KR100334127B1 - Automatic translation apparatus and method thereof - Google Patents

Abstract

The present invention discloses an automatic translator and translation apparatus and method for automatically recognizing the sound of each language using the individual parameters extracted in units of ortho-phonemes and automatically converting them to the sound or letters of the desired national language. The apparatus according to the present invention comprises: a band-to-band ratio detection unit that measures a pronunciation of a genuine triumph of an arbitrary user and detects a ratio of bands of each sound; A reference sound determination unit for determining a reference sound value for an arbitrary user by using the ratio of the respective aggregate sound bands to the genuine triumphs transmitted from the aggregated sound tones ratio detector; And a voice recognition unit for recognizing a voice of an arbitrary user in regular phoneme units using the reference sound value determined by the reference sound determination unit.

Therefore, it is not only effective to expand the range of use of the automatic interpretation and translation device, but also to communicate with everyone in the world using the native language in all communication activities including the Internet.

Description

Automatic translation apparatus and method

The present invention relates to an automatic interpretation and translation apparatus and a method thereof, and more particularly, to an automatic interpretation and translation apparatus and method for automatically recognizing and converting a sound of a language into a desired language or character.

So far, voice recognition technology does not provide a parameter normalization method that can overcome individual voice differences. Because of this, currently proposed speech recognition systems are only dependent systems capable of recognizing only certain speech. Due to the limitation of the speech recognition technology, the apparatus implemented to perform the interpretation and translation processing by recognizing the existing speech has a very limited use range.

The present invention provides an automatic translator and translation device and method for automatically converting the sound of each language into a static sound or a character of a desired national language using individual parameters extracted in units of ortho-phonemes. There is a purpose.

1 illustrates a three-minute profit and loss method for three rates.

2 is an inverse distribution map of the consonants.

3 is a functional block diagram of an embodiment of an automatic interpretation device according to the present invention.

FIG. 4 is an example of an inverse distribution diagram illustrating the operation of the aggregated band ratio detector and the reference sound determiner illustrated in FIG. 3.

FIG. 5 is another embodiment of an inverse distribution diagram for explaining the operation of the aggregated band ratio detector and the reference sound determiner shown in FIG. 3.

FIG. 6 is an exemplary view of a result translated through a first translation unit and a second translation unit illustrated in FIG. 3.

7 is a functional block diagram of an embodiment of an automatic translation apparatus according to the present invention.

8 is a functional block diagram of another embodiment of the automatic interpretation device according to the present invention.

9 is a configuration example of an application system to which the automatic interpretation / translation apparatus according to the present invention is applied.

In order to achieve the object of the present invention, the apparatus according to an embodiment of the present invention, in order to measure the pronunciation of the genuine triumph of any user, the average of a plurality of bands formed in the second sound width for each genuine triumph A aggregated band ratio detector for detecting a second value corresponding to an average of a plurality of bands formed in the first value corresponding to the third and third sound widths as a ratio of the aggregated bands; A reference sound determination unit configured to determine a reference sound value using a third value corresponding to an average of first and second values of each of the pure trine transmitted from the aggregation band ratio detector; It is preferable to include a voice recognition unit for recognizing any user's voice in the regular phoneme unit by using the reference sound value determined by the reference sound determination unit. In addition, the apparatus according to another embodiment of the present invention, in order to measure the pronunciation of the genuine triumph of any user, the second band formed in the first and third aggregate bands formed in the second aggregated band for the pure triumph A aggregated band ratio detector for detecting at least one first value corresponding to an average of the liver as the aggregated band ratio; A reference sound determination unit configured to determine a reference sound value using a third value corresponding to an average of at least one first value of each of the genuine triodes transmitted from the aggregation band ratio detector; It is preferable to include a voice recognition unit for recognizing any user's voice in the regular phoneme unit by using the reference sound value determined by the reference sound determination unit.

