JPH02129686A - Conversation aid apparatus - Google Patents

Abstract

PURPOSE: To use a word or a conversational sentence which are not prepared by a memory by syntheisizing the voice data of a voice storage means with voice data selected by a selecting means as a series of voice data, and outputting it as a voice. CONSTITUTION: Conversational data such as a stylized sentence are prepared as a conversational data group M4, and when the output of the desired conversation is instructed by an operator from an instruction inputting means M3, the voice data of the desired conversation are selected from the conversational data group M4 by a selecting means M5. The voice data are synthesized with the voice data directly obtained from the voice of the operator stored through a voice inputting means M1 in a voice storage means M2 as a series of voice data by a synthesizing means M6. Then, this series of voice data are uttered from a voice outputting means M7. Thus, a desired complete sentence can be automatically and smoothly uttered altogether with a sentence in the memory only by uttering and inputting a voice which is not prepared by the memory.

Description

[Detailed Description of the Invention] え咀Ωdanku [Field of Industrial Application] The present invention relates to a conversation aid device, and in particular, in foreign language conversation, it is possible to have a conversation tailored to the situation at the moment with little knowledge, or to adjust one's own voice. This invention relates to auxiliary devices for conversation.

[Prior Art] Conventionally, as a pronunciation practice device, there is a device in which when a predetermined word is inputted from a keyboard, the utterance corresponding to the word is synthesized and displayed using an electronic circuit (Japanese Patent Laid-Open No. 62-36685).
issue). There is also a device that outputs a predetermined text for a message in the form of voice (Japanese Patent Laid-Open No. 62-40524). There is also a device that electrically stores human voice in a semiconductor memory and reproduces it later (Japanese Patent Laid-Open No. 62-55698).

[Problems to be Solved by the Invention] However, these devices simply allow the operator to pronounce desired conversations and words, or record and reproduce the desired conversation or words in place of a tape recorder. Therefore, it is difficult to use words or conversational sentences that are not prepared in the memory, and it is also impossible to utter the conversational sentences stored in the device in one's own voice.

Structure of the Invention The present invention solves this problem, and provides a conversation aid device that can make full use of the knowledge stored in the device's memory, and can also output words and conversational sentences that are not in the device's memory in voice.
Another object of the present invention is to provide a conversation auxiliary device that outputs the conversation using the operator's voice.

[Means for solving the problem] The configuration for solving the above problem is as follows. That is, the conversation auxiliary device of the first invention, as illustrated in FIG. 1(A), includes: voice input means M1; voice storage means M2 for storing the voice input from voice input means M1; and instruction input means M3. and a selection means M5 for selecting audio data of a predetermined conversation from a conversation data group M4 consisting of audio data based on an instruction inputted from the instruction inputting means M3; The present invention is characterized by comprising: a synthesizing means M6 for synthesizing the voice data selected by the synthesizing means M6 as a series of voice data; and an audio output means M7 for outputting the voice data synthesized by the synthesizing means M6 as voice.

The conversation auxiliary device of the second invention, as illustrated in FIG. 1(B), includes: voice input means Mll; voice storage means M12 for storing the voice input from the voice input means Mll; instruction input means M13; Selection means M15 selects predetermined conversation data from the conversation data group M14 based on instructions input from the instruction input means M13; Characteristic extraction means MIB extracts characteristics of the voice pattern input from the voice input means Mll. , data conversion means M17 for converting the conversation data selected by the selection means M15 into voice data having the voice pattern characteristics extracted by the characteristic extraction means M16, and voice data of the voice storage means M12 and data conversion means M1.
synthesis means M18 for synthesizing the voice data converted in step 7 as a series of voice data; and voice output means M for outputting the voice data synthesized by the synthesis means M1B as voice.
19.

[Operations] Conversation data such as first invention fixed phrases are prepared as a conversation data group M4, and when the operator instructs output of a predetermined conversation from the instruction input means M3, the selection means M5 selects a desired one from the conversation data group M4. Select the audio data of the conversation. The voice data of this conversation is data in which the part uttered by the operator is blank. This voice data is synthesized by the synthesis means M6 into a series of voice data and voice data directly obtained from the operator's voice stored in the voice storage means M2 via the voice input means M1. This series of audio data is output from the audio output means M7.

That is, the contents of the spoken conversation are output as a complete conversation by combining the part inputted by the operator's voice and the conversation in the conversation data group into a series of sentences.

Conversation data such as fixed sentences of the second invention are prepared as a conversation data group M14, and when the operator instructs output of a predetermined conversation from the instruction input means M13, the selection means M15 selects the conversation data group M14.
Select desired conversation data from 14. Furthermore, similar to the first invention, the voice input from the voice input means Mll is stored in the voice storage means M12.

