JPH01112297A - Voice synthesizer - Google Patents

Abstract

PURPOSE: To add an effective sound to a synthetic voice by generating an effective sound from information related to an effective sound added to a character string and inserting the generated effective sound into a synthetic voice. CONSTITUTION: An effective sound symbol processing part 4 extracts information related to an effective sound included in an inputted character string and outputs its time length to an effective sound generating part 11 and a mixing part 12 mixes the effective sound obtained by the generating part 11 with a voice obtained by a voice synthesizing part 10 and a synthetic voice with the effective sound is obtained from a synthetic voice output terminal 13. Thus the information related to the effective sound added to the inputted character string or symbol string is extracted and the effective sound is generated in accordance with the information and inserted or mixed into/with the synthetic voice. Consequently the effective sound can be added to the synthetic voice.

Description

[Detailed description of the invention] Industrial applications The present invention relates to a speech synthesis device.

BACKGROUND ART Conventional speech synthesis devices include, for example, Furui: Digital Speech Processing Special p., I46 (Tokai University Press 1985).
is shown. FIG. 4 is a block diagram showing the configuration of this conventional speech synthesis device. 1 is a character string input terminal where a sentence containing kanji and kana is input.

2 is a text processing unit, which uses a dictionary 3 to divide input sentences containing kanji and kana into words and adjuncts, and calculates the readings and adjuncts of each.
An accent type is assigned. 5 is the prosodic control section for each word,
Determining the accent type of a clause from the accent type of an attached word,
Decisions such as pause position and intonation of the entire sentence are made according to rules. 6 is the phonological control unit, which processes the pronunciation obtained by the text analysis unit to make vowels devoiced and nasalized, and to process cases where the kana notation and the vocalization differ (the ``ha'' in ``watashi wa'' is
), etc., to obtain the phonetic notation. 7
is a synthesis parameter creation unit that obtains a synthesis parameter sequence necessary for actual synthesis from prosody information and phonetic notation information. Synthesis parameters are the fundamental frequency that determines the pitch of the sound, the amplitude that determines the loudness of the sound, the vocal tract descriptive parameters (PARCOR coefficient, formant frequency, etc.) that determines the state of the vocal tract, and the voiced/unvoiced judgment that determines the state of the vocal cords. There are flags, etc., and there are various
It is stored in vertical table 8. Reference numeral 9 denotes a parameter interpolation unit that interpolates the synthesis parameter values obtained in step 7 to obtain a synthesis parameter sequence at fixed time intervals. 10 is a speech synthesis section; 9
Synthesize the actual voice from the synthesis parameter sequence obtained in
It is obtained at the synthesized speech output terminal 13.

Problems that the invention aims to solve When reading out a text expressed as a string of characters or symbols, it is possible to insert natural sounds (such as the sound of waves), animal sounds, or musical instrument sounds as sound effects, or use synthesized speech. (
mixing). Conventional synthesis devices do not take such sound effect insertion into consideration, and can only synthesize the desired sound effect as voice by writing it in text. However, the synthesis method suitable for speech is
Many of them model the vocal tract and vocal cords, making them unsuitable for synthesizing sound effects, and their quality is not sufficient.

Additionally, there are limits to the ability to write sound effects as onomatopoeic words. If you simply want to insert a sound effect between voices, it is possible to achieve this by using the onomatopoeic notation of the sound effect described above, but when it comes to overlapping synthesized voices and sound effects, it is impossible to achieve this with conventional synthesis equipment. .

The present invention has been made in view of this point, and includes a method of including information specifying sound effects in an input character string or symbol string, generating sound effects according to that information, and inserting or mixing them into synthesized speech. The purpose is to add sound effects to synthesized speech.

Means for Solving the Problems The present invention provides a text input means for inputting a character string or a symbol string (hereinafter referred to as text), and a text input means for analyzing the text inputted from the text input means to provide information on pronunciation and sound effects. a text analysis means for obtaining information; a synthesis parameter creation means for creating synthesis parameters according to the analysis results obtained by the text analysis means; a speech synthesis means for synthesizing speech from the synthesis parameters; and from information regarding the sound effects. This is a speech synthesis device characterized by having a sound effect generation means for generating sound effects.

