JPH09244678A - Speech synthesizer - Google Patents

Abstract

(57) [Abstract] [Problem] Conventionally, since the control of the fundamental frequency is performed independently of the speech unit and the information on the fundamental frequency of the unit is not used, only the synthesized sound with poor clarity and naturalness is obtained. There was a problem that I could not get it. SOLUTION: A symbol string input terminal 1 for inputting data related to pronunciation, a voice unit data storage unit 60 for storing a plurality of voice units, and a plurality of fundamental frequencies of the voice units are changed to predetermined fundamental frequencies. In this case, the fundamental frequency corresponding to a part of the speech unit is changed, and the fundamental frequency control unit 40 that outputs the result and a plurality of speech units are output based on the output of the fundamental frequency control unit 40. A voice synthesizing unit 5 for synthesizing is provided.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a speech synthesizer which can be used for controlling intonation (fundamental frequency), for example.

[0002]

2. Description of the Related Art FIG. 6 is a block diagram showing a configuration of a general acoustic processing unit of a conventional rule synthesizing device (speech synthesizing device).

A symbol string such as reading or accent is input to the symbol string input terminal 1. The input symbol string is analyzed by the analysis unit 2 to obtain information about phonemes and information about prosody. The time length control unit 3 sets the time length for each segment, and the basic frequency control unit 4 determines the basic frequency.
Finally, the voice synthesis unit 5 reads out the necessary data from the accumulated voice unit data storage unit 6 and outputs the synthesized sound waveform to the waveform output terminal 7.

As a method used for prosody control, a point fundamental frequency pattern for determining a fundamental frequency for each syllable, a Fujisaki model obtained by a mathematical expression, etc. are generally used.

Here, an example using a point fundamental frequency pattern as a prosody control method and a syllable (CV) as a unit of a phoneme will be described. In advance, depending on the accent type and the number of mora, the fundamental frequency (F
(Also called 0) is created as a table.
When synthesizing the "floor", the pieces / yu /, / ka / are read and connected. Since it is a 2-molar type 0, the fundamental frequency obtained by preliminarily reading from the table and performing linear interpolation is shown in FIG. That is, in the figure, the starting point 71 of / yu /,
A center point 72, a starting point 73 of / ka /, a center point 74, and the like are shown. In FIG. 7, the vertical axis corresponds to the fundamental frequency,
The horizontal axis is the time axis.

[0006]

Generally, a rule synthesizing apparatus stores voices in the form of waveforms or parameters in advance in some unit (syllable, VCV, environment-dependent type, etc.), and according to an input character string. There are many methods of processing them to obtain a synthetic sound.

The fundamental frequency is influenced by various factors such as sentence syntax (phrase component), accent position (accent component), syllable type before and after (microprosody), fluctuation and the like. The accumulated speech unit holds information about the fundamental frequency. However, since the conventional device controls the fundamental frequency independently of the speech unit, it does not use the information about the fundamental frequency of the unit. Therefore, there was a problem that only synthetic sounds with poor clarity and naturalness could be obtained.

The present invention has been made in view of the above problems of the conventional apparatus, and an object of the present invention is to provide a speech synthesizing apparatus capable of synthesizing a synthetic speech having more excellent clarity and naturalness as compared with the conventional apparatus.

[0009]

According to a first aspect of the present invention, there is provided an input means for inputting data relating to pronunciation, a voice element storage means for storing a plurality of voice elements, and a plurality of voice elements included in the voice element. When changing the fundamental frequency to a predetermined fundamental frequency, the fundamental frequency corresponding to a part of the speech element is changed, and the fundamental frequency control means for outputting the result, and the fundamental frequency control means A voice synthesizing device comprising: a voice synthesizing unit for synthesizing a voice by combining a plurality of voice units based on an output.

The present invention according to claim 2 has a symbol string input means for inputting a symbol string such as a phonetic symbol, a fundamental frequency control means for controlling a fundamental frequency, and a voice synthesizing means for synthesizing a voice. The fundamental frequency control means is a speech synthesizer that relatively controls a portion of the fundamental frequencies of the speech unit when changing the fundamental frequency of the speech unit used in the speech synthesis means to the target fundamental frequency. is there.

According to a third aspect of the present invention, there is provided a speech synthesizing device in which a part of the speech unit is substantially in the same range for each speech unit.

The present invention according to claim 4 has a symbol string input means for inputting a symbol string such as a phonetic symbol, a fundamental frequency control means for controlling a fundamental frequency, and a voice synthesizing means for synthesizing a voice. The fundamental frequency control means is a speech synthesizing device that holds the fundamental frequency of the speech unit used in the speech synthesis means in the form of a least squares line and its distance, and controls the slope of the least squares line. .

