The user traces the touch screen 31 with his / her finger. This trajectory corresponds to the graphic representing the above-described voice information, that is, the graphic representing the character string sound length, pitch, tone length, and acoustic effect for the input character string, and will be described later with reference to FIG. Are displayed on the touch screen 31 in association with the input character image. Hereinafter, this figure is referred to as a pitch curve. In this way, the display control means 13 constitutes an input character string with a figure (pitch curve 103) in a coordinate system having a first axis representing time and a second axis representing pitch specified by the user. The graphic display means 13B is displayed on the touch screen 31 in a state associated with the input character image 104 representing each character.

  The character spacing control means 14 controls the spacing between the input character images 104 in accordance with an operation of dragging the input character image displayed on the touch screen 31 and changing the display position. Is stored in the RAM and is input to the display control means 13. Here, the input character image 104 is an image of each character displayed in accordance with the content input by the user in the text box 101 (see FIG. 7). Further, the dragging here means that the input character image 104 is selected by touching it with the fingertip on the touch screen 31, and the fingertip is moved as it is. The character spacing control content includes a numerical value representing the distance between the input character images 104 displayed on the touch screen 31. This number always takes a value of 0 or more. When the numerical value (referred to as character spacing value) representing the distance between the input character images at adjacent positions included in the inputted character spacing control content is 0, the display control means 13 performs these inputs on the touch screen 31. The character image 104 is displayed in a combined state. Here, “combined” means a state in which a rectangle surrounding one input character image 104 and a rectangle surrounding the input character image 104 adjacent to this input character image 104 are in contact (see FIG. 7).

  If the character spacing value is not 0, the display control means 13 displays the input character image 104 in a separated state with a distance corresponding to the character spacing value. Here, “separated” means a state in which a rectangle surrounding one input character image 104 is separated from a rectangle surrounding the input character image 104 adjacent to the input character image 104. That is, the display control unit 13 functioning as the character string display unit 13A changes the position of each character displayed on the touch screen 31 in accordance with a user instruction, and displays each character (see FIG. 7). The trajectory analyzing means 15 analyzes the relationship between the above-described pitch curve and each input character image 104, and calculates the character string sound length, pitch, sound length, and acoustic effect for the input character string as analysis results. Store the result in RAM. At this time, the trajectory analyzing means 15 calculates the pitch and length of each character based on the coordinate value of the figure (pitch curve 103) corresponding to each input character image 104 in the coordinate system on the touch screen 31.

  The voice record generation means 16 uses the input character string, pronunciation record, character interval control content, pitch curve analysis result by the trajectory analysis means 15 and the content of the shortest pronunciation time DB 22 stored in the RAM as input parameters. Performs processing to generate audio records. At this time, the sound record generating means 16 assigns a pitch, a sound length, and a sound effect to each character constituting the input character string. The voice record generating unit 16 stores the generated voice record in the voice DB 23. When the voice synthesizing unit 17 receives the notification that the processing is completed from the voice record generating unit 16, the voice synthesizing unit 17 synthesizes voice data based on the voice record stored in the voice DB 23, and based on the voice data from the voice output unit 40. Output audio.

  As described above, the reference sound length specifying unit 12, the character interval control unit 14, the trajectory analyzing unit 15, and the voice record generating unit 16 cooperate to correspond to each character constituting the displayed character string. Based on the coordinate value of the pitch curve which is a figure, it functions as an assigning means 18 for assigning a pitch and a tone length to each character.

<Operation>
Next, the operation of the speech synthesizer 100 will be described with reference to FIGS.
FIG. 7 is a diagram illustrating the appearance and display contents of the speech synthesizer 100, and FIG. 8 is a process flow diagram of the speech synthesizer 100.
As illustrated in FIG. 7, the speech synthesizer 100 includes a housing 110, a touch screen 31, and a speaker 111. The housing 110 is provided with a touch screen 31 and a speaker 111. On the touch screen 31, a text box 101, a pronunciation reference line 102, a pitch curve 103, an input character image 104, a play button image 105, a return button image 106, and an assigned character image 109 are displayed. In the region where the pitch curve 103 shown in FIG. 7 is displayed, the X axis (first axis) represents time, and time passes from the negative direction to the positive direction on the X axis. In the area where the pitch curve 103 shown in FIG. 7 is displayed, the Y axis (second axis) represents the pitch, and the higher the sound is in the positive direction, the lower the sound is in the negative direction. Become. In the present embodiment, it is assumed that a pitch of one octave is assigned between the minimum coordinate value and the maximum coordinate value on the Y axis.

  The text box 101 is an area where a character string is input. When the user touches an area corresponding to the text box 101 on the touch screen 31, a keyboard image is displayed on the touch screen 31. The user inputs a character string into the text box 101 by touching the keyboard image. The pronunciation reference line 102 is a straight line representing the reference of the pitch when the input character string is pronounced. For example, when the user touches the touch screen 31 so as to trace the pronunciation reference line 102, the input character string is pronounced according to a predetermined pitch (eg, 440 Hz) with respect to the pronunciation reference line 102. As described above, the pitch curve 103 determines the character string tone length, pitch, tone length, and acoustic effect for the input character string. A figure drawn on the touch screen 31 by the user is displayed on the touch screen 31 as the pitch curve 103.

  The input character image 104 is as described above. Further, as described above, when the user drags the input character image 104 on the touch screen 31, the display position of each input character image 104 can be changed, and these can be combined and separated. When two adjacent input character images 104 are combined, the sound length of the previously input character among the two characters corresponding to these input character images 104 is the shortest pronunciation time. On the other hand, when two adjacent input character images 104 are separated, the sound length of the previously input character among the two characters corresponding to these input character images 104 is longer than that at the time of the above combination. That is, the pronunciation is made with the sound extended. The input character image 104 is displayed in a state where the display position is directly below the pitch curve 103 (Y-axis negative direction) and along the horizontal width (length in the X-axis direction) of the pitch curve 103. Here, when the user newly redraws the pitch curve 103 without changing the input character string, the display control means 13 newly maintains the ratio of the intervals between the input character images 104 on the X axis. The input character image 104 is displayed by changing the display position along the horizontal width of the pitch curve 103.

  When the user touches the play button image 105, the input character string is pronounced from the speaker 111 according to the input pitch curve 103. When the user touches the return button image 106, the input character library is displayed in a tree shape on the touch screen 31. The input character library includes a plurality of combinations of a character string input by the user and the pitch curve 103 input to the character string. When the touch of a button image 106 the user returns, for example, "Hello", "Good evening", a plurality of character strings such as "Good night" is displayed in a tree-like. When the user selects any one of the displayed character strings via the touch screen 31, a screen as shown in FIG. 7 is displayed on the touch screen 31 for the selected character string.

