JP2908720B2 - Synthetic based conversation training device and method - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to speech.
The present invention relates to a training system and a method, and more particularly, to a training apparatus and a method for a student having a language disorder. Also, it relates to learning of foreign language conversations of learners other than hearing-impaired persons.

[0002]

2. Description of the Related Art The most basic teaching (training) method of teaching a hearing-impaired child to speak, in particular to speak, and to hear by visually observing the movements of the mouth, etc. There is a method of using a vocal apparatus to indicate a correct sounding operation. by this,
The child learner (trainee) can observe the movement of the external vocal organs such as the lips and chin that occur during the production of language speech, and can also observe the movement of the tongue to some extent. As a tactile teaching method, a method in which a learner touches and compares a teacher's vocal organs with his / her own vocal organs to correct their own mistakes is sometimes used. Such conventional teaching methods have limitations such as that many of the vocal movements cannot be observed from the outside.

In recent years, it has become possible to use a device for analyzing speech or a computer program to teach a method of speaking a language. In the teaching method using these devices and programs, the learner can observe various features associated with the production of many voices, such as display of voice information. An example of an optimal use of such a system is a computer-integrated conversation developed by Matsushita, especially a vocal training aid (CIST).
A, Computer IntegratedSpeed
ch Training Aid). CISTA
Provides multi-channel data collected by several converters. The contents are as follows.

[0004] 1. Dynamic palate diagram. In 1962, a former Soviet scholar, Y. Kuzmin is the first teaching method used, and the contact between the tongue and the palate that occurs during vocalization is shown using the contact of many electrodes on the palate artificially provided in the mouth. When the tongue touches one of the electrodes, a low voltage circuit with the device that goes outside the mouth communicates and detects this. Further, the presence or absence of the contact is displayed on a CRT (display device).

[0005] 2. Nasal sound sensor. Headgear (Head gear) on the side of the nasal cavity or a microphone (Electret microphone) temporarily attached with adhesive tape detects the trembling of nasalization. 3. Throat sensor. A microphone attached to the larynx with a flexible collar detects glottal tremors.

[0006] 4. Air flow sensor. In front of the child learner's mouth, there are several methods that use a device that senses the flow of exhaled air. 5. Perform input for voice analysis using a standard microphone. Teaching conversations to deaf children learners is a difficult task due to the limited time for teachers to teach. Children with no hearing impairments listen for hours on a daily basis and get audio feedback on their utterances, while children with hearing impairments spend about one week a week training conversation. You just get feedback on your utterances.

[0007] Combining a computer with a conversation training device, such as CISTA, allows some of the training of conversation, especially vocalization, to be performed without a teacher, thus greatly extending the training time. Another conventional example is a document (chapter) audio system.
Any utterance (refers to the target language, word, sentence, etc., is not limited to linguistic "utterance";
lance) is typed in automatically. The best known of such document speech systems is currently Digital Equipment (D).
It is "DECTalk" which utters English from a document of the company (Equipment Equipment).

Next, the study of vocalization will be described. The shape of the human vocal tract (the way the voice is produced by the vocal organs) determines the resonance state for the utterance, and further controls the human utterance itself. For this reason, electronic and computational models showing the relationship between the shape of the vocal tract and the resulting sound output have been a major research topic for conversation researchers for many years. And in this study, the shape of the vocal tract was examined by the researchers themselves, and the sound output was measured. More recently, researchers have identified articulatory articulations.
Vocal tract shapes are generated automatically. In this case, the input is a string of phonemes (string
fphonemes), and this group of phonemes is converted into a detailed specification of the shape of the vocal tract. The vocal tract shape is used to form a vocal tract model that emits speech.

[0009]

SUMMARY OF THE INVENTION However, CIS
In conversational training devices that provide feedback directly to the learner, such as TAs, guidance content is limited to individual voices or pre-programmed utterances or conversations. Also, "DE
In CTTalk, the document speech system creates only audible sounds.

[0010] In addition, the display of each vocal organ, air flow, and the like, which affect the accuracy of utterance, cannot be said to be perfect. In particular, the movements of the tongue and palate that play an important role in the actual formation of the vocal tract, specifically the contact patterns, the use of which parts are coming and going at what time, etc., and the display are unclear . And this was not considered to be adopted for learning training and guidance.

