JPH0229800A - Voice synthesizing device - Google Patents

Abstract

PURPOSE:To obtain a composite voice with high naturalness under easy control by holding the accent level of a proper noun constant in vice mode and obtaining its composite voice, and obtaining the composite voice while varying the accent level in the voice output mode of a general document. CONSTITUTION:Word information and document information indicating the proper noun symbolized by the symbol sequence input part 1 of the voice synthesizing device are inputted and a rule synthesis processing part 2 analyzes symbol information inputted through an input part 1 to convert it into a symbol information phoneme sequence and also finds accent type information. Those pieces of information are supplied to composite parameter generation parts 3 and 4, which generate parameters for voice synthesis according to a phoneme sequence and accent information from the processing part 2 and output the composite parameters to a voice synthesis part 5, thus easily performing rule synthesis.

Description

[Detailed Description of the Invention] [Objective of the Invention] (Field of Industrial Application) The present invention provides a speech synthesis system that can synthesize natural noun sequences and general sentences input as encoded code information with a high degree of naturalness. Regarding a synthesis device.

(Prior Art) Recently, an input character code string is analyzed to obtain its phonological parameter sequence and prosodic parameter sequence, and the above-mentioned human character code string is Various speech synthesis devices have been developed that generate and output synthesized speech represented by . This type of speech synthesis device using rule synthesis has an excellent advantage over conventional recording and editing systems in that it can generate arbitrary words and sentences into products with good naturalness. Therefore, along with the development of voice recognition technology, it is attracting attention as an important technology for realizing highly natural man-machine interfaces.

By the way, most of these rule synthesis technologies are being researched and developed exclusively for converting general text information into speech. For this reason, there is a problem in that it is not necessarily possible to obtain highly natural synthesized speech for applications where only proper nouns are to be output as speech, such as in various information retrieval systems.

Specifically, for example, when synthesizing a general sentence such as ``Otoshite/Kudasai'', the accent level of the word ``Kudasai'' is changed to the accent level of the word ``Otoshite'', taking into account the importance of the words that make up the sentence. By setting it lower, we are trying to obtain a synthesized voice with a high degree of naturalness. However, when sequentially synthesizing proper nouns with the same number of moras and the same accent type, if the accent level is variably controlled based on the connection of the words, it is undeniable that the unnaturalness increases.

(Problem to be Solved by the Invention) As described above, since conventional speech synthesis devices only synthesize general sentences, it is difficult to ensure naturalness when sequentially synthesizing proper nouns. There were problems such as difficulty.

The present invention has been made in consideration of these circumstances, and its purpose is to synthesize speech with a natural manner regardless of whether the input character code string is a general sentence or a proper noun. The purpose of the present invention is to provide a speech synthesis device that can perform the following tasks.

[Structure of the Invention] (Means for Solving the Problems) A speech synthesis device according to the present invention includes a means for synthesizing speech by generating a synthesis parameter that always keeps the accent level of the pitch pattern of the speech to be synthesized and output constant. , a means for synthesizing speech by generating a synthesis parameter that makes it possible to vary the accent level of the pitch pattern of the speech to be synthesized and output, and determining whether the human character code string is a general sentence or a proper noun. It is characterized by being selectively used depending on the situation.

(Function) According to the present invention, when in the speech output mode of proper nouns, the accent level is kept constant and synthesized speech is obtained, and when in the speech output mode of general sentences, the synthesized speech is obtained while varying the accent level, so that the synthesized speech is highly natural. It becomes possible to easily obtain synthesized speech under simple control.

(Example) An example of the present invention will be described below with reference to the drawings.

FIG. 1 is a schematic configuration diagram of an embodiment of the apparatus, and l is a symbol string input section for inputting word information and sentence information indicating encoded proper nouns. The rule synthesis processing section 2 analyzes the symbol string information inputted through the symbol string human power section 1, converts the input symbol string information into a phoneme sequence, obtains information on its accent type, and converts this information. are given to the synthesis parameter generation units 3 and 4, respectively. These synthesis parameter generation units 3.4 generate synthesis parameters for speech synthesis based on the phoneme sequence and accent type information given from the rule synthesis processing unit 2, and output these synthesis parameters to the speech synthesis unit 5. This is used for the regular synthesis of speech.

The synthesis parameter generation unit 3 for the proper noun utterance mode generates synthesis parameters that keep the accent level of the pitch pattern of the voice to be synthesized and output constant at all times, as shown in FIG. 2(a). In other words, the synthesis parameters are generated so that the accent level al of the first accent phrase is equal to the accent level a2 of the second accent phrase. In FIG. 2, the broken line indicates the falling tone component, and the solid line indicates the pitch pattern.

On the other hand, the synthesis parameter generation unit 4 for the sentence utterance mode generates the synthesized parameters according to the arrangement of the accent phrases, for example, as shown in FIG.
), the accent level is (al −a2
>a3) Generate variable synthesis parameters.

The mode switching section 6 generates the above-mentioned synthesis parameter generating section 3.4 according to the mode switching information indicating whether the symbol string to be speech synthesized is a string of interior angle nouns, or a general sentence.
is selectively used to control the speech synthesis. Through this switching control, a synthetic confession with a constant accent level is generated and output for proper nouns, and a synthesized sound with a variable accent level is generated and output for general sentences.

As a result, whether it is continuous speech synthesis of proper nouns such as addresses, names, and company names, or multistage synthesis of general sentences, it is possible to achieve a sufficiently high level of naturalness. It becomes possible.

Note that the present invention is not limited to the embodiments described above. For example, the number of accent phrases to be synthesized is 2 or 3.
The accent level is not limited to this, and the form of accent level variable control is not particularly limited either. It is also possible to automatically switch and control the generation form of synthesis parameters according to the result of part-of-speech analysis of a human-powered symbol string.

In addition, the present invention can be implemented with various modifications without departing from the gist thereof.

[Effects of the Invention] As explained above, according to the present invention, it is possible to provide a simple and highly practical speech synthesis device that can effectively generate and output highly natural synthesized speech through simple control. It can bring about great effects, such as being able to.

[Brief explanation of the drawing]

FIG. 1 is a schematic configuration diagram of a speech synthesis device according to an embodiment of the present invention, and FIG. 2 is a diagram schematically showing a control form of the accent level of synthesized and output speech. ,
2... Rule synthesis processing unit,... Synthesis parameter generation unit, 5... Speech synthesis unit, mode switching unit.

Claims (1)

[Claims] In a speech synthesis device that generates and outputs synthesized speech indicated by the symbol string information using an analysis result of symbol string information and a predetermined rule, the accent level of the pitch pattern of the synthesized speech is always kept constant. 1. A speech synthesis method characterized by comprising: a means for generating a synthesis parameter for voice synthesis by generating a synthesis parameter that maintains the pitch pattern; and a means for generating a synthesis parameter for voice synthesis by generating a synthesis parameter that makes it possible to vary the accent level of the pitch pattern of the voice to be synthesized and output. Device.