JPH01106000A - Voice encoder - Google Patents

Abstract

PURPOSE: To make a synthesizer on an audio responding device side small-sized by analyzing inputted text data and writing voice data consisting of rhythm information and pronunciation information in a storage element. CONSTITUTION: A document is inputted through a text data input means consisting of a keyboard and other document input means. Its text data me converted by a document analyzing means 8 into the voice data consisting of the rhythm information (c) and pronunciation information (d) by a text synthesizing method according to a word dictionary and a rule, and the converged voice data are further encoded into, for example, a bar code by a voice data writing means 9 and written on printed matter, e.g. a teaching material, etc., to obtain the storage element 1. An IC card is usable as the storage element 1. Thus, the synthesizer on the audio responding device can be made small-sized.

Description

DETAILED DESCRIPTION OF THE INVENTION [Technical Field] The present invention provides a voice encoding device that creates voice data to be used in a voice response device and stores it in a storage element.

[Background Art] As the content of information and communication services has expanded, various types of terminals have come into practical use as terminals for transmitting computer-processed results to humans. Conventionally, visual means such as printers and displays have been mainly used as an interface between humans and machines, but it is often easier to use and more economical to convey information to humans by voice.
Recently, it has become widely used for various services.

Such a voice response device is generally called a voice response device.

This voice response device stores data for outputting voice used in response, and creates and outputs voice in accordance with the service request based on this data.

The storage element 1 that has accumulated this data is shown in Figure 4(a).
As shown in FIG. 4, a ROM built into the voice response device 2, and an external memory such as an IC card attached to the voice response device 2 as shown in FIG. 4(b) are used.

In the voice response wave fi2 of FIG. 4(a) in which the storage element 1 is incorporated, necessary data in the data b of the storage element 1 is selected by the selection signal a from the outside, and the synthesizer 3
It is now output to .

In this case, since the only information input from the outside is the selection signal a, the amount of input information is small, but since the storage element 1 is internal, there is a problem that the device becomes larger.

On the other hand, in the case of FIG. 4(b) in which the storage element 1 is external, the data b selected by the selection signal a is manually inputted to the voice response device 2, so the amount of input information to the device is large, but the storage element 1 is stored internally. Therefore, the device is not as large as the Moesha.

For example, when the voice response device 2 is used as a learning device, each person uses the voice response device 2, so the voice response wave r!
It is desirable that the 12 is small and inexpensive, and if the content of the output audio (stored data) is frequently changed or added, it is better to have the data stored externally. It is thought that

By the way, when recording and storing data, the audio data must be encoded in a form that can be synthesized into speech.Speech synthesis methods include recording editing method, parameter editing method,
Examples include rule synthesis method. Comparing these three methods, the results are shown in the table below.Among them, the rule synthesis method has inferior voice quality compared to the other two methods, but has a very high information compression rate, making it possible to create a small and inexpensive voice response device. suitable for cases.

This rule synthesis method is a method for creating speech based on phonetic and linguistic rules from pronunciation information (character strings) and prosodic information such as accent and intonation. especially,
A method that creates speech only from pronunciation information (character strings) is called a text synthesis method, and it can be said that this method is the final form of speech synthesis that extends to the intellectual activity of human speech. If the information contained in speech is divided into M-word information and naturalness such as individuality, the current text synthesis method can only reproduce linguistic information, but there are applications in which it is only necessary to convey linguistic information. is considered to be wide. It is also a useful method for learning the standard way of speaking in language learning materials. Typical text synthesis methods, for example, for Japanese, are
“Self-speech, small door, Sagayama speech synthesis from Japanese text”
, Tsuken Monthly Report, 32, 11. pp, 2243-2252 (
1972)", and for English, "^l l
en*J, Carlsonkai R0 Granstromv
B, lunicutt, Kiatt, D, and
PisoniyD: MrTALK-79″: C
version of unrestricted
ed Englisb II text・to・
'5peech+MrT (1979) J.

Synthesizers based on such a text synthesis method can be roughly divided into, as shown in Figure 5, input text data and generate prosodic information C1 such as accent/intonation and phoneme information and pronunciation information d using word dictionaries and various rules. A text analysis unit 4, a control information generation unit 5 that generates control information e from the information generated by these text analysis units 4, and a control Speech synthesis unit 6 that synthesizes an audio signal based on I report e
In particular, the text analysis section 4 that generates accent/intonation and prosody information is large, and if this text synthesis system is used as a synthesizer, it is inevitable that the device will be large.

Therefore, if the voice response device 2, which stores text data in the storage element 1 and synthesizes speech from the text data, becomes obsolete, the synthesizer There is a problem in that it becomes larger.