[0011] In addition, it is insufficient for effective learning for those who still have some hearing. It is also inconvenient for learning a foreign language. The present invention has been made in view of the above problems, and provides a training device and a method that enable a child learner to obtain feedback information for any utterance without the guidance of a teacher. The purpose is.

Another object of the present invention is to provide a training apparatus and a training method for making it possible to understand the state of vocal tract formation, which plays an important role in vocal formation. At the same time, it aims to provide useful things for learning foreign languages.

[0013]

In order to achieve the above objects, the conversation training apparatus and method according to the present invention provide a learner (trainee) with any utterance (utt) desired to learn (train).
erence) and foreign languages are also typed into the computer, and articulatory model mov, which is the correct operation of various vocal organs (necessary) that are not necessarily visible due to the utterance of the typed document.
elements, especially the state of contact between the tongue and palate (tong)
ue-palate contact pattern
s. ), And the state of the airflow and the movement of each organ accompanying the form are displayed on a CRT and can be learned.

For those who still have some hearing,
Make the correct utterance sound loud. In order to achieve another object, the conversation training apparatus and method of the present invention provides a method for learning similarities and differences between an actual utterance of a learner and an exemplary utterance when the learner utters a typed utterance or a foreign language. Are compared, and the evaluation result is displayed on the CRT, thereby enabling the feedback that the correct utterance is trained and learned.

[0015]

According to the above arrangement, first, the editing means edits and converts a typed document (chapter) for utterance, and sends it to the speech synthesizer. The speech synthesizer analyzes the type-input document into a series of phoneme groups, and creates parameters defining the pronunciation characteristics of the document every 10 milliseconds. The generated set of parameters is returned to the system and converted into a set of parameters needed to utter the typed document, for example, an articulatory pattern for tongue-palate contact. Then, a video of a contact pattern indicating the location and time relationship between the tongue and the palate for the articulation is displayed on the CRT.

[0016] A set of transducers is used to measure the nasalization vibration, glottal vibration, exhaled air flow, tongue contact, and utterance discrimination when the learner actually utters the input utterance that is documented and input. You. The conversation training device evaluates the sound uttered by the learner based on, for example, how similar it is to an exemplary tongue-palate contact pattern. This evaluation is displayed on the CRT as a screen of the contact pattern of the tongue and palate.

[0017] Also, training in a foreign language to give an exemplary utterance,
The same goes for learning.

[0018]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of a speech ability training system based on synthesis using the present invention will be described below with reference to the drawings and tables. (First Embodiment) FIG. 1 shows the configuration. The speech training apparatus based on synthesis according to the present embodiment may use a digital computer for a laboratory. Such a device has means for inputting an utterance to be learned, and as such means, the keyboard 1 constituting an input unit of the conversation training device is suitable.

The input utterance is sent to a voice training device of the ASCII notation system. The conversation training device
Utterances that have been type-input using the document speech device are automatically synthesized. Here, "DECTalk" manufactured by Digital Equipment Corporation is the most well-known as a document speech conversion device for English. However, in this embodiment, Panasonic Technology (Panasonic Technology) is used.
ies) Conversation technology belonging to the company (Speech Tech)
Novel) is a state-of-the-art text-to-speech system "STLtalk" developed subsequently by the Institute.

Communication between the synthesizing unit 3 and the conversation training device is performed by RS
-232 port. RS-232 is XO
It is configured for the N / XOFF communication protocol (protocol) and its transmission setting is 9600 bits / second (ba
ud), 8 data bits, 1 stop bit, and 0 parity bit. The synthesizing unit 3 receives an input document which is a series of phonemes, and stores the input document in an input buffer.

The document utterance unit 2 analyzes the input utterance. Syntax analysis of utterance (synactact)
ic analysis) is a function word (func) such as the position of a comma in spelling and a preposition detected at the time of dictionary search.
Tion words) and the syntactic role of the verb. Next, in accordance with the procedure shown in FIG. 2, phonemic information of the utterance (phonemic representation)
n) is created. Hereinafter, the contents of this drawing will be described.

