JPH0419799A - Voice synthesizing device - Google Patents

Abstract

PURPOSE:To enable voice synthesis of high quality by storing voice data which are high in frequency in use as variable words by sound recording and information compression, and reading voice data corresponding to an input character string out of a storage means and outputting the voice data when the voice data on a variable word corresponding to the input character string is stored. CONSTITUTION:This device is provided with storage means 1 and 2 where voice data on a fixed sentence and voice data high in frequency of use as the respective variable words are stored by sound recording and information compression respectively and a synthesizing means 4 which reads the voice data corresponding to the fixed sentence selected with an input select signal out of the storage means 1 and outputs the voice data. Then whether or not the voice data on the variable word corresponding to the input character string is stored in the storage means 2 is retrieved and when so, the corresponding voice data is read out of the storage means 2, but when not, a text is composed by composing means 6 and 7 and outputted. Consequently, the variable words which are high in frequency of use can be synthesized with good quality as well as the fixed sentence.

Description

[Detailed Description of the Invention] [Field of Industrial Application] The present invention provides a speech synthesis device that outputs a combination of synthesized speech of variable words that can be replaced in fixed sentences and synthesized speech of identified sentences that cannot be replaced. It is related to.

[Prior Art] There are three types of speech synthesis: a recording editing method, a parameter editing method, and a text synthesis method.

This recording method and parameter editing method (hereinafter referred to as the recording and editing method) have a high quality synthesized voice, but the natural voice corresponding to the voice you want to output must be recorded (compressed) in advance. It is necessary to memorize the information.

In the case of the text synthesis method, this work is necessary, and any voice can be output by simply inputting a string of characters, but the quality of the synthesized voice is low at the current technological level.

Here, first, we will realize a speech synthesis device that performs speech synthesis where the content of the output speech is a fixed phrase and some words in the fixed phrase change in various ways, such as a person's name, using only a recording/editing method. Let's think about it. In this case, the words that change in various ways as mentioned above are called variable words, and the words obtained by removing variable words from a fixed sentence are called fixed sentences.The simplest way to memorize the output audio is to write all the words that correspond to each variable word. This method involves memorizing fixed phrases, but there is a problem in that the storage capacity becomes large.

In order to solve this problem, it is conceivable to store each fixed sentence and each variable word and output the fixed sentence and variable word in combination. In this case, the naturalness of the prosody generally deteriorates, but the naturalness can be improved by combining the prosody information of the variable word with the prosody information of the fixed sentence.

The drawback of this method is that only pre-stored sounds of variable words can be output between variable words.

To further solve this problem, it is conceivable to output variable words by text synthesis. As shown in Fig. 2, the configuration of this conventional example includes a fixed sentence audio data file 1 that stores audio data of fixed sentences, and a character string signal for specifying a variable word for an output audio instruction signal inputted from the outside. The selection signal for selecting a fixed sentence and the signal distribution are shown in Part 3.
The signal distribution is performed by a recording/editing synthesis unit which reads the audio data of the fixed sentence specified by the selection signal distributed in unit 3 from the fixed sentence audio data file 1 and synthesizes it using the above-mentioned recording/editing method. 4, the above signal distribution is obtained by adding the text synthesizer 7 which takes in the character string data of the character string signal distributed in section 3 and synthesizes the text, and the audio synthesized by each of the above synthesis sections 4.7. It is composed of an adder 8 that outputs synthesized speech as a fixed phrase.

[Problems to be Solved by the Invention] In the case of the conventional example shown in Figure 2, the quality of the synthesized speech is low because the text synthesis method is used for speech synthesis of variable words, and the quality of the synthesized speech of fixed sentences and variable words is low. The difference has become large,
There is a problem in that the overall sound is unnatural.

The present invention has been made in view of the above-mentioned problems, and its purpose is to improve the speech synthesis device consisting of fixed sentences and variable words as described above, so that the speech range that can be outputted is wide.
Another object of the present invention is to provide a speech synthesis device capable of high-quality speech synthesis.

[Means for Solving the Problem] In order to achieve the above-mentioned object, the present invention according to claim 1 combines a fixed sentence whose words do not change and a variable word consisting of words that can be changed. A speech synthesis device that outputs speech of fixed sentences includes a storage means for storing speech data of fixed sentences and frequently used speech data for each variable word by pre-recording and information compression, and an input selection signal. A synthesis means reads and outputs audio data corresponding to the selected fixed sentence from the storage means, and searches and stores whether audio data of a variable word corresponding to the input character string is stored in the storage means. The text is read from the storage means and outputted, and if the text is not stored, the text is synthesized and outputted.

The invention according to claim 2 is characterized in that each voice data stored in the storage means is composed of data consisting of phonological information and tonality information, and the voice synthesis using the voice data read from the storage means is performed by rule synthesis. 6 [Operations] According to the invention described in claim 1, frequently used variable words can be synthesized with high quality in the same way as fixed sentences, and there is no difference in quality between them and fixed sentences. In addition, if the specified variable word is not registered, the quality will be degraded, but since it will be output by text synthesis, any variable word can be output, and as a result, the range of speech that can be output is wide, and the quality is high. This enables high-quality speech synthesis.

According to the invention as claimed in claim 2, the capacity of the storage means can be made very small, and the circuit for synthesis can be reduced to 2 to 91 people.
P/IXJ-仄＾吾吾し丑助RU助（φへl [Example] The present invention will be explained below with reference to an example.