The automatic translator / translation apparatus may include: an all-round ambassador for storing character information for all-language languages arranged in alphabetical order; A first translator which breaks down the recognition information transmitted from the voice recognition unit in units of words and translates them into letters of a national language corresponding to the pronunciation of a user by referring to the entire country's ambassador; A second translator which breaks down the phonetic symbols transmitted from the first translator in word units and translates them into letters of a desired national language with reference to the entire world ambassador; It is preferable to further include an output means for outputting to recognize the character transmitted from the second translation unit.

In order to achieve the object to be achieved by the present invention, the method according to the present invention, in order to detect the ratio of the aggregate band to the genuine triad of any user, to the average of a plurality of bands formed in the second sound amplitude for each of the genuine triumph Detecting a corresponding first value and detecting a second value corresponding to an average of a plurality of sound bands formed in the third sound width for each genuine triode; Determining a reference sound value using a third value corresponding to an average of each of the first value and the second value of pure three; Recognizing speech provided from an arbitrary user on a regular phoneme level using a reference phonetic value; It is preferable to include the step of translating the voice information recognized in the unit of the phoneme to a constant sound of the desired national language.

The output may include forming information recognized in a phoneme unit in syllable units; a first translation step of breaking information formed in syllable units into word units and translating the information into a national language used by an arbitrary user; A second translation step of translating the sound into a static sound of a desired national language when the translated result is transmitted in the step; Constructing sentences translated from the second translation step into sentences; It is preferable to include the step of outputting the sentence to be recognized in the desired national language.

Hereinafter, with reference to the accompanying drawings will be described in detail the present invention.

Fig. 1 is a three-profit method for 12-Yul-Lee, organized by the maintenance of musical instruments and musical scores in Sejong University. The three-profit and loss method was applied to the creation of Jeong-eum, and the pure rate of each letter was determined. . It is observed that the pure rate matches well when the frequency of the two tones is an integer ratio, i.e. 1: 1, 1: 2, 3: 2, 3: 4, and falls off when it is not an integer multiple. It is the harmony rate that originated.

The three stars (·, ㅡ, ㅣ), which are the basis of vowels, have a relationship of 2/3: 1: 4/3 with the cross-sectional ratio of saints according to the above-mentioned three-profit method. That is, if the above-described acoustic frequency of the three stars is analyzed, as shown in FIG. 2, the position 205 of the position 205 of the third sound width based on the position 203 of '-' as a reference (1 (3/3)) is 205. ) Is present and '·' is present at the position 201 of 2/3 times. In this connection, it can be seen that the vowel 3 is based on the basic pure rate of 12 rate.

In addition, as shown in FIG. 2, when the second and third sound widths of the vowel three stars described above are simultaneously viewed, in the case of the '·' sound, the second sound width and the band (or node) near the highest 710 Hz among the three vowel stars are shown. Is formed, and the third band is formed around the lowest 1100 Hz among the three vowels, and in the case of the '-' sound, which is the reference sound, the band is formed around 550 Hz in the second and around 1500 Hz in the third. In the case of the '' 'sound, the band is formed at around 400 Hz in the second sound width, and the band is formed at around 2100 Hz in the third sound width. By connecting the six sound bands described above, a triangular composition of inverse symmetry (or inverse) is obtained as shown in FIG. 2, and the speech recognition technique according to the present invention is based on the theory that such composition is applied to all consonants and vowels. It is done.

FIG. 3 is a functional block diagram of an embodiment of an automatic interpretation apparatus according to the present invention, wherein the aggregated band ratio detection unit 301, the reference price determination unit 303, the voice recognition unit 305, and the first translation unit 307 are shown. ), A universal translator 309, a second translator 311, a sentence constructor 313, and a voice synthesizer 315.