First, each word is compared with a headword of a small pronunciation dictionary (step 12). If there is no corresponding headword in the dictionary, the common suffix such as “ed” or “ing” is removed to convert the word into a morpheme (morphem).
es) (step 14). After the division, the remaining roots are stored in a phonemic dictionary (phonemic dic).
(ionary) heading sound (step 1)
6). If there is no heading comparable to the root, the pronunciation is guessed by referring to the conversion rules from characters to phonemes (step 18). In addition, when converting a target word into a phoneme, a stress (stress) pattern to be used for a syllable is also examined. That is, when the target word is not described in the dictionary in the system or when the stress pattern of the root decomposed from the word changes depending on the prefix, it must be inferred. The stress level of a syllable is represented by inserting a stress symbol immediately before the vowel represented by the phoneme. When there is no stress sign, no stress is placed. When uttering, etc., if the common suffix has been removed, it is added again (step 2).
0).

The sound parameter synthesizing unit 3 processes the buffer to generate one set of 20 sound parameters for each 10 msec continuous phoneme. The sound parameter defines a feature of the learned utterance as a sound. These parameters represent the transmitted information as basic frequencies, amplitude numbers, and formants (strictly speaking, constituent vowels and formants. However, in this specification, the parameters may also refer to sounds to be trained. ) Is compared with a predetermined value based on the frequency, bandwidth, and noise source of the target utterance. Table 1 shows examples of sound parameters.

[0024]

[Table 1]

Two sets of the same sound parameters are created. Upon receiving the period, which is the symbol representing the end of an utterance, the first set is stored in an array (array, a storage device with an array of storage elements in the CPU). Next, the second set of sound parameters is sent to the formant synthesizing unit 4 and converted into an output signal in the form of an analog audio signal. The output signal converted by the formant synthesizing unit 4 is reproduced by the speaker 5 at a large volume, so that a learner who has some hearing can hear it.

The first set of sound parameters stored in the array of sound parameter synthesizers (excluding formant synthesis) 3 is returned to the conversation training device via the RS-232 port. The conversation training device reads these values and stores them in another array. Then, the sound parameters are converted into the articulation parameters by the conversion unit 6 of the sound parameters into the articulation parameters. The articulatory parameter represents three types of typed utterances.

The conversation training apparatus converts one group of sound parameters into articulation parameters indicating the voice characteristics of the utterance typed. The sound parameters used for the conversion include the fundamental frequency, the amplitude of the utterance, voiced or unvoiced, formant 1, 2, 3 frequencies. The document speech device is
The frequency (herz) of the sound parameter is multiplied by 10 and sent to the conversation training device. The conversational training apparatus using the frequency divides the received parameter by 10 and displays the screen on the CRT.

The conversation training apparatus also creates articulation parameters indicating the nasalization of the typed utterance. Nasalization is suitable for sounds such as / m / and / n /, but hearing impaired people often inappropriately nasalize the utterance (speak nasal), which hinders conversation. Nasal sounds are indicated by the presence or absence of both nasal poles and zeros in the frequency domain. Typically, nasalization is represented by one nasal pole and one nasal zero. These appear when the soft palate, the valve between the nose and mouth, is low and descends to the open position. When the soft palate closes, it disappears.

The text-to-speech device provides a sound parameter variable for detecting the nasalization of the typed utterance. The conversation training device creates a nasal index using the sound parameters. The document sounder sets the nasal pole and nasal zero to the same frequency and bandwidth as the non-nasal sound parameters. That is, a nasal pole and a nasal zero set at the same frequency and bandwidth cancel each other out.
In this case, the nasal pole and nasal zero are set to 250 Hz to represent non-nasal sounds. When creating nasal sounds, the nasal pole and nasal zero frequencies move to different values from each other,
Each affects a different part of the sound spectrum. The frequency of zero nasal sound is that the target sound is / n / 34 or / m / 36.
, Temporarily increase to 330 or 360 Hz. The variation in frequency is shown in FIG. The document sounder also provides the number of amplitudes of the nasal formants 30,32. The typical amplitude of a nasal formant is 30
, But when the amplitude number is zero, neither of the nasal formants 30, 32 is provided. The language ability training device that uses the sound parameters provided by the document voicing device performs an operation for creating a nasal index that informs a learner of a nasal volume required for performing a typed utterance.

