JPH0758438B2 - Long sound combination method - Google Patents

Description

Description: TECHNICAL FIELD The present invention relates to a long sound combining method in a speech synthesizing method based on a rule synthesizing method.

(Prior Art) A conventional long sound combining method will be described with reference to FIG. FIG. 4 is a diagram showing a conventional long sound combining method using an amplifier parameter. In the figure, 31 is the time length of the preceding syllable, 32 is the time length of the long syllable, 33 is the time length of the following syllable, 34 is the vowel part of the previous syllable, and 35 is the constant section start frame in the syllable file of the previous syllable. Parameter 36 is the parameter of the stationary section end frame in the syllable file of the previous syllable. 37 is a frame for embedding the same vowel as the vowel in the preceding syllable. 38 (hatched part) is the parameter from the beginning of the syllable file of this vowel to the stationary section start frame 39, and 39 and 40 are the above-mentioned vowels. The parameters of the start and end frames of the steady section in the vowel syllable file are shown. That is, the conventional long syllable combining method uses the parameters of the syllable file of the same vowel as the vowel of the previous syllable in the section corresponding to the long sound, and these parameters are stored in the syllable file of the previous syllable as in the method of combining other syllables. It is connected to the parameter of.

As described above, even in the above-described conventional combining method, if the single vowel vowel in the preceding syllable is replaced with the section corresponding to the long syllable, it can be processed by the same algorithm as other normal syllables.
Further, although FIG. 4 shows the amplifier parameters, the linear prediction coefficient parameters are also subjected to the same method.

(Problems to be Solved by the Invention) However, in the above-described conventional method, since the same vowel as the vowel in the preceding syllable is used in the section corresponding to the long sound, there is a problem that it is hard to hear like a long sound.

The present invention provides a long sound combining method capable of synthesizing long sounds by the rule synthesizing method.

(Means for Solving Problems) In order to solve the above problems, the present invention uses a method of using an amplifier parameter, in which a parameter of a preceding syllable is combined with a constant value up to an end frame of a long syllable, and linear prediction is performed. In the method using the coefficient parameter, the parameters in the syllable file of the same vowel as the vowel of the preceding syllable in the syllable file of the same vowel in the long syllable interval are used, and these parameters and the parameters of the stationary section end frame in the syllable file of the vowel of the preceding syllable are set. They are connected by linear interpolation.

(Operation) According to the present invention, the method described above is used to generate a synthesized sound including a long sound by using the parameters in an existing syllable file without adding a new long sound file to the syllable file. It is possible to create a sequence and obtain a long sound that sounds natural to the sense of hearing.

(Embodiment) An embodiment of the present invention will be described with reference to FIGS. 1 and 2.
FIG. 1 is a diagram showing a combining method using an amplifier parameter in one embodiment of the present invention, and FIG. 2 is a diagram showing a combining method using a linear prediction coefficient parameter in one embodiment of the present invention. In FIG. 1, 1 is the time length of the previous syllable, 2 is the time length of the long syllable, 3 is the time length of the subsequent syllable, 4 is the vowel part of the previous syllable, and 5 is the start frame of the stationary section in the syllable file of the previous syllable. Parameter, 6 is the parameter of the stationary section end frame in the syllable file of the previous syllable, 7 is the section corresponding to the long sound,
Reference numeral 8 (indicated by a dot on the time axis) represents a parameter of a stationary section start frame in the syllable file of the long sound section, and 9 represents a parameter of a stationary section end frame in the syllable file of the long sound section. As shown in FIG. 1, in the method using the amplifier parameter, the parameter of the end frame 6 of the preceding syllable is used as it is with a constant value up to the steady section end frame 9 in the long syllable file.

In FIG. 2 showing the combination method using the linear prediction coefficient parameter, 11 is the time length of the preceding syllable, 12 is the time length of the long sound section,
13 is the time length of the succeeding syllable, 14 is the vowel section of the preceding syllable, 15 is the parameter of the stationary section start frame in the syllable file of the preceding syllable, 16 is the parameter of the stationary section end frame in the syllable file of the preceding syllable, and 17 is A long syllable section, 18 is a parameter of a stationary section start frame in a syllable file of the same vowel as the previous syllable, and 19 is a parameter of a stationary section end frame in a syllable file of a vowel.