In the second invention, the characteristic extraction means M16 extracts the pattern characteristics of the voice input from the voice input means Mll. Next, the conversation data selected above is
The data conversion means M17 converts the audio data into audio data having the above pattern characteristics. This converted audio data is
Similar to the first invention, the synthesizing means M1B synthesizes the audio data stored in the audio storage means M12 into a series of audio data, and this series of audio data is uttered from the audio output means M19.

In the case of the second invention, all of the synthesized conversation sentences are output in the operator's voice or a voice similar to the operator's voice.

[Embodiment] Fig. 2 shows an embodiment of the conversation assisting device of the first invention. This device is configured as an English, French, and Spanish conversation aid.

The appearance of the conversation aid device 1 is card-shaped, and on its side,
Instruction input keys 3 are arranged, and a display liquid crystal panel (L
CD) 5, a speaker 7 for audio output, and a microphone 9 for audio input.

As shown in the block diagram of Fig. 3, this internal structure is configured as a microcomputer, and the main part is CPU11.
．． ROM13. With RAM15, here 5, key 3
and a microphone 90 input circuit 17, and an output circuit 19 for the liquid crystal panel 5 and speaker 7. These components are connected to enable signal transmission and reception via an address bus and a data bus (not shown).

The flowchart in FIG. 4 shows the processing of the conversation assisting device 1 of the first embodiment of the invention. This process is repeatedly executed after the battery 21 is attached to the conversation assisting device 1.

When the process starts, initial settings are first made, and period values of various variables and flags are set (step 110). Next, the system waits for a key input (step 120). If any of the keys 3 is pressed here, the contents of the key are determined (step 130).

When French key 3b is pressed, ROM13 or RAM1
The French text file stored in 5 is set as a text selection target (step 135), and at this time, the message "French conversation has been set" is displayed on the LCD 5 (step 137). ).

Next, when the text selection key 3e is pressed, the corresponding language,
At this time, one French sentence is read from the beginning of the file and displayed (step 140). If this sentence is not the desired sentence, when the next candidate selection key 3h is pressed (step 15o), the next sentence is displayed (step 14).
0). When the desired sentence is displayed, press the confirm key 3d and the audio data corresponding to the confirmed sentence will be transferred to the RO.
The data is stored from M13 to a predetermined address in the RAM 15 (step 160).

For example, a plurality of conversation sentences that are sequentially displayed on the LCD 5, such as rcomb i en y a-t-i 1d'1
cia$location$? ” rQue prendrez-vous.

$ thing * ou $ thing $? ” From among the rQue prendrez-vous.

$things$ou $things$? ” When you select a sentence, the audio data of this sentence is stored in RAM.
15. This audio data may be composed of parameters obtained by various types of analysis table shaping encoding methods, or may be audio data obtained by recording and reproducing methods. Note that "$mono$" at 2 ftf in a sentence or chapter is an audio data insertion part and is not converted into audio data.

Next, a voice input instruction display is displayed on the LCD 5, ``Please press the voice key and then input voice.'' (step 170).

When the voice key 3f is pressed, it becomes possible to input voice from the microphone 9 and store it (step 180). The waveform of the audio input from the microphone 9 is converted into a digital value by an A/D converter provided in the input circuit 17, and the digital value is stored in a buffer set in the RAM 15. For example, if you pronounce rdu 1aitJ, the pronunciation data will be stored in the buffer. This audio data corresponds to the first "$mono$" in the conversational text.

Thereafter, if the voice person key 3f is pressed again (step 184), the voice input and storage process of step 180 is repeated as the previous correction process.

When the confirmation key 3d is pressed (step 186), it is determined whether the number of processing times has reached a predetermined number (step 1).
8th). Since there are two "$ items $" and one more voice data input is required, a negative determination is made and the process moves to the second voice input/storage process (step 180). That is, the audio data corresponding to the second "$thing $" in the conversational text is similarly input and stored. For example, rde
If you pronounce "la creme", it will be stored in the basso as the second pronunciation data.

Next, insert the two audio data stored in step 180 into the two "$thing $" parts in this audio data,
Combine them into one sentence (Ste Mbu 210). That is, rQue prendrez-vous, dufai
'tou de la crome? The audio data corresponding to the sentence ``'' will be synthesized.

Next, a voice output instruction is displayed (step 220), and the system waits for key input (step 120).

Next, when the audio output key 3g is pressed, the audio data synthesized in step 210 is output from the speaker 7 (step 230).

That is, rQue prendrez-vous, dulai
tou de la cr″eme ?J
The sound will be output from the speaker 7.

Therefore, by simply uttering and inputting the desired word, the desired sentence will be automatically and smoothly uttered.