One example of the above operation is as follows. NI = ｛(| F _nf −F _nz |) / (| F _nf −F
_nz |) _max ｝ × [｛(AN-1)-| (AN-1) |｝
/ 2 + 1] where NI = nasalization index F _nf = frequency of the nasal formant
sal formant) F _nz = frequency of the nasal zer
o) AN = amplitude of the nasa
l formant) where all input variables are integers.

In the second element on the right side of the arithmetic expression, the value is multiplied by 0 when the amplitude of nasalization is 0, and the value is multiplied by 1 when the amplitude of nasalization is other than 0. The speech training device also creates articulatory articulation parameters indicative of the tongue-palate contact required to make the typed utterance. Other than providing sound parameters, the document speech system is used to time the next four times. That is, the time when the sound starts, the time when the maximum amplitude is reached, the time when the amplitude starts to decrease, and the time when it ends. These times are shown in combination in a set of palate diagrams.

The maximum area for contact of each utterance unit, whether a phoneme or a syllable, is stored in the converter 6 for converting sound parameters into articulation parameters, and the target of the utterance unit (target)
Treated as a pattern. The voice of the conversation is in various contexts (context, word, sound, etc., for example, "t" means "ct", "th", "st", etc.)
Since the contact pattern changes according to t), it is necessary to store many different contact patterns not only for each sound, but also for each context and context.

Each utterance unit has a start and end time,
The duration of the part of the steady state (duration
on) data is assigned. The contact area between the tongue and palate is defined as a set of 63 points. Here, 63 voices are used for various sounds,
This is in consideration of the fact that the state of contact between the tongue and the palate can be sufficiently understood with respect to the pronunciation unit, that the visibility on the CRT is good, and that practicality in actual production and use.

The classification hierarchy is:
A number of points are defined in such a way that they touch and separate at the beginning and end. In general, the point of initial contact is the part towards the back of the palate, and the point of initial separation is the part in front of the palate. Since it is difficult to convert the sound parameters generated by the text-to-speech system into a tongue-palate contact pattern learned by humans, the present invention (embodiment) uses the consonant start and the tongue-palate contact pattern synthesis. The end is used (diverted).

For example, when the utterance "She Said" is uttered, the text-to-speech system notifies the start and end of / SH / and the start and end of / S / based on the synthesis parameters. FIG. 4 shows all parameters of the utterance “She Said”. For palatal contact,
Current training devices that provide actual contact patterns use only the beginning and ending times.

In current devices, the tongue-palate contact displayed by the learner is included in this document speech system.
FIG. 5 shows a typical pattern, "She
vowel tongue surrounding / S / sound in "said"
7 shows a palate contact pattern. The larger point indicates contact, and the smaller point indicates non-contact. 58 to 10
Each frame up to 4 represents 10 milliseconds.

The continuation of the frame in FIG. 5 represents a flow of a vowel in "She", a following / S / sound, and a vowel in "said". The first seven frames 58,
Numerals 60, 62, 64, 66, 68, and 70 indicate a contact for the vowel of "She" to follow the / S / frame 104 of "said".

Beginning with a complete / S / sound and frames 8 and 72, the tongue begins to separate from the palate at frame 90, nine frames later. By the 24th frame 104, “sai
The contact required for the vowel in d "is completely non-contact. In the conversation training program, the actual tongue contact operation displayed on the CRT as a screen is performed in the following three different ways.

The device has consonants / t /, / d /, / s /, /
When pronounced z /, / sh /, / zh /, / l /, / n /, / r /, the tongue contact pattern is stored in advance (prestored in an available mode, Prestored). . Up to now, the tongue contact pattern of / k / or / g / pronounced immediately before or immediately after the high tongue, the front tongue vowel is not included in the previously stored tongue contact pattern, but the principle applied Is the same.

The teacher stores the tongue contact pattern of each sound in a form usable by the device according to the situation in which it is formed. Also, while the learner and the child uttered the speech once well, the situation was different but the Japanese
For example, the sound determined by the teacher as having to be repeatedly practiced, such as the distinction between r and r, is selected and stored. Since the most appropriate tongue-palate contact is closely related to the shape of the palate of each learner, this storage method performed by the teacher is often most effective.