According to a fifth aspect of the present invention, the fundamental frequency control means is a voice synthesizing device for increasing the absolute value of the slope of the least-squares straight line when the preceding syllable is a consonant.

The present invention according to claim 6 has a symbol string input means for inputting a symbol string such as a phonetic symbol, a fundamental frequency control means for controlling a fundamental frequency, and a voice synthesizing means for synthesizing a voice. The fundamental frequency control means holds the fundamental frequency of the speech unit used in the speech synthesis means in the form of a least squares straight line and its distance, and uses the distance component when changing to a target fundamental frequency. It is a voice synthesizer.

The present invention according to claim 7 has a symbol string input means for inputting a symbol string such as a phonetic symbol, a fundamental frequency control means for controlling a fundamental frequency, and a voice synthesizing means for synthesizing a voice, The fundamental frequency control means is a speech synthesizer that lowers the fundamental frequency when the succeeding syllable has a silent portion.

In the present invention according to claim 1, the input means is
When data regarding pronunciation is input, the speech element storage means stores a plurality of speech elements, and the fundamental frequency control means changes a plurality of fundamental frequencies of the speech element to a predetermined fundamental frequency, The fundamental frequency corresponding to a part of the speech unit is changed, the result is output, and the voice synthesis unit combines a plurality of the speech units based on the output of the fundamental frequency control unit. To synthesize the voice.

In the present invention according to claim 2, the symbol string input means inputs a symbol string such as a phonetic symbol, the fundamental frequency control means controls the fundamental frequency, and the speech synthesizing means synthesizes the speech, and the fundamental speech is generated. The frequency control means relatively controls a part of the fundamental frequencies of the speech element when changing the fundamental frequency of the speech element used in the speech synthesis means to a target fundamental frequency.

In the present invention as set forth in claim 3, a part of the speech units is substantially in the same range for each speech unit.

In the present invention according to claim 4, the symbol string input means inputs a symbol string such as a phonetic symbol, the fundamental frequency control means controls the fundamental frequency, and the speech synthesizing means synthesizes speech,
The fundamental frequency control means holds the fundamental frequency of the speech unit used in the speech synthesis means in the form of a least squares straight line and its distance, and controls the slope of the least squares straight line.

According to the fifth aspect of the present invention, the fundamental frequency control means increases the absolute value of the slope of the least-squares straight line when the preceding syllable is a consonant.

In the present invention according to claim 6, the symbol string input means inputs a symbol string such as a phonetic symbol, the fundamental frequency control means controls the fundamental frequency, and the speech synthesizing means synthesizes speech,
The fundamental frequency control means holds the fundamental frequency of the speech unit used in the speech synthesis means in the form of a least squares straight line and its distance, and uses the distance component when changing to a target fundamental frequency. To do.

In the present invention according to claim 7, the symbol string input means inputs a symbol string such as a phonetic symbol, the fundamental frequency control means controls the fundamental frequency, and the speech synthesizing means synthesizes speech,
The fundamental frequency control means lowers the fundamental frequency when the succeeding syllable has a silent portion.

Accordingly, in the present invention, for example, when changing the fundamental frequency of the accumulated speech unit to the target fundamental frequency, the fundamental frequency of a part of the speech unit is relatively controlled, It is possible to provide a synthetic voice with high clarity and naturalness by making use of the fundamental frequency of the speech unit. Further, the fundamental frequency is relatively controlled within a fixed range regardless of the type of consonant.

[0024]

BEST MODE FOR CARRYING OUT THE INVENTION Embodiments of a speech synthesizer according to the present invention will be described below with reference to the drawings.

(Embodiment 1) FIG. 1 is a block diagram showing a configuration of a speech synthesizer according to an embodiment of the present invention, and the configuration of the present embodiment will be described with reference to the figure.

Note that the substantially same components as those described with reference to FIG. 6 are designated by the same reference numerals, and the description thereof will be omitted.
The main difference between the speech synthesizer of this embodiment and the apparatus of FIG. 6 is that the fundamental frequency control unit 40 corresponds to a fundamental frequency corresponding to a part of the speech units stored in the speech unit data storage unit 60. Is to change the fundamental frequency of the speech unit. As will be described later with reference to FIG. 2, the speech unit data storage unit 60 also stores usage range information that defines in advance which range of the basic frequency information to use. Is.