The aggregate band ratio detection unit 301 measures the pronunciation of each of the genuine triumphs (·, ㅡ. ㅣ) provided by the user who will provide the first voice to detect a corresponding aggregate band ratio. That is, in the example shown in FIG. 4, 401 becomes the aggregated band in the third volume for '·', and 402 is the aggregated band in the second volume for '·', and (411) Is the aggregated band at the third loudness for 'ㅡ', (412) is the aggregated band at the second loudness for 'ㅡ', and (421) is the aggregated band at the third loudness for 'ㅣ' And 422 is the aggregated band in the second tone for 'ㅣ'. As can be seen from FIG. 4, each aggregated band consists of a plurality of notes, which are generated each time a pronunciation of the vowel is input. The aggregate band ratio detection unit 301 detects an average value of frequencies of the tone bands formed in each of the aggregate bands, and outputs the detected average value as a standard aggregate band value of the aggregate band. Therefore, the aggregate band ratio detection unit 301 outputs six standard aggregate band values. The output standard aggregate band value is transmitted to the reference sound determiner 303.

The reference sound value determination unit 303 detects an average value of the standard aggregated band values for pure trine, respectively, and obtains frequencies for the bands 403, 413, and 423, and again calculates average values of the bands 403, 413, and 423. This average value is the frequency of the reference sound value for the user providing the first voice. Information about the frequency of the reference sound value is provided to the speech recognition unit 305.

FIG. 5 is another embodiment of an inverse distribution diagram for explaining the operations of the aggregated band ratio detector 301 and the reference price determiner 303 shown in FIG. When applied, the aggregated band ratio detection unit 301 detects the tone bands 501, 502, 511, 512, 521, and 522 formed in the third and second tone widths of the genuine triodes, respectively, to the reference price determiner 303. to provide. The reference sound value determination unit 303 detects averages between corresponding sound bands, respectively. After performing the operations of the aggregated band ratio detector 301 and the reference price determiner 303 a plurality of times, the reference tone determiner 303 is purely near the frequency band of the corresponding reference value as shown in FIG. 5. The information of the aggregated bands 503, 513, and 523 for each of the three sounds is present. The result of detecting the average value of the average values of the aggregated bands 503, 513, and 523 is used as the frequency of the corresponding reference sound. Provided at 305.

The speech recognition unit 305 configures the inverse distribution of the consonants for the user providing the first speech using the frequency of the reference sound value provided by the reference sound price determiner 303 as shown in FIG. Speech recognition processing in units of ortho-phonemes is performed. The regular phoneme is an independent phoneme unit based on the genuine vowel 3 and the consonant 7 in the phonebook shown in FIG. The information extracted in the phoneme unit is transmitted to the first translation unit 307.

The first translation unit 307 forms a recognition result applied in phoneme units in syllable units, and then breaks it in word units to extract a character of a language corresponding to the first voice by referring to a corresponding region of the universal metabolism unit 309. do. For example, when the first voice is a Chinese pronunciation as shown in FIG. 6 (a), the phonetic notation of FIG. When the Chinese characters are extracted as indicated in () at the bottom, the first translation unit 307 outputs the phonetic notation (development phase) existing at the top as a translation result. If the first voice is English pronunciation as shown in (b) of FIG. 6, the first voice is referred to at the bottom of the phonetic notation of FIG. As shown in (), when the English type of characters is extracted, a phonetic notation such as 'Wo Phraspode ...' is output as a translation result of the first translation unit 307. The translation result is transmitted to the second translation unit 311.

The whole country ambassador 309 is implemented to store information about the language characters of the world in alphabetical order. The storage structure can be implemented by dividing the storage area for each country.

The second translator 311 refers to a phonetic alphabet transmitted from the first translator 307 on a word-by-word basis by referring to the language region of the corresponding country of the international ambassador 309. Translate to. For example, when the second voice is Korean, the second translator 311 refers to the translation of the phonetic transcription provided from the first translator 307 with reference to the Korean language in the storage area of the universal metabolism 309. Translate into Korean. For example, when the first voice is Chinese and the second voice is Korean, as shown in FIG. 6 (c), the translation result transmitted from the first translation unit 307 becomes the top notation information at the top and the second translation. The translated result in the part 311 becomes the contents of 'four to four ...' indicated in (). In addition, when the first voice is English and the second voice is Korean, the translated result is extracted as shown in FIG. The extracted translation result is transmitted to the sentence constitution unit 313.