In this embodiment, even if the insertion and synthesis processing of the stem 210 is not performed, other synthesis processing can be performed on the stem 210.
When outputting in step 30, the outputting may be performed in the order in which they should be uttered.

That is, rQue prendreZ-V from ROM13
After extracting and uttering the OuSJ audio data, the RAM
15, the voice data of ``rd u l a it'' is extracted and uttered, then the voice data of ``Ou'' is extracted from the ROM 13 and uttered, and then the voice data of ``rd u l a it'' is extracted from the RAM 15 and uttered.
The audio data of "de la creme" may be extracted and uttered. In addition, if such a utterance method is used, it is not necessary to store the audio data itself in step 160, but only the memory and address in which the audio data of the sentence is stored may be stored.

In the above embodiment, the microphone 9 corresponds to the voice input means M1, the RAM 15 corresponds to the voice storage means M2, the next candidate selection key 3h corresponds to the instruction input means M3, and the CPU
I corresponds to the selection means M5 and the synthesis means M6, and the speaker 7 corresponds to the audio output means M7. CPL] 11, the processing of steps 140 and 150 corresponds to the processing as the selection means M5, and the processing of step 210 corresponds to the processing as the synthesis means M6.

The flowchart in FIG. 5 shows the processing of the conversation assisting device according to an embodiment of the second invention. The hardware configuration of this embodiment is the same as the embodiment of the first invention shown in FIGS. 2 and 3, so a description thereof will be omitted. This process is repeatedly executed after the battery 21 is attached to the conversation assisting device.

When the process starts, initial settings are first made, and initial values of various variables and flags are set (step 310). Next, the system waits for a key input (step 320). key 3 here
When any one of the keys is pressed, the contents of the key are determined (step 330).

When English key 3a is pressed, ROM13 or RAM1
The English text file stored in 5 is set as a text selection target (step 335). At this time, the message "English conversation has been set" is displayed on the LCD 5 (step 337).

Next, when the text selection key 3e is pressed, the corresponding language,
At this time, one English sentence is read from the beginning of the file and displayed (step 340). If this sentence is not the desired sentence, when the next candidate selection key 3h is pressed (step 350), the next sentence is displayed (step 34).
0). When the desired sentence is displayed, press the confirm key 3d and the audio data corresponding to the confirmed sentence will be transferred to the RO.
The data is stored from M13 to a predetermined address in the RAM 15 (step 360).

For example, a plurality of conversation sentences are sequentially displayed on the LCD 5, such as rWhere can I get $things$?
”Would you page $people$? ”
From the rWould you page $people$? ”, the audio data of this sentence will be stored in RAM1.
5 out of 5 times will be stored. This audio data is composed of parameters obtained by various analysis-synthesis encoding methods.

Note that in the text, "$人$" is a portion into which audio data is inserted, and is not converted into audio data.

Next, a voice input instruction display is displayed on the LCD 5, ``Please press the voice key and then input voice.'' (step 370).

When the voice key 3f is pressed, it becomes possible to input voice from the microphone 9 and store it (step 380). The waveform of the audio input from the microphone 9 is converted into a digital value by an A/D converter provided in the input circuit 17, and the digital value is stored in a buffer set in the RAM 15. For example, if you pronounce rMr, Smit hJ,
The sound data is stored in a buffer. This audio data corresponds to "$人$" in the above conversation sentence.

Thereafter, if the voice person key 3f is pressed again (step 3, day 4), the voice input and storage process of step 380 is repeated as the previous correction process.

If the confirmation key 3d is pressed (step 386), it is determined whether the number of processing times has reached a predetermined number (step 3
88). Here, since only one piece of audio data, "$人$", is required, a negative determination is made, and then pattern analysis of the audio stored in the buffer is performed in step 380 (step 390). .

That is, parameters representing characteristics of the speaker's voice are detected using a generally known analysis-synthesis encoding method such as the PARCOR vocoder method or the LSP vocoder method. In particular, since the spectral envelope and fundamental frequency (pitch) represent special effects, only these two parameters may be captured.

Next, among the standard parameters of the audio data of the sentence stored in step 360, the spectral envelope parameter and fundamental frequency (pitch) parameter are stored in step 39.
It is replaced with the parameter detected at 0 (step 400). Of course, if other parameters are detected by the stem 390, those parameters may also be replaced. In this way, the audio data of the selected sentence is approximated to the speaker's voice.

Next, the voice data stored in step 380 is inserted into the "$人$" portion of the approximated voice data to synthesize one sentence (step 410).

Next, a voice output instruction is displayed (step 420), and the system waits for key input (step 320).

Next, when the audio output key 3g is pressed, the audio data synthesized in step 410 is output from the speaker 7 (step 430).

