TW580651B - Language learning system and method using visualized corresponding pronunciation suggestion - Google Patents

Abstract

The present invention provides a language learning system and method using visualized corresponding pronunciation suggestion, which employs the sound wave diagram in the language sample corresponding to the voice output to notify the user for gradually modification on high/low of the pronunciation and the light/heavy control of pronunciation; and, effectively improving the speaking capability of the user in the language learning through the methods of segmented evaluation and referring modification. The system according to the present invention includes a language database, a follow-reading learning module, a display control module, a segmented evaluation module, a reference and modification module, and a prompt editing module. The present invention includes the following steps: first, abstracting the language sample in whole sentence; outputting the whole-sentence voice message and the whole-sentence combined message; next, prompting the user to emulate the follow-reading, and recording the voice data; conducting the comparing procedure based on the voice data and outputting the similarity value of the whole-sentence voice; finally, verifying the voice data.

Description

5δ〇65ΐ l, invention is described in (1) Field of the invention [^] belongs to the present invention is a sense of language learning system prompts the corresponding pronunciation of language learning [prior art] under the convenient traffic trends in the world 'frequent language communication skills develop potential school education and cram under the system of division 'in the classroom with learners through Xu Xi's interest in learning, so that students dare division sidelines close supervision, but often makes supervise the learner to pronounce effects for every barriers pronunciation for outward learners' teacher standard pronunciation also belong to the main registration 'easily lead to poor performance in, button 0, and methods' particularly to system and method of use. Communication opportunity between people will be increasingly will become the trend of the future of education. Shang interactive language learning system produced by one or more evening old way, and thus stimulate learning = I said, at this same time, the number of older students by far more than the number of teachers, with also affected, can not be achieved for a clear fast corrective i, in addition the concept of determining vary, interest beneath a large scale virtually learner's computer technology it discount bad ...... Cloth, T w Gongben [sigma], of: 'the human voice CD and various electrical "with only limited overall :: exemplary sentence imitation and c 'to other early learning function, such that school:' t: cognitive modulation, to study their effect as;" (iv) a sheet from the pronunciation learners in the learning process familiar to many outside: of: this τ, or maybe listen to: * training effect 'then again with an unusual language families' Yong' :: also reached the debut performance to pronunciation 1 pick the language for H: sleeve: 9% could not learn the fine put bottleneck s

Page five 5580651580651 invention will be described (2) to a ΐ ^ Xi will fall into parrot, dazzling merchandising war when an effective system and method. (Iv) improved by the + and pronunciation but the kick sound waveform shown in FIG chime tone sounds in the BU 1, 1 ^ Shi said calculating a quantization rendered through computer graphics, the weight of Laid + in high and low tones and the tone of light path 2 production engineers can sound information, the shape difference recording work, in order to achieve higher recording quality. And Summary of the Invention = currently on the market in any system which does not have the use of this method and one white. In view of this, the present invention is to solve the above-mentioned problems: the language learning system corresponding to the pronunciation prompts = and == visual display of the sound waveform diagram that makes the voice output, and is used to extract and evaluate through segmentation and comparison correction mode correction, learning the ability to "say." 'There jamb Jian Zheng Yi in the user language split :: To achieve the above object, the present invention benefits mentioned = language learning system' comprising: said language data = not significant control module, segment assessment module to control repair : white modules, modules. u-hwan is set and tips compiled a lucky i put

580651 five invention is described in (3) and repeat copy data included in the speech, the speech sentence and then order and outputs a similarity value, P f for the speech data using the visual comparison of the process of the present invention corresponds to voice data sent Γ ^ identified. The purpose law, may by the tone of voice and language learning system and the severity of side presented, in order to allow the user to see the heart-shaped map to achieve the objective differences in visual voice messages, and language sample into said record

Jr ρ; ΤΙ 4- I ^ ♦ stone itself Cui user tone correction scheme and the severity of, to achieve the most effective known Embodiment [correcting said purpose. Moon system and the system as a ΪΓ use visual prompts the corresponding pronunciation of language learning system "%, specific examples illustrate the feasibility of the present invention will next Tony Shi, see Figure 1", which is described as follows: (η corpus 100, used to store language samples, where the language samples have voice messages and combined messages. Voice messages refer to the contents of language samples that are output through sound in the voice samples, and combined messages refer to sound waveforms that are synchronized with the voice messages of the language samples. maps and text prompts billion (2) repeat 2 thousand and learning modules for corpus since 1 thousand and retrieve language samples and provide output voice messaging users hope to listen, and then prompts the user to repeat it to imitate in accordance with the contents included in the user voice message repeat of the voice data is generated. (3) the display control module thousand and 3, repeat learning module for outputting the combined message 200 retrieved and included in the user After the voice message, a corresponding sound wave pattern is output according to the pronunciation of the user.

