JPH01186996A - Sentence intonation processing method for voice synthesizing device - Google Patents

Abstract

PURPOSE:To obtain a synthesized voice having an intonation approximating a human voice by determining a breath position in accordance with boundaries of clauses and phrases of a text and adding the expiratory intonation determined based on the breath position to a fundamental intonation and word accents to obtain a sentence intonation. CONSTITUTION:A sentence analyzer 5b refers to a grammar dictionary 6 and a word dictionary 7 to perform the sentence analysis of a Japanese-language input text such as clause division and phrase division and obtains boundaries of clauses and phrases and the number of syllables of each boundary with respect to the sentence divided to clauses and phrases and determines a breath position in the sentence in accordance with these boundaries and numbers of syllables. An expiratory intonation which has a logarithmic characteristic having the breath position and the number of syllables as the start point and the end point respectively is generated and is superposed on the fundamental intonation and word accents and a pause or the like is added to obtain metrical information of one sentence. Thus, a natural voice is synthesized.

Description

DETAILED DESCRIPTION OF THE INVENTION A. Field of Industrial Application The present invention relates to a speech synthesis device using a rule synthesis method, and more particularly to a method for processing sentence intonation.

B8 Summary of the Invention The present invention, in obtaining prosodic information of a text from sentence analysis, determines the breathing position from the boundaries of the text into clauses and phrases, and uses the exhalation intonation determined from this breathing position as the basic intonation and word accent. By adding this to the sentence intonation, it is possible to obtain a synthesized voice with an intonation close to that of a human voice.

C1 A speech synthesis device based on the conventional technical rule synthesis method has, for example, the configuration shown in FIG. The text analysis unit 1 performs text analysis on a character string of a Japanese input text using a dictionary 1a and a text analysis device 1b. In addition to a dictionary for converting the pronunciation of words, the dictionary 1a has a Japanese grammar dictionary for dividing words into clauses and phrases, and also has a dictionary of rules for word accents and basic intonation. The text analysis device 1b converts the input text into a phoneme symbol string of phonemes or syllables by referring to the dictionary 1a, and also generates prosodic information such as word accents and basic intonation.

The speech synthesis rule section 2 is composed of a file 2a and a parameter generation device 2b. The file 2a stores feature parameters of phoneme units, their connection rules, and prosodic information control rules. The parameter generation device 2b generates a control parameter sequence in which feature parameters for phoneme information are linked together with information such as connection time, and also generates a sound source pattern sequence in which pitch, energy, and intonation processing of the sound source is processed based on prosody information.

The speech generation section 3 is composed of a sound source generation device 3a, a speech synthesis digital filter 3b, and a speech converter 3c. The sound source generation device 3a generates sound source signals such as pitch and energy according to the sound source pattern sequence. In the digital filter 3b, the parameters of the Percoll coefficient, transfer function, or formant frequency are changed according to the control parameter string, and a synthesized speech data string is obtained as a response output to the sound source signal with these parameters. The audio converter 3c is a filter 3b
The output of the converter is converted into an analog signal to obtain an audio waveform, and synthesized audio is output by an electro-sound conversion means such as a speaker.

In the speech synthesis device described above, the prosody information is the basic intonation for the text data.

A single sentence intonation is created by combining word accent, stress, pause, duration, etc. In this process of creating sentence intonation, the Japanese input text is divided into phrases, phrases, and sentences by syntactic analysis with reference to Dictionary 1a, and words are divided into series by morpheme (the smallest linguistically meaningful unit) classification. Convert to a phonetic string with word boundaries and word boundaries. For this phonetic sequence information, basic intonation determined from the number of clauses (mora) for each punctuation mark, accent for each word, pause for each phrase, stress in words determined from prefixes and suffixes, etc., and phoneme units. The duration etc. of the period are determined.

D0 Problems to be Solved by the Invention In conventional sentence intonation processing methods, the pitch frequency in which basic intonation and word accent are superimposed becomes dominant, which tends to result in monotonous sentence intonation, which is inferior in quality to human speech. It was easy. In other words, due to the nature of the human voice's generator, its amplitude, pitch, and formant frequency change significantly over time, making it difficult to obtain synthesized speech that closely resembles the natural voice using only conventional basic intonation and word accents. .

