JPH0549998B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Application of the Invention] The present invention relates to a speech synthesis device, and particularly to a speech synthesis device suitable for outputting speech expressing complex content in an easy-to-hear manner.

[Background of the invention]

An example of conventional speech synthesis that is easy to hear is the Japanese patent publication
It is described in Publication No. 56-29297. However, when converting complex sentences, such as sentences with annotations between parentheses ( When composing ``hitonokoeniha samazamanashiyurui, hirakikatsuko tatoeva otokoto onnano koe tojikatsuko, gaaru,'' the sentence becomes complex and not enough to make out the meaning of the sentence.

[Purpose of the invention]

An object of the present invention is to provide a speech synthesis device that converts such complex written sentences into speech in an easily audible format.

[Summary of the Invention] The present invention provides a method for adding predetermined symbols (for example,
Input an input symbol string that includes a sentence with an explanation surrounded by brackets, analyze the input symbol string, and select a predetermined input symbol string from the input symbol string that ignores the input symbol string in the explanation part surrounded by the specified symbols. First speech synthesis information including information indicating the position of the explanatory part surrounded by predetermined symbols is generated using a rule synthesis method, and from the input symbol string of the explanatory part surrounded by the ignored predetermined symbols, a control means for controlling the generation of the second speech synthesis information using a predetermined rule synthesis method; a storage means for storing the first speech synthesis information and the second speech synthesis information, respectively; The first speech synthesis information stored in the storage means is read out, and the first speech synthesis information is read out using the speech segment of the first type (for example, the first speaker).
If information indicating the position of the explanatory part surrounded by predetermined symbols is read out when reading the first speech synthesis information, the second speech synthesis information stored in the storage means is The present invention is characterized by comprising a synthesizing means that switches to reading the speech synthesis information and synthesizes speech of the second type using speech segments of the second type (for example, from a second speaker).

The main text of the sentence to be converted into audio and the part inside the cutlet that explains part or all of it are grammatically independent. Therefore, in the present invention, the above objective is achieved by simulating the format in which the main text is uttered by a first speaker and the parts in the cutout are explained by a second speaker. It is something. In this way, by listening to the voice of the first speaker, the main text can be easily heard, and the explanation can be heard as an explanation using the voice of the second speaker.

[Embodiments of the invention]

Hereinafter, one embodiment of the present invention will be described with reference to the drawings, taking the first object as an example. In FIG. 1, a symbol string representing an input sentence is input to a control unit 1. As shown in FIG. Control unit 1
Then, we analyze the input symbol string, ignore the parts enclosed in brackets, determine the accent type of each word using a well-known rule synthesis method, estimate the duration of the clause corresponding to each character, and calculate the pitch. The pitch pattern of the pitch of the first speaker's voice is extracted from the portion A of the pattern memory 3, compiled into control information for speech synthesis using the segment editing method, and written into the first portion of the control information buffer 6. This information also includes the information that it is the voice of the first speaker, and a mark is also attached so that the location of the ignored katsuko can be determined. Next, the control unit 1 takes out the sentence inside Katsuko, takes out the pitch pattern information of the second speaker from B of the pitch pattern memory 3, creates control information using the same procedure, and stores the pitch pattern information in the control information buffer 6.
Write in the second part of. This control information also includes information indicating that the voice is the voice of the second speaker. When the creation of the control information is completed in this way, the control section 1 instructs the editing and synthesis section 7 to synthesize speech. The editing/synthesizing unit 7 is configured to control the first
The information on the part is extracted, and the switch 5 is controlled in accordance with the command to select the segment part 4-1 of the first speaker in the speech segment memory 4, and the speech is successively synthesized.
When a mark indicating the position of the cutter is input from the data in the first part of the control information buffer 6, the data jumps to the second part, and accordingly the switch 5 switches to the data for the second speaker and stores the data in the memory. The second speaker's segment is extracted from 4-2, and the sentences in the cutlet are successively synthesized, and when finished, the first speaker's data is restored and the synthesis continues. The speech synthesis method using speech segments is described in, for example, the Journal of the Institute of Electronics and Communication Engineers Vol. 58-D No.
9p522 "Method for creating speech segments in a monosyllabic editing type speech synthesis method using speech segments", etc. Also regarding rule synthesis, see the Acoustical Society of Japan speech study group material "Monosyllabic editing type using speech segments", etc. Voice response method”
This is a well-known technique as shown in numerous materials such as .

In this embodiment, the phoneme one-sided method is used in the synthesis section, but it goes without saying that other synthesis methods such as the PARCOR method may also be used.

In the above embodiment, the cutter is placed in one place, but it goes without saying that it is easy to extend the cutter to multiple places.
Furthermore, the system is not limited to two types of voices, and can easily be expanded to include many more types. If two types are used, a male voice and a female voice would be appropriate. Naturally, it is also possible to change the output level of one side, or to change the sound by making it softer.

In order to achieve the second objective, the basic configuration is the same as that shown in Figure 1, and a code is added to the input signal to distinguish whether it is control information or data (a distinction is generally made in conventional terminals as well). This can be easily realized by switching and selecting the audio using the control unit 1 according to the code.

[Effects of the Invention] As explained above, according to the present invention, in character expression, multiple types of information that can be easily distinguished by brackets, positions, colors, etc. can be classified by using differences in sounds to classify information types. By doing this, you can easily hear, categorize, and understand the words.

[Brief explanation of the drawing]

FIG. 1 is a diagram illustrating an embodiment of the present invention. 1...Control unit, 2...Input signal line, 3...Pitch pattern memory, 4...Speech unit memory, 5...
switch, 6... control information buffer, 7... editing synthesis section, 8... synthesis output signal line.

Claims (1)

[Claims] 1. Inputting an input symbol string including a sentence with an explanation surrounded by predetermined symbols in an input sentence, analyzing the input symbol string, and analyzing the explanatory part surrounded by the predetermined symbols. Using a predetermined rule synthesis method, first speech synthesis information including information indicating the position of the explanatory part surrounded by the predetermined symbols is generated from the input symbol string in which the input symbol string is ignored. a control means for controlling to generate second speech synthesis information using a predetermined rule synthesis method from the input symbol string of the explanatory part surrounded by the predetermined symbols; and the first speech synthesis information. and the second speech synthesis information; and a storage means for reading out the first speech synthesis information stored in the storage means and using the first type of speech segment to produce the first speech synthesis information. If information indicating the position of the explanatory part surrounded by the predetermined symbols is read out when reading the first speech synthesis information, the information stored in the storage means Said second
A speech synthesis device comprising: synthesis means for switching to reading speech synthesis information and synthesizing speech of a second type using speech units of a second type. 2. The speech synthesis device according to claim 1, wherein the explanatory portion surrounded by the predetermined symbols is an explanatory portion surrounded by brackets. 3. The speech synthesis device according to claim 1, wherein the first type and the second type are a first speaker and a second speaker.