  The assigned character image 109 is a reflection of the height in the Y-axis direction representing the pitch of the input character image 104. As shown in FIG. 7, each of the assigned character images 109 has the same display position in the X-axis direction as the corresponding input character image 104, and the display position in the Y-axis direction has a pitch curve. 103 and immediately below the intersection of the perpendicular line extending in the positive direction of the Y axis with respect to the pronunciation reference line 102 from the center of the rectangle surrounding each input character image 104. Here, when the user newly redraws the pitch curve 103 without changing the input character string, the display control means 13 newly maintains the ratio of the intervals between the assigned character images 109 on the X axis. The display position in the X-axis direction of each assigned character image 109 is changed so as to follow the horizontal width (length in the X-axis direction) of the pitch curve 103, and the new pitch curve 103 is adapted to the height in the Y-axis direction. Thus, the display position of each assigned character image 109 in the Y-axis direction is changed and displayed.

  In FIG. 8, when the user inputs a character string into the text box 101 (step S1; YES), the character string acquisition unit 11 stores the input character string in the RAM (step S2). For example, as represented in FIG. 7, when the user inputs a character string "Hello" in the text box 101, a character string obtaining means 11 stores the input character string "Hello" to the RAM. Next, the reference sound length specifying means 12 searches the pronunciation dictionary DB 21 using the input character string, specifies the corresponding pronunciation record, and stores the specified pronunciation record in the RAM (step S3). Here, the reference tone duration specifying means 12 uses the string "Hello", the result of searching the pronunciation dictionary DB21 represented in Figure 2, the identification ID is to store the sound record "000002" to the RAM.

  Next, the display control means 13 (here "Hello") above the input string based on, and displays an input character image 104 on the touch screen 31 (step S4). At this time, the display control means 13 displays the input character string in a manner based on the reference assigned sound length in the pronunciation record stored in the RAM. Specifically, as represented in FIG. 2, for the character string "Hello", "this", "the", "Chi", and the respective character "ha", "0.2 The pronunciation time of "second" is assigned, and the pronunciation time of "0.1 second" is assigned to the character "n". Here, as described above, these numerical values of 0.2 seconds or 0.1 seconds mean the ratio of the pronunciation time of each character constituting the character string, so the display control means 13 are adjacent to each other. The input character image 104 is displayed on the touch screen 31 by being separated by a distance corresponding to the ratio of the pronunciation time. As a result, the display mode of the input character image 104 is as shown in FIG. If the character string corresponding to the input character string does not exist in the pronunciation dictionary DB 21 and the pronunciation record related to the input character string is not stored in the RAM, the display control means 13 displays the characters constituting the input character string. The displayed input character image 104 is displayed at equal intervals.

  Next, when the display position of the input character image 104 is changed by the user's operation, the character interval control unit 14 controls the interval of each input character image 104 and stores the character interval control content in the RAM. The character spacing control content is input to the display control means 13 to control the display mode of the input character image 104 (step S5).

FIG. 9 is a schematic diagram for explaining the character spacing control process.
Details of step S5 will be described with reference to FIG. When the user drags the input character image “n” whose rectangle is represented by a broken line toward the position where the rectangle is represented by a solid line, the following processing is performed. When the distance β between the input character image 104 “n” and the input character image 104 “ni” is calculated as the character interval value, the character interval control means 14 stores the calculated character interval value in the RAM and displays it. Input to the control means 13. The display control means 13 changes the display position of the input character image 104 “n” based on the input character interval value. That is, the display control means 13 changes the display position of the input character image 104 “n” from the position where the rectangle is represented by a broken line to the position where the rectangle is represented by a solid line. As a result, in FIG. 9, as a result of the display position of the input character image 104 being changed, the input character images 104 “ko” and “n” are combined, and “ni”, “chi”, and “ha” are combined. The input character image 104 is separated.

  When the sound record generation means 16 assigns a sound length to the input character “n”, the sound length corresponding to the distance γ before the change is calculated by calculating the sound length corresponding to the distance β. In comparison with, a longer sound length is assigned. As a result, when the input character “n” is pronounced, it is pronounced as an extended sound, such as “n-oo”. Also, since the length of the entire character string is defined as the character string sound length, if the sound length calculated for the input character “n” is changed to a longer length, the input character “ko” is correspondingly changed. On the other hand, the calculated sound length is shorter than before being combined. Here, the sound length assigned to the input character “ko” is never shorter than the length of time stored as the shortest pronunciation time corresponding to the character “ko” in the shortest pronunciation time DB 22. That is, the sound record generating means 16 assigns a sound length equal to or longer than the shortest sound length stored in the shortest pronunciation time DB 22 which is the shortest pronunciation time storage means to each character constituting the character string.

  After step S5, the user does not touch the play button image 105 (step S6; NO), and the user does not draw a figure on the touch screen 31, that is, the pitch curve 103 is not input. In the case (step S7; NO), the process returns to step S4, and the above process is repeated.

  On the other hand, after step S5, the user does not touch the play button image 105 (step S6; NO), and the user draws a figure on the touch screen 31, that is, the pitch curve 103 is input. In the case (step S7; YES), the trajectory analyzing means 15 analyzes the input pitch curve 103 (step S9). Specifically, in step S9, the trajectory analyzing means 15 analyzes the relationship between the input pitch curve 103 and each input character image, and constructs a character string sound length for the input character string and this input character string. The pitch, length, and sound effect for each character are specified and stored in the RAM.

  The process of step S9 will be described in further detail. First, the trajectory analysis means 15 determines the entire character string according to the moving speed of the user's fingertip when the pitch curve 103 is input (that is, the time required for input from the beginning to the end of the pitch curve 103). A character string sound length that is a sound length to be assigned at the time of pronunciation is calculated. If the user draws a figure on the touch screen 31 at a high speed, the character string sound length is short. If the user draws a locus at a low speed, the character string sound length is long. . For example, when the trajectory analyzing means 15 analyzes the character string sound length as 3 seconds from the drawing speed of the pitch curve 103 and assigns it to the input character string, the entire character string is pronounced with a length of 3 seconds. . That is, in the example of FIG. 7, so that the character string "Hello" is pronounced by the length of 3 seconds.

  Next, the trajectory analyzing means 15 obtains the pitch of each character of the input character string. Specifically, first, the trajectory analyzing means 15 is a perpendicular line (input) extending in the positive direction of the Y axis with respect to the pronunciation reference line 102 from the center of the rectangle surrounding each input character image 104 in the area where the pitch curve 103 is input. A character line) is virtually drawn. The trajectory analyzing means 15 then uses the pitch assigned to the pronunciation reference line 102 as a reference, and the sound of each character constituting the input character string according to the Y coordinate value of the intersection of the pitch curve 103 and the input character line. Calculate the high. That is, if the above-described Y coordinate value of the intersection is larger than the Y coordinate value of the pronunciation reference line 102, the pitch of the input character at the intersection is higher than that assigned to the pronunciation reference line 102. On the other hand, if the Y coordinate value of the intersection is smaller than the Y coordinate value of the pronunciation reference line 102, the pitch of the input character at the intersection is lower than that assigned to the pronunciation reference line 102.