The teacher specifies which of the above three methods is to be used for each consonant. Unless otherwise specified, a method of storing a tongue contact pattern in advance is adopted. Common to all of the above methods, the conversation training device provides a series of contact patterns that begin before the consonant begins and continue until after the consonant ends. When a speech requesting a consonant with tongue-palate contact is typed into the device, the document speech system notifies the training device of the time for the start and end of the consonant. The series of patterns stored by one of the three methods described above is then used to display the appropriate contact pattern to the learner.

After confirming the state of the operation of contact between the tongue and the palate necessary for performing the utterance typed on the screen on the CRT, the learner actually utters the utterance. Here, the conversation training device uses the conversation training assistance device (CISTA) described above. Then, this device has several transducers (speech transduce).
rs) 8, various patterns associated with the utterance of the learner,
Measure the parameters. The learner's tongue and palate contact, nasal vibration, glottal vibration, exhaled airflow and sound are measured, and data necessary for evaluation during vocalization is acquired. Then, the data input to the converter is analyzed.

The evaluation unit 9 of the conversation training apparatus compares the measurement result accompanying the utterance of the learner with the stored sound parameters to detect and evaluate the similarity (difference). Thereafter, based on the difference between the sounding organ operation and the articulatory model operation actually performed by the learner, feedback is provided to the learner through the display unit 10. (Second Embodiment) In this embodiment, a learner having a normal hearing ability learns a foreign language conversation using a conversation training apparatus based on synthesis. Except as described below, the speech training (learning) device used in this embodiment is the same as the conversation training device used in the first embodiment.

The learner inputs a foreign language document to be learned into the input unit (keyboard) 1 of the conversation training apparatus.
The document audio unit 2 edits the input document for conversation. The edited document is sent to the sound parameter synthesizing unit (excluding the formant synthesizing unit) 3. The synthesizing unit 3 processes the document to create a sound parameter set including 20 sound parameters. Here, the sound parameter specifies a predetermined feature, in this case, a speech feature of a foreign language document to be learned.

Two sets of the same sound parameters are created. 2
The sound parameter set of the set is sent to the formant synthesizing unit 4 and is converted into an analog audio signal output signal. The output of the formant synthesis unit 4 is a predetermined foreign language,
The sound is reproduced from the speaker 5 as an utterance part. Then, 1
The sound parameter set of the group is sent to the conversion unit 6 for converting sound parameters into articulation parameters of the conversation training device. The conversation training device converts the sound parameters into the articulatory parameters by the conversion unit 6, and shows an articulatory model operation for producing an input document in a predetermined foreign language. The operation of each part, including the contact pattern of the tongue and palate including the consonant corresponding to the converted articulation parameters, is displayed on the CRT 7 on the screen.

The learner pronounces the input document in a foreign language by referring to the articulatory model operation including the contact pattern of the tongue and palate for uttering a document input in a predetermined foreign language. The conversation training device measures the learner's pronunciation. Next, the measured value accompanying the utterance of the learner is evaluated by comparing it with the document sound parameter input for the foreign language. The evaluation unit 9 of the conversation training device evaluates the learner by comparing it with a sound parameter for a document input in a foreign language. The conversation training device detects the difference between the articulation operation performed when the learner utters a document input in a foreign language and the articulation model operation stored in advance, and displays this to the learner. The feedback is provided using the display 10 for the feedback of the section.

While the present invention has been described with reference to preferred embodiments so that those skilled in the art can clearly understand and practice the present invention, the present invention is not limited to the description of the embodiments in the specification. Needless to say, various changes can be made without departing from the gist of the invention. Specifically, for example, an English language system uses phonemes as basic utterance units. For this reason, the embodiment describes an English system using phonemes as basic units. However, some other languages, such as Japanese, mainly use syllables as basic utterance units. Therefore, in the case of such a language, a system is used in which syllables are used as basic utterance units.

[0048]

As described above, according to the present invention, even a child with hearing impairment is required to correctly utter any word or document as long as he / she can enter with a keyboard, but it is not essential. The movement around the palate, the contact of the tongue, etc., which is difficult to visually recognize, is displayed on the CRT, so that it is possible to visually observe the movement. For this reason, it is easy to understand how to move the palate and tongue to make accurate sound, how to make contact, and the like.