That is, rWould you page Mr. Smith? J The sound will be output from the speaker 7.

Furthermore, among these sounds, rMr, Smith thJ are completely the sound quality of the speaker, and rWould youpageJ
has a sound quality that approximates that of the speaker.

Therefore, by simply speaking and inputting the desired word, the desired sentence will be automatically and smoothly uttered in one's own voice.

In this embodiment, even if step 410 is not specially inserted and synthesized, step 43 can be performed as another synthesis process.
Similar to the embodiment of the first invention, the output may be performed in the order in which the utterances should be made when outputting the utterances at 0.

In the above embodiment, the microphone 9 corresponds to the voice input means Mll, the next candidate selection key 3h corresponds to the instruction input means M13, the CPU II corresponds to the selection means M15, and the characteristic extraction means M1.
6. corresponds to the data conversion means M17 and the synthesis means M1B, the RAM 15 corresponds to the voice storage means M12, and the speaker 7 corresponds to the voice output means M19. Among the processes of the CPU II, the processes of steps 340 and 350 are performed by the selection means M15.
The processing at step 390 corresponds to the processing as the characteristic extraction means M16, the processing at step 400 corresponds to the processing as the data conversion means M17, and the processing at step 410 corresponds to the processing as the synthesis means M18. Applies to.

In each embodiment, a dictionary function is provided in case the word to be pronounced is unknown, and when Japanese is input from the separately provided kana key when processing Ste-Tsub 180°380,
The information may be converted into English, French, or Spanish and output to the LCD 5. At the same time, the pronunciation may be output as audio from the speaker 7. The operator can look at it or ask it and input it by voice.

Further, in the display processing of STEP 140.340, when displaying sentences in each language, Japanese may also be displayed.

for example, rQue prendrez-vous.

$things$ou $things$? Which would you prefer, $things or $things? ”rWool
d you page ＄人＄? Could you please call someone $? ” will be displayed.

Furthermore, the device may generate a sound at the same time as this display. This has the advantage that the operator can memorize the correct pronunciation in advance.

According to the first invention, by simply speaking and inputting the desired word, the desired complete sentence is automatically and smoothly uttered together with the sentence in memory. Become.

According to the second invention, the desired sentence can be uttered smoothly in one's own voice.

By using a conversation aid device such as those disclosed in the inventions, it is possible to immediately and smoothly utter the necessary conversation sentences. The speaker pronounces only one part, and the rest is pronounced accurately by the device, so even if the speaker is not good at pronunciation, it is easy for foreign listeners to understand. In particular, in the conversation assisting device of the second invention, the entire text has a uniform sound quality that is close to the speaker's voice, making it easier for the listener to understand without feeling any discomfort.

[Brief explanation of drawings]

FIG. 1(A) is a diagram illustrating the basic configuration of the first invention;
B) is a diagram illustrating the basic configuration of the second invention, FIG. 2 is an external perspective view of each embodiment of the first invention and the second invention, FIG. 3 is a block diagram thereof, and FIG. 4 is an implementation of the first invention. FIG. 5 shows a flowchart of the processing contents of the example of the embodiment of the second invention. Ml, MI M2. MI M3. MI M4. MI M5. MI M6. Ml 1... Voice input means 2... Voice storage means 3... Instruction input means 4... Conversation data group 5... Selection means 8... Synthesis means M7. M19...Audio output means M2S...Characteristic extraction means M17...Data conversion means 3h...Next candidate selection key 7...Speaker 9.
...Microphone 11...CPU13...R
OM 15...RAM agent Patent attorney Tsutomu Sadatsu (2 people)

Claims (1)

[Claims] 1. A voice input means, a voice storage means for storing the voice input from the voice input means, an instruction input means, and a conversation consisting of voice data based on the instructions input from the instruction input means. a selection means for selecting voice data of a predetermined conversation from a data group; a synthesis means for synthesizing the voice data in the voice storage means and the voice data selected by the selection means as a series of voice data; and synthesis by the synthesis means. A conversation auxiliary device comprising: a voice output means for outputting the voice data as voice; 2 voice input means; voice storage means for storing the voice input from the voice input means; instruction input means; and selecting predetermined conversation data from the conversation data group based on the instructions input from the instruction input means. a selection means; a characteristic extraction means for extracting the characteristics of the voice pattern input from the voice input means; and a characteristic extraction means for extracting the characteristics of the voice pattern input from the voice input means; a data converting means for converting; a synthesizing means for synthesizing the audio data in the audio storage means and the audio data converted by the data converting means as a series of audio data; and outputting the audio data synthesized by the synthesizing means as audio. A conversation auxiliary device comprising: a voice output means for outputting a voice;