The prosody control method of this embodiment will be described with reference to FIG. FIG. 2A is a diagram in which the content of the voice unit data stored in the voice unit data storage unit 60 regarding / ka / is displayed in an easy-to-understand manner. In the figure, the fundamental frequency mark 22 included in the speech unit substantially corresponds to the peak position of each amplitude in the amplitude of the waveform data 21 of the speech unit and is represented by a short vertical line. The fundamental frequency in the transfer section 23, which is a part of the speech unit, is
It is easily affected by the type of consonant that precedes it, and such an effect is generally called microprosody. As shown in FIG. 2A, in addition to the waveform data 21 of the speech segment and the fundamental frequency mark 22 as the data of the speech segment, the basic frequency information included in the waveform data 21 at the time of speech synthesis is included. The use range information (t0, t1) predetermined for each speech unit as to which range of the basic frequency to be used is included. In FIG. 2A, as the usage range information (t0, t1), the time point corresponding to the basic frequency mark shown at the leftmost end of the basic frequency mark 22 is set to t0, and from that time point, the positive time axis is indicated. The time point corresponding to the second fundamental frequency mark in the direction is defined as t1. The usage range information (t0, t1) is determined so that the microprosody possessed by the segment can be effectively used. In FIG. 2A,
It shows a case where the usage range information (t0, t1) is determined so as to correspond to the range of the passing section 23. In this way, the range in which the basic frequency information is used is marked on the speech unit, and the basic frequency information of the speech unit is used for the basic frequency control in that range.

In FIG. 2 (b), when synthesizing the "floor", the fundamental frequency pattern of / ka / determined by the conventional method (corresponding to the portion of the fundamental frequency of / ka / in FIG. 7). 2C, and FIG. 2C shows the fundamental frequency pattern obtained by the present embodiment. 2 (b) and (c)
The start point 73 and the center point 74 shown in (1) correspond to the points with the same reference numerals described with reference to FIG.

A method of obtaining the fundamental frequency according to this embodiment will be described. First, the fundamental frequency f by the conventional method
Find (t). Next, regarding the bridging portion 23, the fundamental frequency is determined as follows in order to effectively utilize the microprosody possessed by the element. The time is t, the target fundamental frequency is F (t), and the fundamental frequency of the segment is F0.
(T), the fundamental frequency according to the conventional example at time t1 is f
(T1). Fundamental frequency F of the fragment at time t1
The difference between 0 (t1) and f (t1) is defined as the relative frequency SA.

The following equation gives the target fundamental frequency F
It is an expression for obtaining (t). (1) When t0 ≦ t ≦ t1: F (t) = F0 (t) + S
A (2) At other times: F (t) = f (t) where SA = F0 (t1) −f (t1) F (t) obtained using the above equation is shown in FIG. It is the fundamental frequency corresponding to the vertical axis of the figure.

F (t0) corresponding to the time t0 is calculated from the fundamental frequency F0 (t0) of the segment and the relative frequency SA.
It is obtained from the above equation (1), and in FIG.
Corresponds to the point marked with 4. Similarly, in FIG.
The fundamental frequency corresponding to 5 is also obtained. Further, the point denoted by reference numeral 26 in the same figure corresponds to the value of t = t1 in FIG. 2 (b).

As described above, according to this embodiment, the fundamental frequency information of the segment can be effectively used, that is,
Since it is possible to effectively reproduce the microprosody of the segment, it is possible to provide a synthetic sound with high clarity and naturalness.

Incidentally, in this implementation order, C is used as a fragment.
As the portion that uses V and the fundamental frequency information, the transition portion 23 is provided as a part of the voice unit, but the present invention is not limited to this, and a consonant portion may be included.

(Second Embodiment) The configuration of the speech synthesizer according to the present embodiment is the same as that shown in FIG.

In the above-described first embodiment, the case where the portion using the fundamental frequency information of the speech unit is a variable type in which the optimum range is individually predetermined for each speech unit has been described. In the form, the data of the voice unit has the following configuration.

In other words, the usage range information (data of each speech unit stored in the speech unit data storage unit 60 is set so that the range in which the basic frequency information is used is the same range in all the speech units. Predetermine t0, t1).

As a result, the fundamental frequency information of the element can be effectively used by the simpler method as in the first embodiment, that is, the microprosody of the element can be effectively reproduced. It is possible to provide a synthetic sound that is highly natural and natural.

(Embodiment 3) Next, a speech synthesizer according to another embodiment of the present invention will be described with reference to FIGS.
A prosody control method using the same device will be described with reference to (c).

The structure of the speech synthesizer of this embodiment is the same as that of FIG. 1 except for the contents of the basic frequency control unit 40, and therefore its explanation is omitted.