580,651 V. invention is described in (4) refers to information corresponding to the sound waveform diagram of voice messages and text labels, presenting and providing user understand the meaning of the sentence fragments voice data through the screen, and the sound waveform diagram in accordance with the built-based live The sounds produced by the pronunciation, as well as the graphics constructed by the pauses, can be used to correct the inconspicuous problems such as the tone and weight of the pronunciation by the user through his own pronunciation and then the sound wave pattern generated in the system. This improves the user's pronunciation level. (4) assessment module segment 40 billion, according to a voice-based message as a standard for comparison for the voice data included in the user program generates pronunciation, the calculated speech sound produced in the speech samples of voice messages with the user via computer The value of the difference between the data, and then proposed a voice similarity value, which is a percentage value that indicates the difference between the voice data and the voice message, and output the entire voice similarity value to display the user's pronunciation pronunciation value of the language of each line segment to generate a similarity value of ^ generated through an averaging operation, so achievement sentence f assessment of desirable criteria (e.g., voice sentence similarity value is higher than 8 billion%), there may exist between the non- to standard segment (e.g., less than voice 8〇% similarity values) square (5) of 2, 5 thousand and correction module for the voice traffic repeated Ii similarity value determination is performed 'in accordance with its whole i.e. sentences of the speech similarity value or similarity values 2 i-generation section, thereby determining whether or not to repeat Bofang d I a and the recording of audio data to enhance the user's pronunciation water

Page 8 580 651 V. made before ⑸ invention to repeat the learning module to retrieve the speech samples, the speech message information and performs segmentation composition = annotation program, also according to the user's intention ^ = ^ multi one words constituting a whole sentence voice or a combination of several messages into only a single or a ^ ^ a voice or a combination of words constituting the message, which in practice are "inch &# Fo knife and I s samples, i.e. a subparagraph sentence pronunciation teaching materials said word pronunciation. / Cloth of next refer to "Figure 2", illustrate the present invention disclosed the use of visual wave 1 language learning should prompt the pronunciation, through the rapidly-type voice samples of the sound map with, prompt the user to carry out pronunciation correction, the method comprising the following ^ lure Baixian from corpus 100 captured sentence language sample (step 10), then 3〇 (Γ_ study with ^ 1 module 200 output sentence The voice message and the display control module output an orthodox combination message (step 20). The user can imitate and read in the textbook and collect the user's repetition language-like audio data; advance :: beib: (.step , 3〇),. 妾Assessment module 400 segments for 40 words), ( ', inflammation test " D ^, the program and outputs a speech sentence similarity value (step 5〇 step) :, f is determined whether voice data most meat does not reach standard (if this step aging WVt standard set of voice similarity value of 80%, to inform repeat learning module ^ / = / re-reading // ^ (iv);? police their own pronunciation of the sound; at the same time on this kind of tone and pause (iii) Point difference;

η Ti 9 sentence record audio messages and engagement from m (n manufactured prong 300 outputs the sentence language group

580,651 V. described invention (β) 'e- ------] ^ portion of the vehicle to be amended; also followed by the same sequence of steps until the validity period & 80% said similarity value, to complete the sentence repeat times task selection step may be segmented assessment (step 2〇), to enhance the depth. Sub from the state "process in this method, in addition to the training Bofang repeat the sentence, and the method further comprises process segment F practice, based, previously provided by the electrical system 150 prompts editing mode, the user's intention through a combination of voice messages and performing segmentation standard message: ΐΪί (refer to Annex Shu), then this system can produce the program later provide reinforcement beam ,, white spoon segmented voice messages and message segments in combination, refer to "FIG. 3 "* j first the user selects segments of speech samples (step 61〇), and then with the chin learning module outputs 200 segment verbal messages as well as message segmentation combination (step 620), which repeat the learning module by the 2〇 then the user is prompted to imitate the square with the trip% Li voice data included in the segment (step 63〇), speech followed by data for the segment, and outputs a program to compare similarity of the speech (step 64〇) (refer to Annex 4) Represents the degree of similarity between this segment and the sound waveform presented by the user's actual pronunciation, so as to identify whether the segment voice data does not meet the standard? (Step 65 billion), if not up to standard "speech segment similarity values thus only 78% (assuming the standard previously set to 80%), then the control segment based on the correction module 5〇〇 assessment module voice analysis of the similarity values 400, in order to link learning modules 2〇〇 repeat repeated Bofang this segment, as well as more than a collection of steps until this segment similarity values up to standard voice, was able to complete repeat its training mission, after which the user can choose whether or not to exercise the entire sentence (step 660), to practice the rest of this language samples of, or the end of the segment repeat this training mission. The above are merely preferred embodiments in which the creation of the present embodiment it is not