  Next, the trajectory analyzing means 15 obtains the sound length of each character of the input character string. Specifically, the trajectory analyzing means 15 normalizes the reference assigned sound length in the pronunciation record recorded in the RAM, assuming that the time length in the character string sound length is 1, and thereby the sound length assigned to each character. Calculate the ratio.

  Here, the details of step S9 will be described with reference to FIG. In FIG. 9, the touch screen 31 is shown in an enlarged manner, and a part of the display content is omitted for convenience of explanation. In FIG. 9, an input character line 107 is shown. However, this is not actually displayed on the touch screen 31. Intersections A, B, C, D, and E are intersections where each input character line 107 and pitch curve 103 intersect, and hold X coordinate values and Y coordinate values, respectively. For example, when the trajectory analyzing unit 15 calculates the pitch of the input character “KO”, the trajectory analyzing unit 15 determines the pronunciation reference because the Y coordinate value of the intersection A is smaller than the Y coordinate value of the pronunciation reference line 102. The difference length α is obtained by subtracting the Y coordinate value of the intersection A from the Y coordinate value of the line 102. The trajectory analyzing means 15 inputs a pitch that is lower than the pitch (for example, 440 Hz in this case) on the pronunciation reference line 102 by a pitch corresponding to the difference sentence length α with reference to the Y coordinate value of the pronunciation reference line 102. Calculate for the character “ko”.

  On the other hand, for example, when the trajectory analyzing unit 15 calculates the pitch of the input character “ni”, the Y coordinate value at the intersection C is larger than the Y coordinate value of the pronunciation reference line 102, so the trajectory analyzing unit 15 The difference length α ′ is obtained by subtracting the Y coordinate value of the pronunciation reference line 102 from the Y coordinate value of the intersection C. Then, the trajectory analyzing means 15 uses the Y coordinate value of the pronunciation reference line 102 as a reference and outputs a pitch higher than the pitch on the pronunciation reference line 102 by a pitch corresponding to the difference sentence length α ′ to the input character “ni”. To calculate.

  Next, as described above, the sound length of each character constituting the character string is determined according to the character string sound length and the ratio of the sound length of each character based on the contents of the pronunciation dictionary DB 21. For example, returning to the example of FIG. 7, an input character string is "Hello", a character string tone length is "3 seconds", the sound length of each character, as shown in FIG. 2, the character "ko "," "Ni", "chi" and "ha" are "0.2 seconds", and the character "n" is "0.1 seconds". Accordingly, the trajectory analyzing means 15 normalizes the character string sound length “3 seconds” as a reference, so that the characters “ko”, “ni”, “chi” and “ha” are “0.67 seconds”. Is assigned the length of "0.33 seconds" for the character "n" (rounded to the second decimal place). Here, the rounding position may be changed as appropriate in the design. Thus, the trajectory analyzing means 15 calculates the character string sound length, which is the sound length when the entire character string is pronounced, based on the time required for the input from the beginning to the end of the figure. Based on the character string sound length and the ratio stored in the pronunciation dictionary storage means for each character constituting the character string, a sound length is assigned to each character.

  Here, assuming that the display position of the input character image 104 “n” has been changed by the user as shown in FIGS. 7 to 9, the trajectory analyzing means 15 will input the input character image 104 “after the display position has been changed. X coordinate value at the intersection B for "" is acquired. Further, the trajectory analyzing means 15 acquires the X coordinate value at the intersection C for the input character image 104 “ni” adjacent to the input character image 104 “n” in the positive X-axis direction. Then, the trajectory analysis unit 15 calculates the difference between the X coordinate values at the intersection B and the intersection C by subtracting the X coordinate value corresponding to the intersection C from the X coordinate value corresponding to the intersection B. The trajectory analyzing means 15 assigns a longer sound length to the input character “n” than before the display position is changed based on the calculated difference of the X coordinate values. As a result, as described above, the input character “n” is pronounced as an extended sound. On the other hand, as long as the pitch curve 103 is not changed, the character string sound length maintains the character string sound length based on the pitch curve 103 that has already been input. The sound length of “ko” is changed to be shorter. Here, the sound length assigned to the input character “ko” is shorter by the length of the input character “n” that is longer than before the change. Here, at least the shortest pronunciation time corresponding to the character “ko” stored in the shortest pronunciation time DB 22 is assigned to the input character “ko”.

  Further, when another figure is drawn on the already drawn pitch curve 103 at a position superimposed on the figure (pitch curve) already drawn by the user, the trajectory analyzing means 15 stores the shapes of the plurality of figures. From the acoustic effect DB 24, an acoustic effect graphic having a similarity equal to or higher than a predetermined threshold is specified for the graphic drawn so as to be superimposed. Then, the trajectory analyzing unit 15 stores the type of the acoustic effect associated with the identified acoustic effect graphic as the type of acoustic effect for the character displayed at the position corresponding to the Y coordinate value of the superimposed graphic. To remember. Here, a known method may be used as a method of obtaining the similarity between figures. At this time, the display control means 13 displays on the touch screen 31 a graphic drawn by the user so as to be superimposed on the input pitch curve 103.

  After step S9, the display control unit 13 functioning as the graphic display unit 13B displays the pitch curve 103 on the touch screen 31 in a state of being associated with the input character image 104 (step S10). Next, the voice record generating means 16 constructs the input character string based on the input character string, the pronunciation record, the character interval control content, the analysis result of the pitch curve 103 and the shortest pronunciation time DB 22 stored in the RAM. A sound record is generated by assigning a pitch, a sound length, and a sound effect to each character to be performed, and the generated sound record is registered in the sound DB 23 (step S11). At this time, the voice record generating means 16 assigns a sound length to each character so that the sound length in the entire character string is the same as the character string sound length based on the ratio of the sound length assigned to each character. Here, as described above, the sound length assigned to each character does not become shorter than the length of time stored as the shortest pronunciation time of the corresponding character in the shortest pronunciation time DB 22. When the calculated character string sound length is less than the total time compared to the total time when the shortest pronunciation time is assigned to all the characters constituting the input character string, the voice record generating means 16 accumulates the shortest pronunciation time of each character from the top in accordance with the input order of the characters. Then, when the result of this integration exceeds the calculated character string sound length, the voice record generation means 16 does not set the subsequent characters as sounding targets and assigns no sound length.