FIG. 3A is a diagram showing the fragment data of / ka / together with the fundamental frequency mark. FIG. 3 (b) is a graph showing the fundamental frequency as a value,
The least squares straight line (y = at + b) of the part <t0-t1> affected by the consonant is entered. FIG. 3 (c)
It is the figure which showed the difference (Sa (t)) of the fundamental frequency of a least squares straight line and an element.

That is, Sa (t) = F0 (t) -y (t)
It is.

Microprosody is affected by the vocalization speed, average vocalization height, and the like. In the present invention, the microprosody possessed by the segment is defined as the slope (least squares line
(T)) and the fluctuation (Sa (t)), and the slope of the least squares line is changed and applied according to the utterance speed and the average height. For example, when speaking speed is high,
When the absolute value of the inclination a is reduced and the vocalization speed is slow,
The absolute value of the inclination a is increased to make microprosody remarkable. The formula is as follows.

Least-squares line: y (t) = at + b Fluctuation: Sa (t) = F0 (t) -f (t) (Example 1) When speech rate is slow: a '= 2a (double slope) y' (T) = 2a (t) + b When t0 ≦ t ≦ t1: F (t) = y ′ (t) + Sa
(T) At other times: F (t) = f (t) (Example 2) When the vocalization speed is fast: a ″ = a / 2 (half inclination) y ″ (t) = a / 2 (t) + b When t0 ≦ t ≦ t1: F (t) = y ″ (t) + Sa
(T) At other times: F (t) = f (t) As described above, according to the present embodiment, the fundamental frequency information possessed by the segment is effectively used, and control according to the utterance speed is performed. By doing so, it is possible to provide a synthesized voice with high clarity and naturalness.

(Embodiment 4) Next, a speech synthesizer according to another embodiment of the present invention will be described with reference to FIGS.
A prosody control method using the same device will be described with reference to (c).

The structure of the speech synthesizer of this embodiment is the same as that of the third embodiment except the contents of the basic frequency control section 40, and the basic structure is the same as that of FIG. The description is omitted.

Basically, it is similar to that of the third embodiment, but the influence (microprosody) of the consonant type on the fundamental frequency pattern is remarkable when the preceding syllable is a consonant. For example, the microprosody of the consonant "k} in" yuka "and" yukka "is more prominent in the latter case, that is, the slope is larger.
However, in a general synthesizer, / ka / and / kka /
It is common to use the / ka / segment instead of holding the individual segments of. Therefore, when the preceding syllable is a consonant, a more natural fundamental frequency pattern is obtained by increasing the absolute value of the slope of the least-squares straight line that describes the microprosody, as in the case of the slow vocalization rate of the third embodiment. Can be generated.

(Embodiment 5) Next, FIG. 4A to FIG. 4 will be used for a speech synthesizer according to another embodiment of the present invention.
A prosody control method using the same device will be described with reference to (c).

The structure of the speech synthesizer of this embodiment is basically the same as that of FIG. 1 except for the contents of the basic frequency control section 40, and therefore its explanation is omitted.

FIG. 4A is a diagram showing the fragment data of / ka / together with the fundamental frequency mark. FIG. 4B shows the fundamental frequency as a graph, and the least squares straight line of the portion t1-t2 excluding the influence of consonants is entered. FIG. 4C is a diagram showing the difference between the least square line and the fundamental frequency of the element, that is, the fluctuation of the fundamental frequency. In the present invention, the fundamental frequency is controlled as follows in order to utilize the fluctuation of the fundamental frequency of this segment. When the fluctuation component is y (t) and the fundamental frequency given by the conventional method is f (t), (1) When t0 <t <t1: F (t) = Fo + f (t1) (2) Other When: F (t) = f (t) + y (t).

As described above, according to the present embodiment, it is possible to effectively utilize the fluctuation of the fundamental frequency information of the segment and provide a synthesized voice with high clarity and naturalness.

(Sixth Embodiment) Next, a speech synthesizer according to another embodiment of the present invention will be described with reference to FIG.
A prosody control method using the same device will be described with reference to (b).

The configuration of the speech synthesizer of this embodiment is basically the same as that of FIG. 1 except for the contents of the fundamental frequency control section 40, so its explanation is omitted.