Effect of the present invention With the above-described configuration, the present invention extracts information regarding sound effects added to an input character string or symbol string,
By generating sound effects according to the information and inserting or mixing them into the synthesized speech, it is possible to add sound effects to the synthesized speech.

Embodiment FIG. 1 is a block diagram showing the configuration of a speech synthesizer according to an embodiment of the present invention. ■: A sentence containing kanji and kana is input at the character string input terminal. Reference numeral 2 denotes a text processing unit, which divides the inputted kanji-kana-mixed text into words and adjunct words using a dictionary 3, and assigns readings and accent types to each word. 4 is a sound effect symbol processing unit that extracts information regarding sound effects included in the input character string. 5 is a prosody control unit which determines the accent type of each word and adjunct, the accent type of the clause, the pause position, the intonation of the entire sentence, etc., according to rules. Reference numeral 6 denotes a phoneme control unit which performs processes such as devoicing and nasalization of vowels and processing when the kana notation differs from the uttered voice to obtain a phonetic notation for the pronunciation obtained by the text analysis unit. 7
is a synthesis parameter creation unit that obtains a synthesis parameter sequence necessary for actual synthesis from prosody information and phonetic notation information. When a sound effect and a synthesized voice are overlapped, it is necessary to create a sound effect for the time length of the synthesized voice, so the time length is output to the sound effect generation section 11. The synthesis parameters are stored in a parameter table 8. Reference numeral 9 denotes a parameter interpolation unit that interpolates the synthesis parameter values obtained in step 8 to obtain synthesis parameters at fixed time intervals. Reference numeral 10 denotes a speech synthesis unit, which synthesizes actual speech from the synthesis parameter sequence obtained in step 9. Reference numeral 11 denotes a sound effect generation unit which generates sound effects according to the information regarding the sound effects obtained in step 4. Details of the operation will be described later. 1
2 is a mixing section which mixes the sound effect obtained in step 11 and the voice obtained in step 10 to obtain a synthesized sound with sound effects at a synthesized speech output terminal 13.

FIG. 2 is a block diagram showing the configuration of the sound effect creation section 11. As shown in FIG. lla is a sound effect information input terminal, into which information regarding the type and duration of the sound effect is input.

Reference numeral 11b denotes a sound effect data reading unit which calculates the address where the necessary sound effect is stored and reads out the data for the necessary time. When the time length of the stored sound effect is shorter than the required time length, the read data is used repeatedly. llc is a sound effect data storage unit in which sound effect data is compressed to 8 bits using a sampling frequency of 10 kHz and tt-1awPcM and stored therein. In this example, it is assumed that there are 8 types of sound effects, and that approximately 0.8 seconds worth of data is stored for each sound effect (total 64 types).
). lid is μm1aw in the decoding section
PcM compressed sound effect data is decoded and converted into 12-bit data. The conversion formula etc. are already known. lid is the amplitude processing section, and the rise of the sound effect,
Performs amplitude processing of the falling portion (start portion, end portion). FIG. 3 shows the function f (t) used for amplitude processing. The horizontal axis is time (msec) and the vertical axis is the function value. Amplitude processing is performed according to the following equation.

At rise x(t)-f(t)*x(t) o<t
≦100 falling x(t)-f(T-t)*x(t
) T-100<t≦Tt: Time (msec) T: Time length x(t): Sound effect data f(t): Amplitude control function The sound effect data obtained in this way is output to the mixing section 12.

Next, details of each process will be explained using an example. “■
This is the sound of waves. Consider a case where the sentence ``'' is input. Text analysis 2 divides the input text into the following parts to obtain information on accent type and pronunciation. Here, ■ is a symbol that indicates that the sound of waves is overlaid on the synthesized voice as a sound effect. For the sake of explanation, we have used the symbol O for convenience, but any symbol may be used; if you use an unused code with a blank or null code assigned to the JIS code or Shift JIS code, the presence of the sound effect symbol will be detected in the input string. does not appear. In the text analysis section, this ■ symbol is not processed and is sent to the sound effect symbol processing section.

(Input text) “This is the sound of waves.” (Word division)
This is the sound of waves.