Further, it also assists in hearing by mouth by observing the movement of another person while speaking. As described above, since hearing impairment prevents speech information from being obtained, it is suitable for a child learner who learns typing early. In addition, if the hearing is somewhat enhanced, it becomes possible to listen by setting the volume to a large volume, so that it is possible to perform better conversation training. In addition, we compared the measured values of the vocal organs such as the palate, the tongue, and the associated airflow, as well as the typical vocalized and vocalized organ movements, as well as the utterance itself when actually speaking. Since the results are displayed on the CRT, the training of conversation, especially vocalization, is more effective.

Also, a dedicated teacher is not required for conversation training. It can also be used by non-hearing people to learn foreign languages. In this case as well, the operation for accurate sounding of the palate, the tongue, etc., and the actual utterance and operation of the user, as well as the accurate utterance, are compared and displayed on the CRT.

[Brief description of the drawings]

FIG. 1 is a configuration diagram of one embodiment of a conversation training device.

FIG. 2 shows a phoneme, a word written based on the ASII system.
Stress and part-of-speech information (parts-of-speech)
9 is an operation flow chart of rewriting the information into (information).

FIG. 3 shows nasal zero and nasal pole during one utterance (pole,
(Pole) is a sound spectrum indicating sound parameters.

FIG. 4 is an example of a sound spectrum showing a contact pattern of the tongue with the palate occurring during one utterance.

FIG. 5 is a diagram showing an actual contact pattern between the tongue and the palate, which occurs during one utterance.

[Explanation of symbols]

Reference Signs List 1 Keyboard input unit for conversation to be learned 2 Document speech unit 3 Sound parameter synthesizing unit (excluding formant synthesizing unit) 4 Formant synthesizing unit 5 Speech unit of synthesized conversation 6 Conversion unit of sound parameters to articulation parameters 7 Articulation Parameter display section 8 Measuring device for learner's utterance 9 Evaluation section for evaluating learner's utterance using parameters 10 Display section as feedback

──────────────────────────────────────────────────の Continued on the front page (51) Int.Cl. ⁶ Identification symbol FI G10L 9/02 301 G10L 9/02 301A H04R 1/42 H04R 1/42 (72) Inventor Brian, Allen, Hanson United States of America California Goleta San Carpino 7457 (56) References JP-A-6-348297 (JP, A) JP-A-4-359299 (JP, A) (58) Fields investigated (Int. Cl. ⁶ , DB name) G09B 21 / 00 G09B 5/06 G09B 19/06 G10L 3/00 G10L 3/00 551 G10L 9/02 301 H04R 1/42

Claims (29)

