JPH0313597B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION (1) Technical Field of the Invention The present invention relates to a speech synthesis device that synthesizes speech using rules from symbol strings such as kana character strings, and particularly relates to an accent-type speech synthesis device.

(2) Background of the technology Speech synthesis technology is expected to become an increasingly important technology in the future as a more advanced man-machine interface. For example, this includes voice input and voice output to an information processing device, or key input and voice output. This speech output must be as natural as possible and must be reliably recognized by humans, and accent is also an important element for recognition.

(3) Prior Art and Problems In conventional speech synthesis, accent type generation is limited to either no accent or automatic accent type generation using an accent dictionary or rules. If there is no accent, the speech synthesizer will be simpler, but the speech output will be extremely difficult to hear. On the other hand, in the case of automatic accent type generation using an accent dictionary or rules, it is not practical to have a huge amount of accent dictionaries in order to generate accent types for all words. It is impossible to generate by Therefore, in many cases, it is necessary to either limit the word category (for example, to personal names, etc.) and have an accent dictionary for that category, or to generate accent types based on rules for that category. However, when such a category is limited, if a word outside the category is input, the wrong accent type may be given, and the meaning may be confused.

(4) Purpose of the Invention The purpose of the present invention is to reduce the confusion of meanings and unnaturalness even when a word outside the word category is input in automatic accent generation with a limited word category as described above. It is something to do.

(5) Structure of the Invention The present invention provides a speech synthesis device that synthesizes speech from a symbol string using rules, which comprises: a fixed accent type generator that generates a fixed accent type for the symbol string; an automatic accent type generation unit that generates an automatic accent type for the symbol string; a mode switching unit that generates either the fixed accent type or the automatic accent type for the symbol string based on a mode switching signal; a rule synthesis processing section that rulesically synthesizes either the fixed accent type or the automatic accent type generated by the unit and the symbol string; and a rule synthesis processing section that converts the speech parameters output from the rule synthesis processing section into a speech signal. This is a speech synthesis device characterized by having: a means for outputting speech using a voice.

(6) Embodiment of the invention FIG. 1 is a diagram showing an embodiment of the invention. The symbol string first enters a symbol string buffer 1, and then enters a rule synthesis processing section 2 and a mode switch 3. The mode switch 3 instructs the fixed accent type generator 4 to generate a fixed accent type by the mode switching signal, or transfers the symbol string to the automatic accent type generator 5.
and generates an accent type that matches the symbol string according to rules. These accent types enter the rule synthesis processing section 2, where they are processed together with symbol strings to generate speech parameters. The audio parameters are input to the audio synthesis LSI, where they are converted into audio signals and output. In addition to the above method, the automatic accent type generator may refer to an accent dictionary.

Note that the mode switching signal can be extracted from the conversation step in a conversational man-machine interface. Further, as the fixed accent type, a so-called 0 type without an accent is preferable in relatively many cases, but it is not limited to the 0 type.

(7) Effects of the Invention According to the present invention, a correct accent can be given to a symbol string within a limited category using an automatic accent type generator based on rules or a relatively simple automatic accent type generator having a small accent dictionary. ,
For symbol strings outside of the limited range, misinterpretation due to incorrect accents can be reduced.

[Brief explanation of the drawing]

FIG. 1 is a diagram showing one embodiment of the present invention, 1
The names of ~6 are as written in the figure.

Claims (1)

[Scope of Claims] 1. A speech synthesis device that synthesizes speech using rules from a symbol string, comprising: a fixed accent type generator that generates a fixed accent type for the symbol string; an automatic accent type generator that generates an automatic accent type; a mode switching unit that generates either the fixed accent type or the automatic accent type for the symbol string based on a mode switching signal; a rule synthesis processing section that rulesically synthesizes either the generated fixed accent type or the automatic accent type and the symbol string; and a rule synthesis processing section that converts the speech parameters output from the rule synthesis processing section into speech signals to generate speech. A speech synthesis device comprising: means for outputting;