Therefore, a method has been proposed in which rule synthesis is performed without using the document analysis unit 4, but in this case, when writing the prosody information to the storage element 1, the prosody information C to be written is created manually, which is extremely inefficient. It was a target.

OBJECT OF THE INVENTION] In view of the above-mentioned points, an object of the present invention is to easily store voice data consisting of prosodic information and pronunciation information in a storage mary, and as a result, the voice response device side can easily store voice data consisting of prosodic information and pronunciation information. An object of the present invention is to provide a speech encoding device that allows a synthesizer to be miniaturized.

[Disclosure 1 of the Invention As shown in FIG. The apparatus is characterized in that it includes a sentence analysis means 8 for analyzing the voice data, and a means 9 for writing the audio data into the storage element 1.

The operation of the embodiment of the speech encoding device of the present invention will be described below.

FIG. 2 shows a 70-chart of the operation of the embodiment,
First, the apparatus n of the present invention is started, and a text is inputted using the text data manual means 7 consisting of a keyboard or other text input means. This text data is the text analysis method 8
Prosodic information (accent, intonation, pause, speech level, phonological duration, etc.) and pronunciation information (
d (indicating a sequence of synthesis units during speech synthesis), and this converted speech data is encoded as a bar code, for example, by the speech data writing means 9, and then sent to a printed material, e.g. Written in teaching materials, etc. In other words, the printed material on which this barcode is printed becomes -a product index 1.

In this way, in this example, if the text data is manually processed,
The information can be automatically converted into prosodic information C and pronunciation information d used for rule synthesis, encoded, and stored in the storage element 1. Here, since the amount of information in the audio data is small, M can be multiplied by the barcode as mentioned above, and since it is a barcode, it can be randomly accessed by humans, and special selections can be made on the audio output means side. It does not require any means, and since it is a printed matter, it can be easily reproduced.

FIG. 3 shows storage element 1 created by the speech encoding device of the present invention.
This figure shows a voice response device 2, which is a voice output means for reading voice data from a computer and synthesizing it.
! ! 2 is equipped with a barcode league that reads code data from bar food, decodes the code data read from the barcode into prosody information C and pronunciation information d, and converts these information c
, d, a control information generation section 5 that creates control information e necessary for speech synthesis, and a speech synthesis section 6 that synthesizes an audio signal based on the control information e from the control information generation section 5. It is equipped with a so-called rule synthesis system. In this case, unlike the text synthesis method, there is no need to incorporate a text analysis section, so the voice response device 2 can be manufactured in a small size and at low cost.

Another advantage is that there is less manual input compared to the recording editing method or the parameter editing method.

Note that since the voice response device 2 is small, R is used as the storage element 1.
Even if an internal point storage element such as OM is used, the voice response device 2 does not become larger. Also, the C card may be used as the M product element 1. Of course, if the type of storage element 1 changes, the configuration of the voice writing means 9 will also change, and the code input means of the control information generating section 5 of the voice response device 2 will also change.

[Effects of the Invention] The present invention includes a text data input means for inputting text data as described above, a text analysis means for analyzing the input dext data and outputting audio data consisting of prosody information and pronunciation information, Since it is equipped with a means for writing the above-mentioned audio data into the storage element, just by inputting text, the pronunciation information and prosody information used for rule synthesis can be transferred to the M product means!
Since it is possible to memorize products, efficiency is greatly improved compared to the conventional method of creating information manually.Furthermore, unlike text synthesis, each synthesizer does not require a text analysis means, so it can be used as an audio output means. In addition to being able to reduce the size of the device, the manufacturing cost is also low, and it is possible to satisfy the requirements for use as a learning material, for example.

[Brief explanation of the drawing]

FIG. 1 is a circuit configuration diagram of the present invention, FIG. 2 is a 70-chart for explaining the operation of an embodiment of the present invention, and FIG. 3 is a circuit configuration diagram of a voice response device used in the above. The diagrams R4& and Pt55 are also single diagrams of the explanatory circuit of the conventional example. 1...M product elements, 7...Text data input means,
8: Sentence analysis means; 9: Audio data writing means. Agent Patent Attorney Ishi 1) Ai 7th Diagram 2 (A) December 28, 1986

Claims (3)

[Claims] (1) A text data input means for inputting text data, a text analysis means for analyzing the input text data and outputting audio data consisting of prosody information and pronunciation information, and a means for writing the audio data into a storage element. A speech encoding device comprising: (2) The speech encoding device according to claim 1, wherein a bar code is used as the storage element. (3) The speech encoding device according to claim 1, characterized in that the storage element is an IC card.