JPS63210898A - Voice synthesizer - Google Patents

Abstract

(57) [Summary] This bulletin contains application data before electronic filing, so abstract data is not recorded.

Description

DETAILED DESCRIPTION OF THE INVENTION (Field of Industrial Application) The speech synthesis device of the present invention is used for reading aloud character strings input into a word processor or the like, for reading by blind people, etc.

(Prior art) Fig. 4 shows the configuration of a conventional speech synthesizer. 1 is a vocal tract parameter file containing vocal tract parameters obtained by comparing vocal tract characteristics with resonance and anti-resonance; The road parameters are composed of information on formant frequency and bandwidth obtained by analyzing speech every 1 On+s, LSP parameters obtained by converting a spectrum into a line spectrum, and the like. 2, when character strings are input, the vocal tract parameters of the syllables included in those character strings are selected from the vocal tract parameter file 1, arranged temporally, and the vocal tract parameters calculated by interpolation calculation are 3 is an amplifier calculation unit that determines the amplitude of the pulse train and white noise when the amplifier control information is input; 4 is the intonation; When control information is input, an intonation calculation section 5 determines the pulse interval (hereinafter referred to as pitch) of the pulse train, and 5 outputs a pulse train based on the amplitude determined by the amplifier calculation section 3 and the pulse interval determined by the intonation calculation section. A pulse train generator, 6 a white noise generator that outputs white noise based on the amplitude determined by the amplifier calculator 3, and 7 a transmitted wave and reflection in the vocal tract when the pulse train and white noise are input to the vocal tract. This acoustic calculation section 7 obtains a desired audio signal as a transmitted wave from the lips by calculating waves, and this acoustic calculation section 7 is composed of a digital computer. 8 is a D/A that converts the digital audio signal output from the acoustic calculation unit 7 into an analog audio signal.
Converter 9 is a speaker driven by an analog audio signal.

In the conventional example configured in this way, one character string.

When the amplifier control information and intonation control information are input, the intonation calculation unit 4 calculates a pitch value [calculated from the first tone component [see FIG. 4(a)] and the accent component [see FIG. 4(b)].
Hereinafter referred to as control point pitch. (see Figure 4(c))] was given to approximately the center of the vowel of each syllable, and linear interpolation was performed between these spectral stationary points.

(Problem to be Solved by the Invention) However, in the conventional pitch generation method, only one spectral stationary point is given for each syllable.
The pitch drop in the voiced consonant part of syllables such as ``and paba'', and the pitch drop in the voiced consonant part of syllables such as ``he is'', as shown in Figures 4 (a), (b) and (c), for example in the sentence ``he is'' There was a problem that the pitch drop between the syllables ``ga'' and ``i'' in ``Imasu'' could not be expressed.

The present invention has been made in view of these problems, and an object of the present invention is to provide a speech synthesis device that can express complex pitch changes like natural speech.

(Means for Solving the Problems) The present invention calculates the control point pitch between a plurality of spectral stationary points formed within a syllable based on tone components and accent components of intonation control information, and then Interpolation is performed between these spectral stationary points using a nonlinear interpolation function determined by the position of the spectral stationary points.

(Function) According to the present invention, since a plurality of spectral stationary points are provided for each syllable, complex pitch changes can be expressed for each syllable, and nonlinear interpolation between adjacent spectral stationary points is possible. Because it is interpolated with a function.

It can also express sudden changes in pitch between syllables.

(Example) Hereinafter, an example of the present invention will be described in detail with reference to the drawings.

FIG. 1 shows the configuration of an embodiment of the present invention, and the same reference numerals as those in FIG. 3 indicate the same parts, and
lO is a tone/accent component file 11, an intra-syllable pitch file 12. This is an intonation calculation section consisting of a control point pitch calculation section 13, an interpolation function file 14, and an interpolation calculation section 15.

In this embodiment configured in this way, when an input character string and suppression field control information are input, first, the control point pitch calculation unit 1
3 is the starting point frequency Fjj of the tone dF from the tone accent component file 11 based on the intonation control information.
and the accent component value/(n, T) [pitch of the center of the vowel], and the control point pitch correction coefficient 9 (k, c) within the syllable from the intra-syllable pitch file 12 for each syllable.
is read out, and the control point pitch ft, h is calculated using the following equation.

However, /z, b: Pitch of the second control point in the input i-th character N: Number of moras of the input character string/(i, T): Accent component value of the i-th mora of the T-shaped accent 9 (k, c): The pitch correction coefficient of the second control point within the syllable of input character C, with 9(k, c) = 1.0 at the vowel center.

Furthermore, the number and position of spectral stationary points within a syllable are:

Determine when creating vocal tract parameter file 1. i.e. by calculating the spectral changes during the analysis of each syllable.

If the spectral change threshold is 1 dB or less and the duration threshold is 20 m5 or more, and the center of the section with a small spectral change and long duration is the spectral stationary point, then 2~ Three stationary point positions are obtained. At this time, the center of the vowel is always the position of the spectral stationary point.

Next, the interpolation calculation unit 15 interpolates between these spectral stationary points by calculating an interpolation function read from the interpolation function file 14 based on the position information of the spectral stationary points.

There are five types of position information: beginning of a sentence, within a syllable, between syllables within a bunsetsu, between bunsetsu, and during a bunsetsu, and each table of average pitch change patterns found in natural speech is created using an interpolation function. It is said that

(Effects of the Invention) As explained above, according to the present invention, pitch control within a syllable is performed at multiple positions, and a nonlinear interpolation function that simulates pitch changes in natural speech is used for interpolation. This has the effect of allowing pitch changes within a syllable or sudden changes in pitch between phrases to be expressed like natural speech.

[Brief explanation of the drawing]

Fig. 1 is a block diagram of an embodiment of the present invention, Fig. 2 is a block diagram of an intonation calculation section in an embodiment of the present invention, Fig. 3 is a block diagram of a conventional speech synthesizer, and Fig. 4 is a block diagram of a conventional speech synthesizer. FIG. 2 is an explanatory diagram of pitch change over time according to the pitch calculation method of FIG. 1...Vocal tract Harameter file, 2...Vocal tract parameter combination unit, 3...Amplifier calculation unit, 5
...Pulse train generation section, 6...White noise generation section,
7...Acoustic calculation unit, 8...D/A converter, 9...Speaker, 10...Intonation calculation unit, 1
1...Tone/accent component file, 12...Intra-syllable pitch file, 13...Control point pitch calculation unit. 14... Interpolation function file, 15... Interpolation calculation section. Patent applicant Matsushita Electric Industrial Co., Ltd.-1,'1 Figure 1 Figure 2 Figure 3 Figure 4

Claims (2)

[Claims] (1) In a speech synthesis device that vocalizes an input character string, the control point pitch between multiple spectral stationary points formed within each syllable is calculated based on the tone component and accent component of intonation control information. A speech synthesis device characterized in that: (2) In a speech synthesis device that vocalizes an input character string, it is assumed that the spaces between multiple spectral stationary points formed within each syllable are interpolated using a nonlinear interpolation function determined by the position of the spectral stationary points. Characteristic speech synthesizer.