  Then, the voice synthesizer 17 synthesizes voice data based on the contents of the voice record stored in the voice DB 23 (step S12). That is, the voice synthesizing unit 17 synthesizes voice data that causes each character constituting the character string to be pronounced with the pitch and the tone length assigned by the assigning unit 18. When synthesizing voice data, the voice synthesizing unit 17 performs a process of connecting a pitch assigned to a certain character and a pitch assigned to a character inputted next to the character by pitch bend. The voice synthesizer 17 synthesizes voice data for the corresponding character in a state in which the assigned acoustic effect is reflected. The audio output unit 40 outputs audio based on the audio data (step S13).

  On the other hand, when the user touches the play button image 105 after step S5 (step S6; YES), that is, when the pitch curve 103 is not input (steps S9 to S11 are not passed), the voice record generation means 16 Registers the input character string stored in the RAM in the voice DB 23. Then, the sound record generating means 16 updates the item “sound length” in the sound DB 23 according to the reference assigned sound length in the pronunciation record stored in the RAM, and at the same pitch as the sound generation reference line 102, the item in the sound DB 23. “Pitch” is updated (step S8). For example, since in the example of FIG. 7 the input string is "Hello", the sound record generating unit 16, the item "tone length" in the voice DB 23, in accordance with the contents shown in FIG. 2, the character "ko", "in "," "Chi" and "ha" are updated with numerical values "0.2 seconds", and the character "n" is updated with numerical values "0.1 seconds". Further, for example, in the example of FIG. 7, since the pitch of the pronunciation reference line 102 is “440 Hz”, the voice record generation unit 16 performs “ko”, “n”, “ni”, “chi”, and “ha”. The item “pitch” in “” is updated at “440 Hz”. Thereafter, the process proceeds to step S12.

10 to 12 are schematic diagrams for explaining the speech synthesis process.
Note that the speech synthesizer 100 has a “drawing mode” that is a mode for setting processing when the pitch curve 103 is drawn. “Drawing mode” includes “overwrite drawing mode” and “continuous drawing mode” set in the initial setting. In the “overwrite drawing mode”, when the pitch curve 103 is already drawn on the touch panel 31 and displayed, when the user newly draws the pitch curve 103, the already drawn pitch curve 103 is erased and newly added. This is a mode in which the pitch curve 103 is displayed according to the trace when drawn. On the other hand, in the “continuous drawing mode”, the pitch curve 103 is already drawn on the touch panel 31 and, when displayed, the user newly sets the pitch curve 103 in an area of the touch panel 31 that does not overlap with the already drawn pitch curve 103. In this mode, the pitch curve 103 that has already been drawn is not changed, and the pitch curve 103 is additionally displayed in accordance with the newly drawn locus. This drawing mode can be appropriately changed by the user through the UI unit 30. The drawing mode is not limited to “overwrite drawing mode” and “continuous drawing mode”, and other setting contents may be selectable.

  10 to 12, input character images 104 of “O”, “Ha”, “Yo”, and “U” are displayed. 10 and 11 show examples when the drawing mode is the “continuous drawing mode”. FIG. 10 shows a case where all the input character images 104 are combined in the input character image group 108A. In FIG. 10, the voice “Good morning” is output according to the input pitch curve 103a. Since the pitch curve 103b displayed on the right side in FIG. 10 does not intersect the input character line, no processing is performed. Alternatively, such a pitch curve 103b may be settable by the user through the UI unit 30 so as to repeat the output of sound according to the input character string. When repeating the above-described voice output according to the input character string, the trajectory analyzing unit 15 stores the flag in the RAM in a state (for example, a value “1”) indicating that the input character string is repeated. The voice record generating means 16 refers to this flag, and when the flag takes a value (in this case, “1”) which means that the input character string is repeated, the same contents as the registered voice record are Register additional to DB23. Alternatively, the voice record generating means 16 additionally registers a predetermined symbol (for example, “*”) indicating that the input character string is repeatedly generated for the item “input character” in the voice DB 23. Thus, the voice synthesizer 17 may synthesize voice data so that the input character string is repeatedly pronounced.

  In FIG. 11, an input character image group 108B in which “o” and “ha” are combined and an input character image group 108C in which “yo” and “u” are combined are separated. Further, the pitch curve 103c for the input character image group 108B and the pitch curve 103d for the input character image group 108C are not continuous, and the locus is interrupted. In such a case, when sound is output according to the input pitch curve 103, the sound of “Oha” and “ A silent period (referred to as a silent period) is generated between the voice and the voice (where □ means a silent period). Specifically, the trajectory analyzing means 15 calculates the length of the silent period based on the length in the X-axis direction at the location where the trajectory is interrupted. Then, the trajectory analyzing means 15 associates the calculated length of the silent period with two input characters sandwiching the silent period (that is, “ha” and “yo” here) and stores them in the RAM. The voice record generating means 16 generates a voice record corresponding to the silent period according to the contents stored in the RAM. FIG. 13 is a diagram showing the contents of the voice DB. Here, the voice record whose character order ID is “0003” is a voice record corresponding to the silent period. In order to distinguish audio records corresponding to the silent period, for example, a space may be allocated as an input character. In FIG. 13, the space is represented by “Δ” for convenience. When the voice data synthesized by the voice synthesizer 17 based on the voice record in FIG. 13 is output as voice, a silence period of 0.5 seconds occurs after “Oha” is pronounced. Will be pronounced.

  FIG. 12 shows an example when the drawing mode is “overwrite drawing mode”. In FIG. 12, an input character image group 108D in which “o” and “ha” are combined and an input character image group 108E in which “yo” and “u” are combined are separated. Further, the pitch curve 103e for the input character image group 108D and the pitch curve 103f for the input character image group 108E draw a continuous locus without interruption. Further, a wave-like figure to which the vibrato acoustic effect is applied is superimposed on the pitch curve 103e and the pitch curve 103f. As described above, when the user designates a graphic to be superimposed on the graphic (that is, the pitch curve 103) displayed on the touch screen 31 that is the display means, the display control unit 13 displays the graphic to be superimposed. It is displayed on the touch screen 31. The Y-coordinate value of this wave-shaped figure corresponds to the display positions of “O”, “Ha”, “Yo”, and “U”. Accordingly, in FIG. 12, according to the input pitch curve 103, the pronunciation for each character is pronounced in a state where the pronunciation for “ha” is extended as “Oh ~~~~~~~~”. Audio is output with vibrato applied.

  As described above, according to the speech synthesizer 100, after a character string is input, the user touches the touch screen 31 so as to draw a trajectory, thereby easily assigning a note string to the character string. Is possible.