(57) [Claims] 1. An input means for documenting and inputting a conversation to be trained, and a set of utterance units for using the utterances constituting the documented and input conversation for use in practical education. Editing means for converting into data relating to the start and end of each utterance unit, and the characteristics as speech in the practical training of the conversation to be trained based on the above-mentioned one set of utterance units and the data relating to the start and end thereof A combination means for composing a set of tongue and palate contact patterns, and a display means for displaying the combined set of tongue and palate contact patterns. A conversation training device based on synthesis. 2. The synthesizing means includes a three-time determination control unit for controlling the synthesis so as to determine the start and end times of each utterance unit and the duration of a stable contact state. A speech training device based on synthesis according to claim 1. 3. The three time determination control unit for each utterance unit for controlling synthesis so as to determine the start and end times and the duration of a stable state for each utterance unit. 3. The speech training apparatus based on synthesis according to claim 2, wherein the speech training apparatus has a combination. 4. The three kinds of time determination control part for each utterance unit includes, for each utterance unit, a region where the contact between the tongue and the palate is maximized, and a different context in which the utterance unit is included. 4. The synthesis-based conversation training apparatus according to claim 3, further comprising a context-reflecting small portion that enables synthesis by dividing at least one contact area according to the following. 5. The apparatus according to claim 1, wherein said synthesizing means has a 63-point expression control section for expressing a contact portion between the tongue and the palate with 63 points. A speech training apparatus based on synthesis according to claim 4. 6. An utterance measuring means for performing utterance measurement including measurement of a contact pattern between a tongue and a palate.
6. A conversation training device based on synthesis according to claim 5. 7. An evaluation means for evaluating the similarity of a tongue-palate contact pattern between an exemplary utterance in training and a measured trainee's utterance.
A conversation training device based on the described composition. 8. The synthesis-based conversation training apparatus according to claim 7, further comprising display means for displaying a result of the evaluation of the similarity. 9. A conversation input step of inputting a conversation to be trained, a conversation data creation conversion step of converting the input conversation into utterance units and start and end data of the utterance units, Tongue and palate contact pattern creation step of converting the converted conversation data into a corresponding tongue and palate contact pattern, and a display step of displaying the converted tongue and palate contact pattern. Conversation training method based on synthesis. 10. The tongue and palate contact pattern creating step includes: data indicating an area of contact between the tongue and palate for each vocal unit;
Furthermore, it has a contact area time reflection creation small step of holding data on the start and end times of each utterance unit and the duration of a stable state, and creating a contact pattern of the tongue and palate based on both data. 10. The method for training conversation based on synthesis according to claim 9, wherein: 11. The contact area time reflection creating sub-step includes, for each utterance unit, a region where the tongue and the palate come into contact with each other and a different context in which the utterance unit is used. The data of at least one other contact area is retained, and from this data and data on the start and end times of each vocal unit and the duration of the steady state, the tongue and palate contact depending on the context 11. The synthesis-based conversation training apparatus according to claim 10, further comprising a context reflecting small step for creating a pattern. 12. The tongue and palate contact pattern creating step is to hold each contact area between the tongue and palate as data comprising 63 points as data indicating an area of contact between the tongue and palate in each vocal unit. 12. The conversation training method based on synthesis according to claim 9, wherein the method has a small step of holding dot data. 13. An input means for documenting and inputting a conversation to be trained, an editing means for converting the input conversation into a phoneme set for use in practical education, and Sound parameter synthesizing means for converting into a sound parameter set that characterizes the speech in the implementation education of conversation, and converting the converted sound parameter set into an articulatory parameter set related to the operation of a vocal organ required for the utterance. Articulation parameter synthesizing means for converting, operation display means for displaying the operation of the vocal organ based on the converted articulation parameter set, and synthesized speech conversion means for converting the converted sound parameter set into synthesized speech output. A speech creation means for creating speech that can be heard by a hearing-impaired person from the converted synthesized speech output; A means for measuring a trainee's tongue contact pattern when a student utters; a voice measurement means for measuring a trainee's voice; and an air flow measurement means for measuring a trainee's exhaled air flow. A nasal vibration measuring means for measuring the trainee's nasal vibration; a neck vibration measuring means for measuring the trainee's neck vibration; and a measurement value obtained from the trainee and an exemplary model for practical education. A conversational training apparatus based on synthesis, comprising: comparison means for comparing and evaluating similarities with sound parameter sets of various utterances; and comparison display means for displaying a result of the comparison and evaluation by the comparison means. . 14. An input means for documenting and inputting a conversation to be trained, an editing means for converting the input and documented conversation into a phoneme set for use in practical education, and the converted phoneme set. And a synthesizing means for converting the sound parameter set into a sound parameter set that characterizes an exemplary speech in the practical education of the conversation of the training target. A speech training apparatus based on synthesis, comprising: articulation parameter synthesizing means for converting to an articulation parameter set; and display means for displaying the operation of a vocal organ based on the converted articulation parameter set. . 15. The speech training apparatus based on synthesis according to claim 14, further comprising formant synthesis means for converting the converted sound parameter set into a synthesized speech output. 16. The apparatus according to claim 14, further comprising a synthesized voice output means for outputting an audible synthesized voice.