580651

Page 11

Using the language learning pronunciation corresponding to prompt visualization purposes, Yan 713 prompts the feel of pronunciation corresponding to the language learning formula Brief Description of Drawings FIG 1 a schematic view of the present invention based system referred to; Fig. 2 a flow chart of the present invention based proposed method; first FIG 3 based benefits of the proposed method of the present invention, the segment repeat training stream _ visual prompts corresponding to the pronunciation of language learning system in Annex 1 of the present invention disclosed a schematic diagram of a message composition "; sample text sentence in synchronization with the voice message Annex 2 of the present invention disclosed based map; and a wheel system which is a similarity value of the speech sentence Annex 3 schematically schematic speech similarity value out of the present invention disclosed based. For voice data segment to segment assessment module controls the amount of control module 20 LV Tip Database Editor Mark me repeat learning module display module outputs a square alignment program [step 100 of FIG. SYMBOL DESCRIPTION steps 150,200,300,400,500 correction module to retrieve a sample of the output of each language sentence a sentence voice message and a message prompting the user to sentence combinations repeat and imitate the voice data included in a

580651 drawings briefly described a step 40 and the whole alignment program step 50 outputs a speech sentence similarity value of the speech data does not reach the standard for assessment step 60 the amount of segmented step 610 for selecting a language of the speech segment data a voice message segment a segment, and step 620 samples combining step 630 outputs a message prompting the user to imitate the repeat step 640 and included in a segment for which the speech segment data for a voice data output and a specific program step 650 speech segment speech similarity value of the data have not reached the step 660 for the entire sentence practice

Page 13

Claims (1)

Sixth, the scope of patent Lee f 1 • one visual prompts corresponding to the pronunciation of language learning system, a language-based content output samples of sound waveform corresponding to one repeat FIG provide a user, thereby prompting the user according to the assessment of pronunciation correcting the results, the system comprising · a language database for storing the language sample, wherein the sample has a language and a voice message composition message; - repeat learning module 'from the language used for information; captures the sample taken language and outputting the voice message, the voice and the latter included a display resource control message to take a combination of tone FIG type; a segment of the voice commentary sound resource similarity was repeated with a control Shuichi 2. language learning patent Noriyuki systems, and complex 3-word sample as patent Fan of language learning system, now used in the learning module for outputting the repeat of the day ^ D ", the heart muscle suddenly hearing the heart, and to repeat the learning mode The group records the phrase, and outputs the corresponding sound wave control group according to the voice message. Based on the voice message, Alignment programs, and the wheel so as to display a sentence words to provide a user correction value pronunciation; and f modules to perform the read job determined based on the voice similarity values. Using visual around the first item corresponding to the pronunciation of the language sample extract wherein six lines includes a plurality of segments Chile venturi language sample. ... n ~ surrounded with a visual of the pronunciation corresponding to the first one mentioned 3: 5 5 Hyun system further comprises a prompt editor module before selling 4 * s ^ Mark me module fetches the samples, the 580651 Sixth, the scope of patent non-prompt message and the message prompt implementation of all points label program. 4. utilizing visual prompts corresponding to the pronunciation of language learning, the speech samples in a line through one of the sound waveform diagram, to prompt the user for pronunciation correcting 'the method comprising the steps of: capturing a sample language sentence; outputs a voice message sentence and a sentence combinations · News fortunate to prompt the user to imitate and repeat it included a voice-owned ^ be the voice data for a specific program and outputs a voice sentence similarity value; and I mouth Verify that the voice data meets the criteria. 5. The use of the visual item 4 of the patent application range corresponding Wu s learning pronunciation of tips, wherein the step of the predetermined clock information is less than the voice 'to the normal time, repeating the steps of: ..... outputs a sentence voice message and a combined sentence message; prompting the user to imitate and repeat it included a sub-vertical shoots, '10 the Wu sound data were compared to a program and outputs a voice Jiao similarity value; and positive j identify the voice data ports up to standard. 6. If the application using the visual item 4 of patentable scope of the library Yi Li language learning method, wherein the step of identifying the speech to ten voice V: V up after said standard, the method further comprises a segment assessment a step for correcting the sentence language pronunciation practice the sample fragments, the linear step process comprising: selecting a segment of speech samples; 580651 Outputs a sub-verbal messages as well as a good-news segments combined to prompt the user to imitate and repeat it included - points ^ ^-owned language program for performing a comparison of the segmented voice data and outputs a voice similarity value; identification The segmented voice material meets the standard. The patent application range of the of language learning method using said utterances visual prompts for the item 6, wherein the step of identifying the segment language - the sound%% bucket does not ^ reached the standard, repeat steps:, " output a divided respectively segment-verbal messages and take a segmented portfolio News · prompt the user to imitate and repeat it included a phrase ^ segmented data; said, be a take off clothes sequence comparison and outputs a voice similarity value for the segmented voice data; identification The segmented voice material meets the standard. Page 16