The above embodiment can be modified as follows. In addition, you may implement the following modifications suitably combining.
<Modification 1>
In the embodiment, a smartphone provided with the touch screen 31 is cited as an example of the speech synthesizer 100, but the present invention is not limited to this. The speech synthesizer 100 may not include the touch screen 31. For example, the speech synthesizer 100 may include a mouse, a keypad, or a pen tablet as the UI unit 30. The voice synthesizer 100 may be a PDA (Personal Digital Assistant), a portable game machine, a portable music player, or a PC (Personal Computer). When the speech synthesizer 100 is a PC, the result drawn by the user using the mouse for the content displayed on the display is recognized as the pitch curve 103 or recognized as the interval control for the input character image 104. Or

<Modification 2>
In the embodiment, vibrato is given as an example of the acoustic effect, but the embodiment is not limited to this. For example, as shown in FIG. 9, when a trajectory surrounded by a circle is superimposed on the input pitch curve 103, the corresponding character may be pronounced in a false setting. In addition to this, various acoustic effects may be assigned to the corresponding characters in accordance with how the pitch curve 103 is drawn or the shape of the figure drawn in an overlapping manner.

<Modification 3>
In the embodiment, the pitch curve 103 is input by the user. However, the present invention is not limited to this, and the storage unit 20 may store a plurality of pitch curves 103 having a specific shape as preset data. For example, when a pitch curve 103 corresponding to dialect inflections such as standard language, Kansai dialect, and Tohoku dialect is stored in the storage unit 20 as preset data, the user can select a specific pitch curve from the preset data through the UI unit 30. 103 may be designated. In short, it is only necessary that the user can specify a graphic in the coordinate system having the first axis representing the pitch and the second axis representing the time.

<Modification 4>
In the embodiment, a pitch of one octave is assigned between the minimum coordinate value and the maximum coordinate value on the Y axis, but this pitch is not limited to this. For example, the user may be able to change the pitch range assigned to the Y axis narrower or wider by setting via the UI unit 30. For example, consider a case where the user sets a pitch of 2 octaves between the minimum coordinate value and the maximum coordinate value on the Y axis. At this time, it is assumed that the pitch of the pronunciation reference line 102 is “261 Hz”. In this case, a virtual pronunciation reference line having a pitch of “523 Hz” that is one octave higher than “261 Hz” exists in the positive Y-axis direction of the pronunciation reference line 102 with the pitch of the pronunciation reference line 102 as the center. . Further, in the negative Y-axis direction of the sound generation reference line 102, there is a virtual sound generation reference line having a pitch of “130 Hz” that is one octave lower than “261 Hz”. In the embodiment, the trajectory analyzing means 15 always calculates the pitch for a certain input character from the difference length between the Y coordinate value of the pronunciation reference line 102 and the Y coordinate value of the intersection. However, as described above, when there is a virtual pronunciation reference line other than the pronunciation reference line 102 displayed on the touch panel 31, the Y coordinate value of the pronunciation reference line closest to the intersection in the Y-axis direction is used as a reference. The pitch of the input character may be calculated.

<Modification 5>
In the embodiment, the user can perform a drag operation on the input character image 104 so that the input character images 104 can be combined and separated. It is not limited. For example, when the user touches and drags the right side or the left side of a certain input character image 104, the length that the input character image 104 is displayed in the X-axis direction (width of the input character image 104) May be changeable. In this case, the voice record generating means 16 assigns a longer sound length to the corresponding character as the width is longer and a shorter sound length as the width is shorter, according to the changed width of the input character image 104.

<Modification 6>
In the embodiment, the voice synthesizing unit 17 performs a process of connecting the pitch assigned to a certain character and the pitch assigned to the character input next to this character by pitch bend. It is not limited to this. For example, the voice synthesizing unit may synthesize voice data according to only the pitch assigned to each character stored in the voice DB 23 without performing pitch bend.

<Modification 7>
In the embodiment, when the calculated character string sound length is less than the total time compared to the total time when the shortest pronunciation time is assigned to all the characters constituting the input character string, the trajectory The analysis means 15 integrated the shortest pronunciation time in each character from the top in accordance with the input order. The trajectory analyzing means 15 performs control so that subsequent characters are not subjected to pronunciation when the result of the integration exceeds the calculated character string sound length. Not only this but the locus | trajectory analysis means 15 restrict | limits the length of the locus | trajectory in the pitch curve 103 displayed on the touch screen 31 beforehand, when the pitch curve 103 is input at the speed which does not satisfy the above-mentioned total time. It may be.

<Modification 8>
Further, a keyboard image may be displayed on the left side when the touch panel 31 is viewed from the front so that the user can intuitively understand the pitch.

<Modification 9>
In the embodiment, when the user draws another figure on the pitch curve 103 that has already been input, the acoustic effect corresponding to the other figure is applied during pronunciation. This may be as follows. The speech synthesizer 100 stores in the storage unit 20 an “additional input mode” that represents a processing mode when the user draws another graphic on the input pitch curve 103. The “additional input mode” includes a “sound effect mode” and a “note string change mode”. The “acoustic effect mode” is as described in the embodiment, and the above-described acoustic effect according to another figure drawn in an overlapping manner is applied when the corresponding input character is pronounced. In the “musical note change mode”, when the user touches and drags a specific part of the input pitch curve 103, the display control unit 13 displays the corresponding part of the pitch curve 103 according to the content of the drag. Change the aspect. For example, when the user touches a specific part of the pitch curve 103 and drags in the positive direction of the Y axis, the coordinate value of the corresponding part in the pitch curve 103 moves in the positive direction of the Y axis and the corresponding value in the pitch curve 103. The display mode for the periphery of the point is displayed so as to draw a curve in the positive direction of the Y-axis with this movement. When the user touches a specific part of the pitch curve 103 and drags in the negative Y-axis direction, the coordinate value of the corresponding part in the pitch curve 103 moves in the negative Y-axis direction and The display mode for the periphery of the location is displayed so as to draw a curve in the negative Y-axis direction with this movement. The user can appropriately change the “additional input mode” through the UI unit 30. Then, the trajectory analyzing means 15 calculates the pitch and the sound length assigned to the input character corresponding to the coordinate value according to the changed coordinate value of the corresponding part, and the voice record generating means 16 calculates the calculated result. The pitch and pitch are assigned to the corresponding input characters based on the above.

<Modification 10>
In the embodiment, the pitch and the length of each character are calculated according to the coordinate value at which the pitch curve 103 and the input character line 107 intersect, but the sound information at the time of pronunciation is not limited to this. . For example, in a place where the shape of the pitch curve 103 is flat with respect to the X axis, more characters are to be pronounced, and in a place where the shape of the pitch curve 103 is steep with respect to the X axis, The pronunciation start position of each character may be calculated so that fewer characters are to be pronounced. Specifically, it is as follows.

FIG. 14 is a schematic diagram for explaining the tenth modification.
FIG. 14 is an enlarged view of a part of the touch panel 31. Taking the input character “ko” in FIG. 14 as an example, when calculating the inclination of the tangent L1a with respect to the pitch curve 103 at the intersection A, the trajectory analyzing means 15 further calculates the absolute value of this inclination and stores it in the RAM. Here, the input character image 104a “ko” whose rectangle is represented by a broken line, the input character line 107a represented by a broken line, and the intersection A are in a state before the display position is changed by the processing in the modification 9. Represents. The trajectory analyzing means 15 calculates the absolute value of the inclination described above for the other input characters “n”. Here, as the absolute value of the slope at the intersection is larger, the pitch curve 103 is steeper at the intersection, that is, the shape of the pitch curve is closer to a state orthogonal to the X axis. On the other hand, the smaller the absolute value of the slope, the flatter the pitch curve 103 at the intersection, that is, the shape of the pitch curve is almost parallel to the X axis.