16. A conversation training apparatus based on synthesis according to claim 15. 17. The speech training apparatus based on synthesis according to claim 14, further comprising articulation parameter measuring means for measuring an element related to an articulation parameter accompanying the utterance of the trainee. 18. The speech training apparatus based on synthesis according to claim 18, wherein said articulatory parameter measuring means includes a tongue contact pattern measuring unit for measuring a tongue contact pattern. 19. The speech training apparatus based on synthesis according to claim 17, wherein said articulation parameter measuring means has a sound measuring section for measuring sound. 20. The speech training apparatus based on synthesis according to claim 17, wherein said articulation parameter measuring means has a discharge air flow measuring section for measuring a discharge air flow. 21. The speech training apparatus based on synthesis according to claim 17, wherein said articulation parameter measuring means includes a nasalized vibration measuring unit for measuring a nasalized vibration. 22. The speech training apparatus based on synthesis according to claim 17, wherein said articulation parameter measuring means has a neck vibration measuring section for measuring neck vibration. 23. Similarity comparison evaluation means for comparing and evaluating the similarity between a sound parameter set of an exemplary utterance of a conversation in a practical education and a sound parameter set associated with the utterance of a trainee. 18. A conversation training apparatus based on synthesis according to claim 17. 24. A comparison display means for comparing and displaying the similarity between the sound parameter set of the exemplary conversation in the practical education and the value of the sound parameter set measured when the trainee utters. 24. A conversation training apparatus based on synthesis according to claim 17 or claim 23. 25. Documenting and inputting a conversation to be trained, converting the input conversation to a phoneme set for use in practical education, and converting the converted phoneme set to a conversation to be trained. A sound parameter synthesizing step for converting into sound parameters characterizing as speech in practical education; a synthetic sound output step for converting the converted sound parameters into a synthetic sound output; and a hearing impaired person based on the converted synthetic sound output. A speech creation step for creating an audible speech, an articulation parameter synthesizing step for converting the converted sound parameters into an articulation parameter set related to the operation of a vocal organ required for the speech, and the converted articulation parameters. To convert articulation into a set of articulation parameters related to the behavior of the vocal organs required for its utterance A parameter set synthesizing step, an operation displaying step of displaying the operation of the vocal organ based on the converted articulatory parameter set, and measuring a tongue contact pattern of the trainee when the trainee utters a conversation to be trained A tongue contact pattern measurement step, a vocal measurement step to measure the trainee's vocalization, and a discharge air flow measurement step to also measure the discharge airflow of the trainee, and a nasal vibration of the trainee. Nose sounding vibration measurement step, which also measures the trainee's neck vibration, and similarity between the typical utterance sound parameter set and the measurement values obtained from the trainee in the practical education A similarity comparison evaluation step of comparing and evaluating, and a comparison display step of displaying a result of the step of performing the comparison evaluation. Basic conversations training method the synthesis that. 26. An input step of documenting and inputting a conversation to be trained, an editing step of converting the input and documented conversation into a phoneme set for use in practical education, and converting the converted phoneme set. A sound parameter synthesizing step of converting into a sound parameter set characterizing as an exemplary voice in the practical training of the conversation to be trained; and articulation relating to the operation of the vocal organ necessary for the utterance of the converted sound parameter. A speech training method based on synthesis, comprising: an articulatory parameter synthesizing step of converting to a parameter; and an operation displaying step of displaying an operation of a vocal organ based on the converted articulatory parameter set. . 27. A synthesized voice conversion step of converting a sound parameter set into a synthesized voice output, a voice generation step of generating a voice audible to a hearing-impaired person from the converted synthesized voice output, and a trainee as a training target. An utterance measurement step for measuring the tongue contact pattern, utterance sound, exhalation air flow, nasal vibration, and neck vibration when uttering a conversation corresponding to the input document, and sound parameters of an exemplary utterance in practical education. A comparison step of comparing and evaluating the similarity between the set and a measured value of the sound parameter set when the trainee utters, and a comparison display step of displaying a result of the comparison evaluation in the comparison step. 27. A conversation training method based on synthesis according to claim 26. 28. An input step of inputting a document in a language to be learned, an editing step of converting the input document into a phoneme set, and converting the converted phoneme set into an input document of a foreign language to be learned. A sound parameter set synthesizing step of converting the sound parameter set into a sound parameter set characterizing the utterance; A synthesis-based conversation training method, comprising: a synthesis step; and an operation display step of displaying an operation of a vocal organ based on the converted articulation parameter set. 29. A synthesized speech output step for converting a sound parameter into a synthesized speech output; a foreign language speech creation step for creating a foreign language speech that can be heard by a trainee from the converted synthesized speech output; When uttering a document to be learned in a foreign language, a foreign language utterance measurement step for measuring the tongue contact pattern, vocal sound, exhaled air flow, nasal vibration, and neck vibration, A comparison step of comparing and evaluating the similarity between the sound parameter set of the input document and the measured value when the trainee utters, and a comparison display step of displaying a result of the comparison evaluation in the comparison step. 29. The method for training conversation based on synthesis according to claim 28.