As shown in FIG. 2, in the method using the linear prediction coefficient parameter, the parameter 15 of the stationary section start frame in the syllable file of the preceding syllable, the parameter 16 of the stationary section end frame in the same syllable file, and the vowel of the preceding syllable are the same. 15 to 16,16 using four parameters of the parameter 18 of the stationary section start frame in the syllable file of the single vowel and the parameter 19 of the stationary section end frame in the same syllable file.
To 18 and 18 to 19 are combined by linear interpolation.

As described above, according to the embodiment of the present invention, in the combining method using the amplifier parameter, the parameter of the preceding syllable is used as it is up to the end frame of the long syllable, so that there is an advantage that there is little discontinuity in hearing. In addition, in the combination method using the linear prediction coefficient parameter, since the parameter in the syllable file of the same vowel as the vowel of the preceding syllable is used for the long syllable section, the feature of the phoneme can be retained as it is.

Next, another embodiment of the present invention will be described with reference to FIG. FIG. 3 is a diagram showing a combining method using an amplifier parameter when a long sound comes at the end of a word according to another embodiment of the present invention. In the figure, 21 is the time length of the previous syllable, 22 is the time length when the long sound comes in the word (same as the time length of the long sound shown in 2,12,32 in FIGS. 1, 2, 4), and 23 is Vowel section of the previous syllable, 24 is the parameter of the start section frame of the regular section in the syllable file of the previous syllable, 25 is the parameter of the end frame of the regular section of the syllable file of the previous syllable, 26 is the long section, and 27 is the syllable file of the long section. In the figure, 28 is a parameter of the stationary section start frame, and 28 is a parameter of the stationary section end frame in the syllable file of the long sound section. Similar to the case shown in FIG. 1, the amplifier parameter holds a constant value up to the stationary section end frame 28 of the long sound section, but when the long sound comes at the end, it is shorter than when the long sound comes in the word. Since it is uttered, the constant value is held until the stationary section start frame 27 of the long sound section, and 0 is linearly interpolated at the stationary section end frame 28. When the linear prediction coefficient parameter is used in the same case, the same combination method (combination method shown in FIG. 2) as in the case where a long sound comes in a word is used.

By the word ending processing as described above, a long sound that is natural to the sense of hearing can be obtained.

(Effects of the Invention) According to the present invention, an amplifier parameter and a linear prediction in an ordinary syllable file used in CV (consonant, vowel) syllable combination in the regular synthesis method without specially creating a long sound file. By the combination method using the coefficient parameter,
It is possible to synthesize long sounds that are extremely natural in terms of hearing.

[Brief description of drawings]

FIG. 1 is a diagram showing a combination method by an amplifier parameter in one embodiment of the present invention, and FIG. 2 is a diagram showing a combination method by a linear prediction coefficient parameter in one embodiment of the present invention,
FIG. 3 is a diagram showing a combining method using an amplifier parameter when a long sound comes at the end of another embodiment of the present invention, and FIG. 4 is a diagram showing a conventional long sound combining method using an amplifier parameter. 1,11,21 …… Time length of previous syllable, 2,12,22 …… Time length of long sound, 3,13 …… Time length of succeeding syllable, 4,14,23 …… Vowel section of previous syllable, 5 , 24 …… Amplifier parameter of the start frame of the steady section in the syllable file of the previous syllable, 6,25 …… Amplifier parameter of the end frame of the steady section of the syllable file of the previous syllable, 7,17,26 …… Long section, 8 , 27 …… Amplifier parameter of the start frame of the steady section in the syllable file of the long section,
9,28 …… Amplifier parameter of stationary section end frame in long syllable file, 15 …… Linear prediction coefficient parameter of stationary section beginning frame in syllable file of previous syllable,
16 …… Linear prediction coefficient parameter of the stationary section end frame in the syllable file of the previous syllable, 18 …… Linear prediction coefficient parameter of the stationary section start frame in the syllable file of a single vowel, 19 …… Syllable file of the single vowel Linear prediction coefficient parameter of the stationary section end frame in.

Claims (1)

[Claims] 1. A long sound combining method for a case where a long syllable comes in a succeeding syllable, wherein in a combination using an amplifier parameter, a parameter of a previous syllable is combined with a constant value up to an end frame of a long syllable syllable, and in a combination using a linear prediction coefficient parameter, Using parameters in the syllable file of the same vowel as the vowel of the preceding syllable in the long syllable and combining these parameters with the parameters of the stationary section end frame in the syllable file of the vowel of the preceding syllable by linear interpolation. Characteristic long sound combination method.