  When the absolute value of the inclination exceeds a predetermined threshold value, the trajectory analyzing unit 15 uses the coordinate value of the corresponding input character image as a reference, and determines whether the tangent line is in the positive direction or the negative direction on the X axis. The most recent coordinate value at which the absolute value of the inclination is equal to or less than the above threshold is obtained, and the value taken by the X axis in the obtained coordinate value is calculated as the pronunciation start position of the corresponding input character. On the other hand, if the absolute value of the inclination does not exceed a predetermined threshold value, the trajectory analyzing means 15 uses the value taken by the X axis at the coordinate value at the intersection of the input character line 107 and the pitch curve 103 as the corresponding input. Calculated as the character's pronunciation start position. That is, there is a high possibility that the pronunciation start position of the input character is not assigned to a location where the absolute value of the tangential slope increases toward the threshold, that is, a location where the pitch curve 103 is steep. As a result, in the steep portion described above, fewer characters are targeted for pronunciation compared to the portion where the pitch curve 103 is flat.

  For example, in FIG. 14, since the absolute value of the slope of the tangent L1a at the intersection A exceeds the above-described threshold, the pronunciation start position of the input image character 104a “ko” has moved in the negative direction on the X axis (ie, Forward position on the time axis). Specifically, since the absolute value of the slope of the tangent L1a 'is equal to or less than the threshold value at the intersection A', the trajectory analyzing unit 15 moves the input character display image 104a to the display position of the input character image 104a '. When the pronunciation start position of the input character “n” is calculated to be the same as the value taken by the X axis in the coordinate value at the intersection B, the input character “ko” is on the time axis with respect to the input character “n”. Since it moves forward, the sound length assigned to the input character “ko” is longer than before the movement.

<Modification 11>
In the embodiment, the trajectory analyzing unit 15 calculates a character string sound length that is a sound length to be assigned at the time of pronunciation of the entire character string, according to the time required for the input from the start end to the end of the pitch curve 103. However, the method for calculating the character string sound length is not limited to this. The trajectory analyzing means 15 is based on the distance on the pitch curve 103 from the start end to the end of the pitch curve 103 and the magnitude of the difference between the X coordinate value at the start end and the end X coordinate value of the pitch curve 103. The character string sound length may be calculated. For example, the longer the distance on the pitch curve 103 from the start end to the end of the pitch curve 103, the longer the character string sound length, or the difference between the X coordinate value at the start end and the end X coordinate value of the pitch curve 103. The larger the is, the longer the character string sound length is. As described above, there are various methods for calculating the character string sound length. Regarding the calculation of the sound length of each character constituting the character string, the trajectory analyzing means 15 uses a figure (pitch curve) corresponding to each character. The pitch and length are calculated based on the coordinate values.

<Modification 12>
In the embodiment, when the speech synthesizer 17 synthesizes speech data, the pitch assigned to a certain character and the pitch assigned to the character input next to this character are connected by pitch bend. Although the processing has been performed, the present invention is not limited to this, and the assigning means 18 realizes what is called a correction function so that a pitch according to a predetermined scale determined by the correction function is assigned to each input character. It may be. The assigning means 18 may assign a sound length according to a predetermined length of time determined by the correction function to each input character. In other words, the correction function realized by the assigning means 18 includes a correction function for pitch and a correction function for pitch.

FIGS. 15A and 15B are schematic diagrams for explaining a correction function for pitch.
In FIG. 15A and FIG. 15B, the menu button image 112 is displayed on the touch screen 31. When the user touches the menu button image 112, the control unit 10 displays function options executable by the user (referred to as function options) on the touch screen 31 in a list format. When the user selects a desired function from the displayed function options, the control unit 10 executes the selected function. Here, the function options displayed on the touch screen 31 include “pitch correction” and “pitch correction” realized by the control unit 10. By setting “ON / OFF”, it is possible to select whether or not to realize these functions.

  FIG. 15 (a), is corrected for pitch represents a display state before that takes place, for the input character string "Hello", the pitch curve 103 slope illustrates a state in which the input ing. When the user touches the play button image 105 in this state, as described in the embodiment, voice data in which adjacent characters in the input character string are connected by pitch bend is synthesized. In the state of FIG. 15A, when the user sets “pitch correction” in the function option to “ON”, the assigning means 18 performs correction on the pitch, and as a result, as shown in FIG. Display state. In FIG. 15B, an image having a horizontal stripe pattern imitating a piano roll representing a pitch is displayed on the touch screen 31 as a background image by the display control means 13. Here, the black horizontal stripe image represents a black key, and the white horizontal stripe image represents a white key. In addition, one pitch is assigned to each horizontal stripe image by the assigning unit 18 over the entire width in the pitch direction (Y-axis direction) and the horizontal width in the time-axis direction (X-axis direction). Each of these horizontal stripe images serves as an index when the assigning means 18 assigns a pitch to each character constituting the input character string 104.

  At this time, in FIG. 15A, the assigning means 18 assigns the pitch assigned according to the position corresponding to each of the input character images 104 in the pitch curve 103 to the pitch closest to the pitch in the piano scale. Change to That is, the assigning means 18 corrects the coordinate value of the position corresponding to each character constituting the input character image 104 in the pitch curve 103 according to the horizontal stripe image corresponding to the index, and based on the corrected coordinate value, each character. Assign a pitch to. Accordingly, the display control means 13 controls the display position so that each of the assigned character images 109 is displayed at a position that overlaps the horizontal stripe image that is closest in the pitch direction. For example, in the example of FIG. 15B, if the horizontal stripe image of the white key displayed at the bottom on the Y axis is “C3”, the assigned character image 109 “ko” has a pitch of “D # 3”. The pitch of “G3” is assigned to the assigned character image 109 “NI”. The display control means 13 controls the display position of the assigned character image 109 as described above, and also changes the display mode of the pitch curve 103. Specifically, as shown in FIG. 15B, the display control unit 13 displays the pitch curve 103 in a stepped manner according to the positions of the assigned character images 109 and the horizontal stripe images. That is, the display control means 13 changes the coordinate value of the position corresponding to each assigned character image 109 in the pitch curve 103 designated by the user according to the horizontal stripe image corresponding to the index, and the pitch curve whose coordinate value has been changed. 103 is displayed. As a result, each character is pronounced for the duration of the assigned tone length at the assigned pitch.

