JPH0452479B2 - - Google Patents

Description

[Detailed Description of the Invention] (1) Technical Field of the Invention The present invention is capable of switching between a conversational voice synthesis unit that generates conversational voice using a rule synthesizer based on conversational information and a singing voice synthesis unit that generates singing voice based on musical score information. The present invention relates to a conversational voice and singing voice generator which is added so that the following is achieved.

(2) Conventional technology and problems Conventionally, conversational information including phonetic character strings and accent patterns is input to a rule synthesizer, and the system matches the conversational information by comparing it with a dictionary that stores phoneme data for each phonetic character. Spectral envelope information representing phonemes, pitch information representing accents and intonations, that is, sound frequency information represented by the number of sampling pitches, and amplitude information representing sound intensity are output and sent to a speech synthesizer to generate conversation. Various types of sound generating devices have been proposed and put into practical use.
However, I have not found anything that generates singing voices by combining or switching with this.

In this case, various uses can be considered, such as reciting the lyrics as a conversational voice and generating a switching voice, or adding music to normal conversational voice and generating it in an operatic manner.

(3) Object of the Invention The object of the present invention is to provide a conversational/singing voice generating device which has an output switching function between a conversational voice and a singing voice, is capable of outputting both voices, and is capable of generating a more natural singing voice.

(4) Structure of the Invention In order to achieve the above object, the conversational voice singing voice generating device of the present invention inputs conversational information to a rule synthesizer, and generates spectral envelope information representing phonemes corresponding to the information, accent, and intonation. a conversational speech synthesis section that outputs pitch information representing the pitch and amplitude information representing the strength of the sound and sends it to the speech synthesizer to generate conversational speech; a conversational speech synthesis section that inputs musical score information and stores pitch information corresponding to the scale; a changeover switch that sends pitch information read out from the section and synthesized for each scale, and amplitude information read out from the amplitude information section corresponding to the strength of the sound to the speech synthesizer, and switches the tone color and time length of each scale, respectively. and a singing voice synthesizer configured to output only the spectral envelope information by sending the signal to the rule synthesizer via the above-mentioned rule synthesizer. The feature is that the singing voice is switched and generated.

(5) Embodiment of the invention FIG. 1 is an explanatory diagram of the configuration of an embodiment of the invention.
In the figure, conversation information 1- is stored in a storage unit 1 and consists of phonetic character strings indicating lyrics, etc., and accent types during conversation.
1, and musical score information 1-2 indicating the timbre, scale, and strength to be given to it. In the case of conversational speech, this conversational information 1-1 is input to the rule synthesizer 2, which generates spectral envelope information representing the phoneme corresponding to the aforementioned conversational information, pitch information representing accent and intonation, and sound intensity. It outputs amplitude information representing , and sends it to the voice synthesizer 8 to generate a synthesized conversation voice signal, which is emitted from the speaker 9 .

In order to add music to this conversation voice, musical score information 1-2 consisting of timbre, scale, and sound intensity is read out from the storage unit 1, and the scale is input into the pitch conversion unit 3.
When the pitch values previously stored in the pitch information storage section 4 are read out and synthesized by, for example, 8KHz sampling, the pitch period is 125 μs, and each scale can be represented by a pitch value that is a multiple of this. For example, "hua"
Since the frequency of is 349Hz, the pitch value is 23, which can be expressed as an approximate value of 1/125μs x 23 = 348Hz. The pitch values thus converted corresponding to each scale from the musical score information 1-2 are sent to the pitch synthesis section 6 and synthesized, thereby smoothing the transition waveform of each scale. Figures 2a and 2b show this state; Figure 2a shows the waveform before pitch synthesis for each scale C, Re, and E, and Figure 2b shows the waveform after pitch synthesis.

A vibrato section 7 adds vibrato to the output of the pitch synthesis section 6. Figures 3a and 3b show this state, with figure a showing the waveform before vibrato, and figure b showing the waveform after vibrato. This allows the generation of sounds that are close to natural. This pitch information is
The intensity of the sound in the musical score information is input to the amplitude information section 5 and sent to the speech synthesizer 8 together with the read amplitude information.

On the other hand, a switch 10 switches between the timbre in the musical score information, that is, the information on the ratio of harmonics to the fundamental wave, and the time length of each scale from the pitch conversion section 3, that is, the time length that differs depending on half note, quarter note, etc.
₁ and 10 ₂ to the rule synthesizer 2.

When these switches 10 ₁ and 10 ₂ are closed, the spectral envelope information created by the conversation information 1-1 is added with the timbre and time length from the musical score information, and only the spectral envelope information is output. The voice is then sent to the voice synthesizer 8, where the pitch information and amplitude information described above are combined into a singing voice, and the voice is emitted from the speaker 9.

As described above, when the switches 10 ₁ and 10 ₂ are opened, only conversation information is generated from the speech synthesizer 8, and when they are closed, a singing voice is generated based on the conversation information (lyrics) and set to music using musical score information. .

In this embodiment, a pitch synthesis section 6 and a vibrato section 7 are provided in order to make the singing voice more natural, but other arbitrary effects may be added.

(6) Effects of the Invention As explained above, according to the present invention, a singing voice synthesis section for generating a singing voice based on musical score information is added to a conversation voice synthesis section that generates a conversation voice by inputting conversation information into a rule synthesizer. It is added so that it can be switched.

As a result, it is possible to output only conversation information, and it is also possible to output singing voice based on this conversation information (lyrics), and by applying it to fields such as lyric writing, composition, and exercises, it can be used for various purposes. Conceivable.

[Brief explanation of drawings]

FIG. 1 is a configuration explanatory diagram of an embodiment of the present invention, FIGS. 2 a, b, and 3 a, b are characteristic diagrams showing the operation of the main parts of the embodiment of FIG. Storage part, 1-
1 is conversation information, 1-2 is musical score information, 2 is a rule synthesizer, 3 is a pitch conversion unit, 4 is a pitch information storage unit,
5 is an amplitude information section, 6 is a pitch synthesis section, 7 is a vibrato section, 8 is a speech synthesizer, 9 is a speaker, 10 ₁ ,
10 ₂ indicates a changeover switch.

Claims (1)

[Scope of Claims] 1 Conversation information is input to a rule synthesizer, and spectral envelope information representing phonemes corresponding to the information, pitch information representing accent and intonation, and amplitude information representing sound strength are extracted. a conversational speech synthesizer that outputs and sends it to a speech synthesizer to generate conversational speech, and a conversational speech synthesizer that inputs musical score information, reads out pitch information corresponding to a scale from a pitch information storage section, and synthesizes pitch information and sound intensity for each scale. The amplitude information read from the amplitude information section corresponding to the scale is sent to the speech synthesizer, and the timbre and the time length of each scale are also sent to the rule synthesizer via a changeover switch that respectively switches the timbre and time length of each scale. and a singing voice synthesizing section configured to output a singing voice, and a conversational voice singing voice generating device characterized in that when both the changeover switches are simultaneously opened, only the conversational voice is generated, and when the two changeover switches are simultaneously closed, the conversational voice and the singing voice with music are switched and generated. .