The sentence constitution unit 313 composes a speech by synthesizing the translation result transmitted from the second translation unit 311 in a word unit in a sentence form (the same form as indicated in () of FIG. 6) corresponding to the second voice. The voice synthesizer 315 is transmitted to the unit 315, and the voice synthesizer 315 is operated as an output unit for performing a voice synthesis process so that a language transmitted in the form of a sentence can be output in the form of a second voice.

FIG. 7 is a functional block diagram of an automatic translation apparatus according to an embodiment of the present invention, wherein the aggregated band ratio detector 701, the reference sound determiner 703, the speech recognizer 705, and the first translator 707 are provided. ), A universal translator 709, a second translation unit 711, a sentence constitution unit 713, and a display unit 715.

Among these components, the aggregated band ratio detection unit 701, the reference tone determination unit 703, the voice recognition unit 705, the first translation unit 707, the universal metabolism unit 709, the second translation unit 711, and the like. The sentence constitution unit 713 includes the aggregated band ratio detection unit 301, the reference tone determination unit 303, the voice recognition unit 305, the first translation unit 307, the universal metabolism unit 309, and the like. Since it is configured and operated in the same manner as the second translation unit 311 and the sentence configuration unit 313, description thereof will be omitted to avoid redundant description.

The display unit 715 is operated as an output means for performing signal processing so that sentences of the language corresponding to the second voice provided from the sentence constructing unit 713 can be sequentially displayed as indicated in () of FIG. 6.

8 is a functional block diagram of another embodiment of the automatic interpretation apparatus according to the present invention, which includes a filter 801, a voice recognition unit 803, a first translation unit 805, a universal metabolism unit 803, and a second unit. It consists of a translation unit 809, a sentence construction unit 811, and a speech synthesis unit 813. Except for the dual filter 801 and the speech recognition unit 803, the other components are configured and operated in the same manner as shown in FIG.

The filter 801 detects the vocal tract transfer function of the user providing the first voice by using the user's voice to provide the first voice and the vibration value of the vocal cords. That is, if the voice applied to the filter 801 is x (t), the vibration value of the vocal cords is y (t), and the vocal tract transfer function is z (t), the vocal tract transfer function z (t) is The speech x (t) and the oscillation value y (t) of the vocal cords are obtained by deconvolution. When this equation is made in the frequency domain, the vocal conduction transfer function Z (W) in the frequency domain can be obtained as X (W) / Y (W) in the frequency domain. The use of these vocal fold transfer functions is based on the theory that the vibration of the ulna (vocal cord) becomes the sound source and the saints modulate it to form a waveform of sound that spreads out of the lips.

The voice recognition unit 803 performs recognition processing on a regular phoneme basis for the first voice based on the vocal tract transfer function provided from the filter 801.

9 is a configuration example of an application system to which the automatic interpretation / translation apparatus according to the present invention is applied, in which subscribers using different languages make calls through an exchange using their native languages.

That is, in a situation where a subscriber using the first subscriber station 901 is a subscriber who uses Korean, and a subscriber who uses the second subscriber station 903 is a subscriber who uses English, voice is transmitted through the first subscriber terminal 901. When input, it is transmitted to the automatic interpreter / translator 907 through the exchange 905, the automatic interpreter / translator 907 converts it into a static sound corresponding to English, and transmits it to the exchange 905, and the exchange 905 ) Provides it to the second subscriber station 903. In this case, the automatic interpreter / translation apparatus 907 is configured and operated as shown in FIG. 3, 7, or 8. The above-described exchange 905 may be implemented as a comprehensive communication network including the exchange.

As described above, the present invention extracts each individual parameter in the unit of phoneme, and enables the independent speech recognition of the phoneme unit by using the parameter, and recognizes the voice recognized by referring to the information of the universal metabolism arranged in the phonetic symbol. By translating into the language of the desired country, and providing the automatic interpretation and translation device and method for outputting the sound or text of the country, the effect of expanding the scope of use of the automatic interpretation and translation device is only effective. Rather, it has the effect of facilitating communication with everyone in the world using the native language in all communication activities, including the Internet.