An object of the present invention is to provide a sentence intonation processing method that eliminates the monotony of sentence intonation and brings it closer to the natural voice.

E2 Means and Effects for Solving Problems In order to achieve the above object, the present invention determines breathing positions from the boundaries of bunsetsu divisions and phrase divisions of Japanese input text, and calculates logarithmic characteristics from these six breathing positions. By determining the exhalation intonation that decreases in the human voice and adding this exhalation intonation, basic intonation, and word accent to obtain the sentence intonation, the sentence intonation close to the real voice is obtained by adding the exhalation intonation included in the human voice.

F. Embodiment FIG. 1 is a block diagram of a text analysis section showing an embodiment of the present invention. The dictionary 5a includes a Japanese grammar dictionary 6 and a word dictionary 7 (accent, connection time), as in the past. The text analysis device 5b refers to the grammar dictionary 6 and the word dictionary 7 to convert the Japanese input text into pronunciations of the words, as in the past.

Analyzes sentences such as bunsetsu divisions, phrase divisions, word accent positions, etc. The sentence analysis device 5b calculates the phrase boundaries and phrase boundaries as well as the number of syllables for each boundary, and determines the breathing position in the sentence from these boundaries and the number of syllables. Then, the sentence analysis device 5b generates an exhalation intonation with a logarithmic characteristic starting from the breath position and ending at the number of syllables, superimposing this exhalation intonation on the basic intonation and word accent, and further adding pauses, etc. Obtain prosodic information of one sentence. Note that the creation of phoneme information is performed in the same manner as before.

FIG. 2 shows how sentence intonation is created in the sentence analysis device 5b. When the Japanese input text is ``The cherry blossoms at school bloomed beautifully.'', the basic intonation decreases in pitch from the rising point in a logarithmic manner according to the number of syllables, and the exhalation intonation rises at the break of phrases and decreases in pitch according to the number of syllables. The word accent is added to these logarithmic characteristics with a time constant, and further rounding is performed by filter processing to create one sentence intonation.

Therefore, to determine sentence intonation, the exhalation intonation is superimposed at the breath position determined from the number of syllables and the boundaries caused by bunsetsu and phrase breaks, and changes in pitch frequency similar to the breathing movement when a person speaks. It is possible to obtain a synthesized voice that is close to the real voice.

Note that when a human-written text becomes a long sentence, a similar exhalation intonation is added at the boundary between bunsetsu and phrase breaks, but the basic intonation is handled by dividing it into the subject and predicate and giving them to each. .

G1 Effects of the Invention As described above, according to the present invention, since sentence intonation including exhalation intonation is obtained, it is possible to generate synthetic speech in which the sentence intonation has an intonation change due to exhalation contained in a human voice, It will be possible to synthesize natural speech.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing an embodiment of the present invention, FIG. 2 is a diagram illustrating intonation waveforms of the embodiment, and FIG. 3 is a block diagram of a speech synthesis apparatus using a rule synthesis method. 5a... Dictionary, 5b... Sentence analysis device, 6... Grammar dictionary, 7... Word dictionary. 5a...Dictionary 5b...Sentence analysis device 6...Grammar dictionary 7...Word dictionary

Claims (1)

[Claims] (1) Obtain phonological information and prosody information from the Japanese input text, obtain the control parameters of the digital filter corresponding to the phonological information and the sound source pattern of the sound source generator corresponding to the prosody information, and generate the audio signal corresponding to the text. In the speech synthesis device using the rule synthesis method, the breathing intonation is determined from the boundaries of bunsetsu divisions and phrase divisions of the Japanese input text, and exhalation intonation that decreases with a logarithmic characteristic is determined from each breathation position. A sentence intonation processing method for a speech synthesizer, characterized in that sentence intonation is obtained by adding intonation, basic intonation, and word accent.