  In the state of FIG. 15B, when the user designates a certain assigned character image 109 and moves it so as to drag it in the pitch direction (Y-axis direction) (so-called “drag” in the pitch direction), The control means 13 performs control to display the assigned character image 109 so as to overlap the horizontal stripe image closest to the drag end point in the pitch direction. With this control, the display control means 13 changes the shape of the pitch curve 103 in the pitch direction. As a result, a pitch corresponding to the horizontal stripe image is assigned to the assigned character image 109 by the assigning means 18. That is, the assigning means 18 corrects the coordinate value of the position corresponding to each character constituting the assigned character image 109 in the pitch curve 103 according to the horizontal stripe image corresponding to the index, and based on the corrected coordinate value, each character. Assign a pitch to.

  When the pitch curve 103 is input at a speed exceeding a predetermined threshold, the control unit 10 does not correct the pitch curve 103 in the range input at such speed. That is, in the above range, as in the embodiment, the pitch assigned to a certain character by the assigning means 18 and the pitch assigned to the character input next to this character are connected by pitch bend. Processing is performed. The above processing is the same when the user changes a specific portion of the input pitch curve 103 by dragging as described above in Modification 9 at a speed exceeding a predetermined threshold. is there.

  The display control means 13 and the assignment means 18 perform the above-described processing on the assigned character image 109 and the pitch curve 103 that are already displayed at the same time that “pitch correction” is set to “ON”. Alternatively, the above-described processing may be performed only for the assigned character image 109 and the pitch curve 103 that are input after “pitch correction” is set to “ON”. The timing for performing the above-described processing may be predetermined in the speech synthesizer 100 or may be changeable by the user via the touch screen 31.

FIGS. 16A and 16B are schematic diagrams for explaining a correction function for the sound length.
FIGS. 16A and 16B show a state in which the user has set “sound length correction” in the function options to “ON”. When “sound length correction” is set to “ON”, the display control means 13 displays the time axis scale 113 on the upper part of the touch screen 31 in the positive Y-axis direction. The time axis scale 113 represents the time that elapses in the positive direction of the X axis. In the example of FIG. 16, one scale on the time axis scale 113 represents 0.1 seconds, but a time of a predetermined length other than 0.1 seconds is associated with this one scale. Or a measure or a beat may be associated. When measures and beats are associated with each other, the user may be able to make settings such as “4 beats” and “3/4 beats” for the beats via the touch screen 31. The time axis scale 113 serves as an index when the assigning unit 18 assigns a sound length to each character constituting the input character string 104.

  In FIG. 16A, when the user designates the assigned character image 109 “Chi” and drags along the trajectory as shown in D1, according to the drag trajectory (referred to as drag trajectory), The display control means 13 changes the display position of the assigned character image 109 and displays it. In the drag locus D1, the assigned character image 109 “CHI” is dragged so as to approach the assigned character image 109 “NI”. Accordingly, the display control means 13 changes the display position of the assigned character image 109 “CHI” to the display position shown in FIG. At this time, the display position of the assigned character image 109 is restricted to a position corresponding to one unit of the time axis scale 113 (that is, one scale for each scale). That is, when the display position of the assigned character image 109 is between a certain time axis scale 113 and the adjacent time axis scale 113, it is changed to the position of the closer time axis scale 113. Along with the change of the display position, the assigning unit 18 shortens the sound length assigned to the assigned character image 109 “NI” and lengthens the sound length assigned to the assigned character image 109 “CHI”. That is, the assigning means 18 corrects the coordinate value of the position corresponding to each character constituting the assigned character image 109 in the pitch curve 103 according to the time axis scale 113 corresponding to the index, and based on the corrected coordinate value, Assign a note length to each character. Accordingly, the display control means 13 changes the display position of the input character image 104 and displays it. When the user specifies and drags the left end (or right end) of the rectangle representing the input character image 104, the sound length assigned to the input character image 104 is changed by the assigning means 18. For example, in FIG. 16B, when the user performs dragging along the trajectory represented by the drag trajectory D2, the assigning means 18 shortens the sound length assigned to the input character image 104 “ko” and “n”. The sound length assigned to the input character image 104 is increased. Accordingly, the display control means 13 changes the display position of the assigned character image 109 and displays it.

  For convenience, “pitch correction” and “pitch correction” have been described as separate functions using different drawings. However, both functions are executed in parallel by the control unit 10 in parallel. May be. Further, instead of displaying the background image simulating the piano roll over the entire area in the time axis direction (X-axis direction), the display control means 13 instead of displaying the background image simulating the piano roll only on the left end of the touch screen 31. May be displayed. According to the modified example 12 described above, a user can intuitively assign a note string to a character string.

<Modification 13>
In the embodiment, the storage unit 20 includes a pronunciation dictionary DB 21 composed of a plurality of pronunciation records including the pronunciation time ratio of each character constituting the character string, and the assigning means 18 includes the character string sound length and the pronunciation record. However, the present invention is not limited to this, and instead of the pronunciation dictionary DB 21, an initial value pronunciation dictionary DB in which the pronunciation record includes the absolute value of the pronunciation time for each character is determined. May be included in the storage unit 20.

FIG. 17 is a diagram showing the contents of the initial value pronunciation dictionary DB.
Each pronunciation record included in the initial value pronunciation dictionary DB includes a plurality of items such as an identification ID, a character, and an initial value pronunciation time. The identification ID is an ID for uniquely identifying each pronunciation record, and is composed of, for example, a 4-digit number. The character is one character that is predetermined as a character to be pronounced. The initial value sounding time is an initial value of the sounding time allocated in advance for each character in each sounding record. In the initial value pronunciation time, the initial value of the sound length of each character is determined in advance based on the length of time that is required when the corresponding character is pronounced with natural inflection, obtained experimentally. . For example, in FIG. 17, for the characters “A”, “I”, “U”, and “E”, a pronunciation time of “0.3 seconds” is assigned in advance as an initial value.

18A and 18B are diagrams illustrating display contents of the speech synthesizer 100 according to the modification 13.
FIG. 18A shows a state immediately after the user inputs a character string in the text box 101 and the pitch curve 103 is not input by the user. In FIG. 18 (a), the character string “Atama” is input, and for each character “A”, “TA” and “MA”, the assigning means 18 stores the character string in the initial value pronunciation dictionary DB. Based on the pronunciation record, an initial value pronunciation time of the same length is assigned as a sound length. In FIG. 18A, the default pitch curve 103 is displayed by the display control means 13 based on the horizontal width of the input character image 104 in the X-axis direction and the position of the pronunciation reference line 102 in the Y-axis direction. Yes.