FIG. 5 (a) shows a fundamental frequency pattern according to a conventional method when synthesizing / yuka /. /
Before ka /, there is a silent part, but normally the fundamental frequency pattern falls toward silent parts and unvoiced consonants. Therefore, in order to achieve that, the subsequent consonants are bursts and consonants,
In the case of unvoiced consonants such as fricatives, the fundamental frequency is controlled as follows. The basic frequency pattern according to this method is shown in Fig. 5 (b).
Shown in

As shown in FIG. 5B, the damping start point t1
Is set after the center of gravity of the vowel in / yu /, and t2 is the starting point of the next syllable. When the fundamental frequency given by the conventional method is f (t) and the following consonant is an unvoiced consonant (1) When t1 <t <t2: F (t) = f (t) + g (t) (2) At other times: F (t) = f (t).

However, g (t) is a function representing attenuation and is set by a linear expression so that g (t1) = 0.

As described above, according to the present embodiment, when the subsequent syllable is an unvoiced consonant, the attenuation of the fundamental frequency of the transition portion of the vowel is realized to provide a synthetic sound with high clarity and naturalness. You can do it.

As described above, according to the present embodiment, it is possible to provide a synthesized voice with high clarity and naturalness by effectively utilizing the fundamental frequency information of the segment used for synthesis.

Each means of the present invention may be realized by software using a computer, or may be realized by using a dedicated hardware circuit having each of these functions.

The unit of synthesis may be any of CV, VCV, environment-dependent type, etc., and the method of accumulating the pieces may be a waveform or a parameter. Further, although table reference is used as the fundamental frequency control method, this does not limit the present invention in any way.

[0060]

As is apparent from the above description, the present invention has an advantage that a synthesized sound having a superior clarity and naturalness can be synthesized as compared with the conventional art.

[Brief description of drawings]

FIG. 1 is a block diagram showing a configuration of a speech synthesizer according to a first embodiment of the present invention.

FIG. 2 is a diagram for explaining a prosody control method according to the first embodiment of the present invention.

FIG. 3 is a diagram for explaining a prosody control method according to a third embodiment of the present invention.

FIG. 4 is a diagram for explaining a prosody control method according to a fifth embodiment of the present invention.

FIG. 5 is a diagram for explaining a prosody control method according to a sixth embodiment of the present invention.

FIG. 6 is a block diagram showing a configuration of a conventional speech synthesizer.

FIG. 7 is a diagram for explaining a conventional prosody control method.

[Explanation of symbols]

 1 Symbol string input end 2 Symbol string analysis part 3 Time length control part 4 Basic frequency control part 5 Acoustic processing part 6 Speech element data 7 Synthetic sound output end

Claims (7)

[Claims] 1. An input unit for inputting data relating to pronunciation, a voice unit storage unit for storing a plurality of voice units, and a case of changing a plurality of fundamental frequencies of the voice unit to a predetermined fundamental frequency, Based on the output of the fundamental frequency control unit that changes the fundamental frequency corresponding to a part of the speech unit and outputs the result, the plurality of speech units are combined. And a voice synthesizing unit for synthesizing a voice. 2. A symbol string input means for inputting a symbol string such as a phonetic symbol, a fundamental frequency control means for controlling a fundamental frequency, and a voice synthesizing means for synthesizing a voice, the fundamental frequency control means comprising: A voice synthesizing apparatus, wherein when changing a fundamental frequency of a voice unit used in the voice synthesizing unit to a target fundamental frequency, a part of the fundamental frequency of the voice unit is relatively controlled. 3. The voice synthesizing apparatus according to claim 1, wherein a part of the speech units is in substantially the same range for each speech unit. 4. A symbol string input means for inputting a symbol string such as a phonetic symbol, a fundamental frequency control means for controlling a fundamental frequency, and a voice synthesizing means for synthesizing a voice, the fundamental frequency control means comprising: A voice synthesizing apparatus, characterized in that a fundamental frequency of a voice unit used in the voice synthesizing unit is held in a form of a least squares straight line and its distance, and a slope of the least squares straight line is controlled. 5. The speech synthesizer according to claim 4, wherein the fundamental frequency control means increases the absolute value of the slope of the least squares line when the preceding syllable is a consonant. 6. A symbol string input means for inputting a symbol string such as a phonetic symbol, a fundamental frequency control means for controlling a fundamental frequency, and a voice synthesizing means for synthesizing a voice, the fundamental frequency control means comprising: A voice characterized by holding a fundamental frequency of a voice unit used in the voice synthesizing unit in the form of a least-squares straight line and its distance, and utilizing the distance component when changing to a target fundamental frequency. Synthesizer. 7. A symbol string input means for inputting a symbol string such as a phonetic symbol, a fundamental frequency control means for controlling a fundamental frequency, and a voice synthesizing means for synthesizing a voice, the fundamental frequency control means comprising: A speech synthesizer characterized by lowering the fundamental frequency when a subsequent syllable is accompanied by a silent part.