Figure 1 shows the overall configuration of the embodiment, which includes a fixed sentence audio data file 1 consisting of an appropriate storage element that stores fixed sentence audio data by recording and information compression, and a frequently used variable file. Variable word audio data file 2 consisting of an appropriate storage element that stores word audio data by recording and information compression.
The output voice instruction signal inputted from the outside is divided into a character string signal for specifying a variable word and a selection signal for selecting a fixed sentence. a recording/editing synthesis section 4 which reads and compares the audio data of the fixed sentence specified by the selected selection signal from the fixed sentence audio data file 1 and synthesizes it in the above-mentioned recording/editing method; A function to search and read the voice data of the variable word corresponding to the character string of the character string signal sorted in step 3 from the voice data of the variable word voice data file 2, and if there is no corresponding voice data by search, the input character string is read out. an audio data search and synthesis method determination section 5 having a function of outputting audio data; a recording and editing synthesis section 6 that takes in the audio data read from the audio data search and synthesis method determination section 5 and synthesizes it in a recording and editing manner; A text synthesis section 7 takes in the character string data output from the data search synthesis method determination section 5 and synthesizes the text, and the speech synthesized by each synthesis section 4゜6.7 is added to synthesize a fixed sentence. It is composed of an adder 8 that outputs audio.

Note that a number is assigned to each fixed sentence audio data stored in the fixed sentence data file 1, and the selection signal is based on this number signal, so that the audio data of a predetermined fixed sentence can be read out by specifying the number. It is now possible to do so. On the other hand, the voice data of variable words is called up by a character string signal as described above.

When performing speech synthesis of a predetermined fixed sentence, an output voice instruction signal is created by arranging a number specifying a predetermined fixed sentence and a character string specifying a variable word in the order in which they are output. The supplied signal is divided into a number signal as a selection signal and a character string signal by a signal distribution unit 3, and then sent to a voice data search and synthesis method determination unit 5 or a recording/editing device. For fixed sentences, the audio data specified by the selection signal from the fixed sentence audio data file 1 is synthesized by recording and editing, and for variable words, the audio data corresponding to the character string is synthesized into variable word audio. If it is in the data file 2, it will be synthesized in a recording/editing manner, and if it is not, it will be synthesized as text.If the capacity of the variable word audio data file 2 is large enough, most of the variable words used will be stored. Therefore, the speech synthesis of variable words can be performed using a recording/editing method, and therefore there is no difference in quality between the synthesized speech of fixed sentences and variable words.

Also, in the case of variable words that are not in the variable word audio data file 2, they are output using the text synthesis method, so there will be a difference in quality. ), and the range of sounds that can be output is expanded. Furthermore, if a method of synthesizing phonological information and prosodic information as voice data by rule synthesis is adopted as a recording/editing synthesis method, the capacity of each voice data file 1.2 will become extremely small, and the circuit for synthesis will become very small. Since it can be shared with the circuit for text synthesis, the circuit configuration becomes simple.

Furthermore, regarding pitch control during speech synthesis of variable words, it is possible to memorize only the accent component (generate it during text synthesis) and superimpose that accent component on the phrase component of the variable word part generated from the surrounding fixed sentences. A synthesized sound can be obtained.

[Effects of the Invention] The invention as claimed in claim 1 provides a speech synthesis device that outputs the sound of a fixed sentence by combining a fixed sentence whose words do not change and a variable word consisting of words that can be changed. a storage means for storing the data and voice data frequently used as each variable word by recording and information compression in advance, and reading voice data corresponding to the fixed sentence selected by the input selection signal from the storage means; A synthesis means to output and a search for whether voice data of a variable word corresponding to the input character string are stored in the storage means, and if stored, read the corresponding voice data from the storage means. Since it is equipped with a synthesis means that performs text synthesis and outputs the text if it is not memorized, high-quality speech synthesis can be performed for frequently used variable words in the same way as for fixed sentences. There is no difference in quality between the synthesized speech of fixed sentences, and if the specified variable word is not registered, it will be output as text synthesis even if the quality is degraded. The present invention has the advantage that it is possible to provide a speech synthesis device that can output synthesized speech, has a wide range of outputtable speech, and is capable of high-quality speech synthesis.

In the invention as claimed in claim 2, each voice data stored in the storage means is composed of data consisting of phonetic information and tonality information, and the voice synthesis using the voice data read from the storage means is performed by rule synthesis. Since the capacity can be made very small and the circuit for synthesis can be shared with the circuit for text synthesis, the circuit configuration can be simplified and costs can be reduced.

[Brief explanation of drawings]

FIG. 1 is a circuit configuration diagram of an embodiment of the present invention, and FIG. 2 is a circuit configuration diagram of a conventional example. 1 is a fixed sentence audio data file, 2 is a variable word audio data file, 3 is a signal distribution section, 4.6 is a recording/editing synthesis section, 5 is an audio data search/synthesis method determination section, 7 is a text synthesis section, 8 is an adder. Agent Patent Attorney Ishi 1) Chief 7 August 1990?

Claims (2)

[Claims] (1) In a speech synthesis device that outputs the voice of a fixed sentence by combining a fixed sentence with no change in words and a variable word consisting of words that can be changed, the voice data of the fixed sentence and the frequency of use of each variable word are combined. a storage means for storing high quality audio data by recording and information compression in advance; a synthesizing means for reading and outputting audio data corresponding to the fixed sentence selected by the input selection signal from the storage means; A search is made to determine whether the voice data of the variable word corresponding to the character string is stored in the storage means, and if it is stored, the corresponding voice data is read out from the storage means and output, and if it is not stored, the corresponding voice data is read out from the storage means. A speech synthesis device comprising: a synthesis means for synthesizing text and outputting the synthesized text. (2) Claim 1 characterized in that each voice data stored in the storage means is composed of data consisting of phonological information and tonality information, and the voice synthesis using the voice data read from the storage means is performed by rule synthesis. The described speech synthesizer.