Claims (11)

delete In order to measure the pronunciation of a genuine triumph of an arbitrary user, a first value corresponding to an average of a plurality of bands formed in a second sound amplitude and a mean of a plurality of sound bands formed in a third sound width for each of the genuine triumphs An aggregate band ratio detector for detecting the second value as a ratio of the aggregate bands; A reference sound determination unit configured to determine a reference sound value using a third value corresponding to an average of the first value and the second value of each of the genuine triodes transmitted from the aggregation band ratio detector; And a speech recognition unit for recognizing the voice of the arbitrary user in the unit of a regular phoneme by using the reference sound value determined by the reference sound price determiner. In order to measure the pronunciation of any user's genuine triumph, the ratio of the first to the average value between the first band formed in the second aggregated band and the second band formed in the third aggregated band for the pure triumph A band-to-band ratio detection unit for detecting at least one; A reference sound determination unit for determining a reference sound value using a third value corresponding to an average of the at least one first values of each of the genuine triodes transmitted from the aggregation band ratio detector; And a speech recognition unit for recognizing the voice of the arbitrary user in the unit of a regular phoneme by using the reference sound value determined by the reference sound price determiner. According to claim 2 or 3, wherein the automatic interpreter / translation device, A universal ambassador for storing character information for a universal language, arranged in alphabetical order; A first translator which breaks down the recognition information transmitted from the voice recognition unit in the unit of the phoneme by word unit and translates the recognition information into a character of a national language corresponding to the pronunciation of the user by referring to the universal metabolism unit; A second translator which breaks down the phonetic symbols transmitted from the first translator in word units and translates them into letters of a desired national language with reference to the universal metabolism unit; And an output means for outputting the text transmitted from the second translation part so as to recognize the text. The method of claim 4, wherein the output means, And an audio synthesizer configured to perform a voice synthesis process so that the text transmitted from the second translation unit is output as a voice corresponding to the desired national language. The method of claim 4, wherein the output means, And a display unit which performs a display process so that the characters transmitted from the second translation unit can be output as sentences corresponding to the desired national language. The automatic translator / translator of claim 4, wherein the entire country's ambassador has a structure in which storage areas are separated for each language. In an environment in which subscribers who speak different languages can communicate through a communication network, A voice recognition unit recognizing a voice in a unit of regular phoneme by using a reference sound value for the first subscriber when a voice of a first subscriber is introduced through the communication network; All the ambassadors of the world to store the language of the alphabet in the order of the sign A translator that breaks the recognition result of the phoneme unit transmitted from the voice recognition unit into word units and translates the language into a language of a country used by the first subscriber and the second subscriber to be called by referring to the universal metabolism unit; And an output means for outputting the result translated by the translation unit to be transmitted to the second subscriber through the communication network. delete In order to detect the ratio of the aggregated band to the genuine triumph of any user, a first value corresponding to the average of a plurality of bands formed in the second sound width for each genuine triumph is detected, and a first value is set for each genuine triumph. A clustered band ratio detection step of detecting a second value corresponding to an average of a plurality of bands formed in three sound widths; Determining a reference sound value using a third value corresponding to an average of the first value and the second value of each of the pure threes; Recognizing the voice provided from the arbitrary user by unit of a phoneme using the reference sound price; Automatic translation and translation method comprising the step of translating the speech information recognized in the unit of the phoneme to a regular sound of the desired national language. The method of claim 10, wherein the output step, Forming information recognized in units of syllables in syllable units; A first translation step of breaking the information formed by the syllable unit into word units and translating the information into a national language used by the user; A second translation step of translating the sound to the sound of the desired national language when the translated result is transmitted in the first translation step; Constructing sentences translated from the second translation step into sentences; And outputting the sentence to be recognized in the desired national language.