  FIG. 18B shows a state in which the user has input the pitch curve 103 from the state of FIG. In FIG. 18B, the display position in the pitch direction (Y-axis direction) is changed by the display control unit 13 for each of the input character images 104 according to the shape of the pitch curve 103. Here, since the initial sound generation time is already assigned to each of the input character images 104, the ratio between the character string sound length and the sound generation time is assigned by the assigning means 18 according to the shape of the pitch curve 103 as in the embodiment. A sound length based on a plurality of pronunciation records including is not assigned to each input character image 104. On the other hand, the length from the left end of the input character image 104 representing the first character of the input character string to the right end of the input character image 104 representing the last character of the input character string (that is, the time axis direction of the input character image 104 ( When the width in the (X-axis direction) is shorter than the length in the time-axis direction (X-axis direction) of the input pitch curve 103, the following may be performed. In this case, the assigning unit 18 assigns a sound length that matches the end of the input pitch curve 103 to the input character image 104 representing the last character of the input character string. In FIG. 18B, the assigning unit 18 assigns a sound length corresponding to the end of the input pitch curve 103 to the input character image 104 “ma” representing the last character of the input character string. Yes.

  In addition, the process in which the assigning unit 18 assigns the sound length to the input character image 104 representing the last character of the input character string in accordance with the end of the pitch curve 103 can be set to “ON / OFF” by the user. It is good. Further, when the user inputs the pitch curve 103, the sound generation time assigned to each of the input character images 104 by the assigning means 18 is an initial value to the last. By dragging the left end (or right end) of the rectangle representing the image 104 or the assigned character image 109 in the time axis direction (X-axis direction), the sound length assigned to each character can be changed. Further, when the user changes the setting via the touch screen 31, instead of the tone length assignment process based on the initial value pronunciation dictionary DB as in the modification, the pronunciation time of each character in the embodiment is changed. It may be possible to shift to a sound length assignment process based on the pronunciation dictionary DB 21 storing the ratio. On the contrary, when the user changes the setting via the touch screen 31, instead of the tone length assignment process based on the pronunciation dictionary DB 21 storing the ratio of the pronunciation time of each character, the modification example It may be possible to shift to a sound length assignment process based on such an initial value pronunciation dictionary DB. According to the modified example 13 described above, for each character, the sound length when it is pronounced with natural inflection is assigned as an initial value regardless of the operation of the user. Note that both the pronunciation dictionary DB 21 and the initial value pronunciation dictionary DB in the embodiment correspond to a pronunciation length dictionary storage unit.

<Modification 14>
The hardware configuration of the speech synthesizer 100 is not limited to that described with reference to FIG. The speech synthesizer 100 may have any hardware configuration as long as the functions shown in FIG. 5 can be implemented. For example, the speech synthesizer 100 may have dedicated hardware (circuit) corresponding to each of the functional elements shown in FIG.

<Modification 15>
The program relating to the speech synthesis application described in the above-described embodiment includes a magnetic recording medium (magnetic tape, magnetic disk (HDD, FD (Flexible Disk)), etc.), optical recording medium (optical disk (CD (Compact Disk)), DVD (Digital Versatile Disk))), a magneto-optical recording medium, and a computer-readable recording medium such as a semiconductor memory (flash ROM or the like). The program may be downloaded via a network such as the Internet.

DESCRIPTION OF SYMBOLS 10 ... Control part, 11 ... Character string acquisition means, 12 ... Reference | standard sound length specification means, 13 ... Display control means, 14 ... Character space | interval control means, 15 ... Trajectory analysis means, 16 ... Voice record generation means, 17 ... Speech synthesis Means, 18 ... Allocation means, 20 ... Storage part, 21 ... Pronunciation dictionary DB, 22 ... Shortest pronunciation time DB, 23 ... Audio DB, 24 ... Acoustic effect DB, 30 ... UI part, 31 ... Touch screen, 40 ... Audio output , 100 ... speech synthesizer, 101 ... text box, 102 ... pronunciation reference line, 103, 103a to 103f ... pitch curve, 104, 104a, 104a ', 104b ... input character image, 105 ... playback button image, 106 ... return Button image, 107, 107a, 107a ′, 107b... Input character line, 108A to 108E... Input character image group, 109. 110 ... Case, 111 ... Speaker, 112 ... Menu button image, 113 ... Time axis scale, A to E, A '... Intersection, L1a, L1a', L2 ... Tangent, α, α '... Difference length, β, γ …distance

Claims (5)

A character string acquisition means for acquiring a character string composed of a plurality of characters;
Character string display means for displaying each character constituting the acquired character string on a display means;
When a graphic in a coordinate system having a first axis representing time and a second axis representing pitch is specified by the user, the display means is associated with each character constituting the character string. Graphic display means to be displayed on,
In the displayed graphic, based on the coordinate value of the position corresponding to each character constituting the displayed character string, an assigning unit that assigns a pitch and a sound length to each character;
A speech synthesizer comprising: speech synthesizer that synthesizes speech data that causes each character constituting the character string to be pronounced at a pitch and a pitch assigned by the assigning unit. A pronunciation length dictionary storage means for storing, for a plurality of words, a length of a pronunciation time of each character constituting the word or a ratio of a pronunciation time of each character constituting the word with respect to a pronunciation time when the word is pronounced Prepared,
The assigning means is a sound length used when the entire character string is pronounced and is stored in the pronunciation length dictionary storage means for the character string sound length designated by the user and each character constituting the character string. The speech synthesizer according to claim 1, wherein a sound length is assigned to each character based on the length of the pronunciation time or the ratio of the pronunciation times. Comprising indicator display means for displaying on the display means an indicator for assigning pitch or length to each character constituting the character string;
The allocating unit corrects the coordinate value of the position corresponding to each character constituting the character string in the graphic according to the index displayed by the index display unit, and based on the corrected coordinate value, The speech synthesizer according to claim 1 or 2, wherein pitch and length are assigned to the character. Corresponding to each of the shapes of a plurality of figures, comprising acoustic effect storage means for storing the acoustic effect applied when a character is pronounced,
When the figure superimposed on the figure displayed on the display means is designated by the user, the figure display means displays the superimposed figure on the display means,
The assigning means specifies a shape of a figure whose similarity with the superimposed figure exceeds a threshold among the shapes of the plurality of figures stored in the acoustic effect storage means, and sets the shape of the specified figure. The sound effect stored in association with each other is assigned to the character displayed at a position corresponding to the coordinate value of the superimposed figure. Voice synthesizer. On the computer,
A character string acquisition function for acquiring a character string composed of a plurality of characters;
A character string display function for causing the display means to display each character constituting the acquired character string;
When a graphic in a coordinate system having a first axis representing pitch and a second axis representing time is designated by the user, the display means is associated with each character constituting the character string. Graphic display function to be displayed on
An assignment function for assigning a pitch and a tone length to each character based on the coordinate value of the graphic corresponding to each character constituting the displayed character string;
A speech synthesis function for synthesizing speech data that causes each character constituting the character string to be pronounced at a pitch and a tone length assigned by the assignment function.

Images