(Reading) Kore Ganami
No Odo Death (accent type)/A'B'
b (Part of speech) Pronoun Case auxiliary name Case auxiliary pronoun Here “gaJ
The accent types given to different words such as ``no'', ``desu'', etc. are those listed in the NHK Accent Dictionary/Explanatory Appendix (Japan Broadcasting Publishing Co., Ltd., 1985), and are those that combine with independent words. This figure shows the combining accent types used when constructing clauses. Furthermore, the accent type of each word indicates the pitch of the sound. 4. The sound effect symbol processing section determines the necessary sound effects and the insertion position of the sound effects. In this case, the sound effect shall be overlaid until the end of the sentence where ``su'' was at the beginning of the sentence. Next, the prosody processing unit 5 determines the pause position, the accent type of the phrase, and the intonation of the entire sentence. Further, the phoneme processing section 6 obtains the actual phonetic notation. The outputs of the prosody processing section and the phoneme processing section are shown below. "vinegar"
The O mark on the shoulder of = 9- indicates that the "s" is devoiced.

(phonetic notation) Korega Naminootodes (sentence control) to (, fs) r\to (
Actual synthesis parameters are obtained in the synthesis parameter creation section 6 according to the prosodic information and phonetic information obtained above n (Phrase accent)/"''n. In this example,
The phoneme (C, V) is used as a synthesized segment unit, and the formant frequency is used as a characteristic description parameter of the vocal tract. That is, for each phoneme shown below, parameters such as formant frequency, bandwidth, fundamental frequency, amplitude, voiced/unvoiced determination, and duration of each phoneme are created with reference to parameter table 8. These parameter values are changed using rules in consideration of the phonetic environment and the like.

(Fragment notation) KOREG 8 (Boz) NAMINOO
The TODESU parameter interpolation unit 9 performs linear interpolation of the parameters for each phoneme thus obtained to obtain a parameter sequence every 5 m5ec, which is synthesized by the synthesis unit 10 using a formant type synthesizer.

= 10- Let the audio waveform obtained by the synthesis unit 10 be y(t).

Further, the sound effect generating section 11 generates the sound effect x(t) according to the procedure already described. Mixing section 1
2, the output waveform z(t) is obtained using the following equation using y(t) and x(t) as inputs.

z(t) −w * x(t) + (1−w) *
y(t) where w −0,3 Here, W is a weighting coefficient and indicates the ratio of the volume of the sound effect and the voice.

As described above, according to this embodiment, a sound effect symbol processing section is provided to extract information regarding sound effects from an input character string, create sound effects according to the information, and insert or mix them into synthesized speech. , it is possible to add sound effects to synthesized speech.

In the embodiment, the synthesis unit is the phoneme (C1) and the formant synthesis method is used, but the present invention is not limited to this. In addition, as a method for creating sound effects, μm]awP
Although CM was used, this does not limit the present invention in any way.

As described in detail, according to the present invention, a sound effect is generated from information regarding a sound effect added to a character string, and the sound effect is added to the synthesized speech by inserting it into the synthesized speech or mixing it. can be added, and its practical value is great.

[Brief explanation of the drawing]

FIG. 1 is a block diagram of a speech synthesis device according to an embodiment of the present invention, FIG. 2 is a block diagram showing the configuration of a sound effect generation section, FIG. 3 is an amplitude control characteristic diagram in the sound effect generation section, and FIG. The figure is a block diagram of a conventional speech synthesis device. 1...Character string input end, 2...Text analysis section, 3.
... Dictionary, 4... Sound effect symbol processing section, 5... Prosody control section, 6... Phoneme control section, 7... Synthesis parameter control section, 8... Parameter table, 9... Parameter interpolation unit, 10...Speech synthesis unit, 11...Sound effect generation unit, 12...Mixing unit, 13...Synthesized voice output terminal, 11a...Sound effect information input terminal, llb...・Sound effect data reading unit, llc...data storage unit, l
id...decoding section, 11e...amplitude processing section, ll
f...Sound effect output end. Name of agent: Patent attorney Toshio Nakao and 1 other person = 13-

Claims (1)

[Claims] a text input means for inputting a character string or a symbol string (hereinafter referred to as text); a text analysis means for analyzing the text input from the text input means to obtain information regarding readings and information regarding sound effects; and the text analysis means. a synthesis parameter creation means for creating a synthesis parameter according to the analysis result obtained in the above, a speech synthesis means for synthesizing a voice from the synthesis parameter, and a sound effect generation means for generating a sound effect from the information regarding the sound effect. A speech synthesis device characterized by: