TWI432179B - Interactive speech testing and training platform - Google Patents

Description

Interactive voice test and training platform

The invention relates to speech testing and training, in particular to an interactive speech testing and training platform.

In the medical field, it is often necessary to perform hearing tests on people with hearing impairments, especially due to the aging of the society, and the hearing-impaired population is increasing. In the past, most hearing-impaired people must wear hearing aids to improve hearing. Therefore, it is necessary to understand the hearing of the hearing impaired, and then design a hearing aid according to the hearing situation, in order to effectively improve the communication ability of the wearer.

The general hearing test plays different sounds in the hearing test room, allowing the subject to listen and press the control button to respond to the sound heard according to his or her feelings. Then, the examiner determines the hearing condition of the subject according to the hearing response of the subject.

Based on current medical knowledge, we understand that listening ability is likely to be rehabilitated if properly trained. After rehabilitation training, it can effectively improve the patient's listening ability. For those who are mildly hearing loss, they do not have to rely entirely on hearing aids. And the ability of the hearing aid has its limits, and if it is supplemented by listening and rehabilitation training, it will benefit the patient.

The purpose of the present invention is to propose an interactive voice test and training platform, in which the test subject can be repeatedly tested and can also achieve the purpose of hearing rehabilitation. In this case, the method of using near-tones is used to realize the test of the voice by means of an electronic device, so that the testee can be easily tested, and repeated training can be performed for the wrong result, thereby enhancing the hearing. Furthermore, the selection of the near sounds in the present invention is mainly based on double words. The frequency of occurrence of the phrase is determined by the frequency of use of the phrase, the frequency of use in daily life, and the degree to which the voices are close to each other.

In order to achieve the above object, an interactive voice test and training platform is proposed in the present invention, which comprises: a close-up vocabulary, storing words of different similar sounds and all similar sounds corresponding to the words, each constituting the vocabulary One unit; the first equalizer is used to convert an input voice into electronic voice, the so-called electronic voice refers to the voice played from a general electronic instrument; a selector is similar in accordance with the default selection logic A test title is selected from the sound vocabulary or the similar sentence sentence library; a player plays a voice or video according to the result selected by the selector; the player is divided into a voice play and a video play; the voice play is played in a voice form The result selected by the selector, the audio and video playback is mainly voice plus the playback image; a receiver is used to receive the correctness of the result replied by the subject; mainly input by the subject; a scorer, After the input received by the receiver is compared with the built-in answer, the correctness of the result of the respondent is determined, and after each reply result is accumulated, The scorer includes a pass-through mechanism to determine whether the score obtained by the subject is passed, and if so, the test is ended. If not, return to the selector and re-test the wrong question. Only the above exclusion logic can be applied at this time, and other different near or similar sentences are selected from the same unit for testing to avoid affecting the test result due to the memory of the subject.

The features of the present invention and its advantages are further understood from the following description, and reference is made to the accompanying drawings.

In view of the structural composition of the case, and the functions and advantages that can be produced, in conjunction with the drawings, a preferred embodiment of the present invention is described in detail below. Please refer to Figure 1 and Figure 2 for the interactive and sound testing and training platform of this case.

The present invention includes a database containing the following secondary databases.

a close-to-sense vocabulary 10, storing words of various similar sounds and all similar sounds corresponding to the words, thus forming a unit in the lexicon, if desired, tens of thousands; use, eat; this year, new year, Heart, newcomer; and so on. Among them, the selection of close sounds is mainly based on double words, because the similar sounds of double words are more difficult to distinguish. For each unit, according to the representativeness of the phrase, the frequency of use in daily life, and the degree to which the voices are close to each other determine the frequency of occurrence of the phrase.

A close-to-sentence sentence library 20 stores various sentences and stores the similar words of the word for a word in the sentence. For example, "ㄒㄧㄣ ㄋㄧㄢˊ 要 要 要 要 要 要 ” ” ” ” ” ” ” ” ” ” ” , , , , , , 对于 对于 对于 对于 对于 对于 对于 对于 对于 ㄒㄧㄣ 的 ㄒㄧㄣ ㄒㄧㄣ ㄒㄧㄣ ㄒㄧㄣ ㄒㄧㄣ

The mouth shape database 30 stores mouth shapes corresponding to various pronunciations.

An expression database 40 stores expressions corresponding to various pronunciations.

Generally, if the subject is listening to the voice and supplemented by the mouth shape and expression, it may increase the accuracy of the voice recognition by 20% to 30%. Therefore, performing this test and training for the listeners helps to enhance the overall speech recognition ability.

A noise database 50 stores various noises.

The equalizer 60 is used to convert an input voice into an electronic voice, and the so-called electronic voice refers to a voice played from a general electronic device such as a telephone, a mobile phone, or the like. Therefore, the subject can be made to hear the voice from the electronic device. Using this electronic voice, the subject can be trained to recognize the electronic voice.

A selector 70 selects a test title in the near-sense vocabulary or the close-up sentence library according to a default selection logic.

Wherein the selection logic is freely selected by the tester or selected in a random manner. The selection logic also includes exclusion logic, which can exclude certain speech, such as voice that has been tested, and the exclusion no longer occurs.

After selecting the voice (ie, the close sound or the close sentence), you can choose whether to add the noise, or display it in electronic voice, or display the corresponding mouth shape or expression. When it is decided to add the noise to the selected near or close sentence, the volume of the noise can also be adjusted as needed. When the volume of the noise is louder, it is more difficult to recognize the voice doped therein. When it is decided to add a mouth shape or an expression, when playing a voice, the mouth shape or expression is also played as an image.

A test question group can be established based on the selection result of the selector 70.

A player 80 plays a voice or image according to the result selected by the selector 70. The player 80 is divided into voice playback and video playback. Voice playback plays the result selected by the selector in the form of voice. The video playback is mainly to play the image.

When playing in voice, the word is played first by voice, and the words of other similar sounds in the same unit as the word are displayed on the display screen, and the subject determines which one of the words is displayed corresponding to the voice. Expresser. For example, the tester reads: the sound of ㄒㄧㄣㄋㄧㄢˊ, and then displays the four groups of words on the screen, "This year, New Year, Heart, Newcomer". The subject decides the sound of the word as the tester. The person being pronounced and enter the answer. The test of the sentence is mainly based on the way the tester directly reads it.

A receiver 90 is used to receive the correctness of the result replied by the subject. It is mainly input by the subject.

A scorer 100 determines the correctness of the results replied by the subject according to the input received by the receiver 90 and compares the built-in answers, and accumulates the results of the responses after each time. The scorer 100 also includes a pass mechanism for determining whether the score obtained by the subject is passed, and if so, ending the test, if not, returning to the selector, retesting the wrong question, only The above exclusion logic can be applied to select other different near or similar sentences from the same unit for testing to avoid affecting the test result due to the memory of the subject.

In the test result of the present invention, the sound, the medium, the rhyme, the tone of the characters in the words of the correct words and the subject selected by the subject are displayed, and the wrong sound, media, rhyme, and tone are marked for testing. The analysis of each voice can enhance the training of listening.

The above-mentioned system and operation of the present invention can repeatedly test the subject to achieve the purpose of hearing rehabilitation. In this case, the method of using near-tones is used to realize the test of the voice by means of an electronic device, so that the testee can be easily tested, and the test can be repeated for the wrong result, thereby enhancing the hearing, which is a conventional technique. Those who cannot be reached in the middle. Furthermore, the selection of the near sounds in the present invention is mainly based on double words. According to the representativeness of the phrase, the frequency of use in daily life, and the degree to which the voices are close to each other determine the frequency of occurrence of the phrase, which is not the case of the conventional system, and is not a Innovation.

In summary, the humanized design of this case is quite in line with actual needs. The specific improvement of the existing defects is obviously a breakthrough improvement advantage compared with the prior art, and it has an improvement in efficacy and is not easy to achieve. The case has not been disclosed or disclosed in domestic and foreign literature and market, and has complied with the provisions of the Patent Law.

The detailed description of the preferred embodiments of the present invention is intended to be limited to the scope of the invention, and is not intended to limit the scope of the invention. The patent scope of this case.

10. . . Similar pronunciation dictionary

20. . . Similar sentence library

30. . . Mouth database

40. . . Expression database

50. . . Noise database

60. . . Equalizer

70. . . Selector

75. . . Test question group

80. . . player

90. . . receiver

100. . . Scorer

Figure 1 shows the component block diagram of the database and selector logic in this case.

Figure 2 shows a block diagram of the components of the operation in this case.

10. . . Similar pronunciation dictionary

20. . . Similar sentence library

30. . . Mouth database

40. . . Expression database

50. . . Noise database

60. . . Equalizer

70. . . Selector

75. . . Test question group

Claims (8)

An interactive voice test and training platform comprises: a database; wherein the database comprises: a close-to-sense vocabulary, storing words of different similar sounds and all similar sounds corresponding to the words, each constituting the word a unit in the library; a first equalizer is used to convert an input voice into electronic voice, a so-called electronic voice refers to a voice played from a general electronic instrument; a selector is based on the default selection logic A test question is selected in the database; a player plays a voice or image according to the result selected by the selector; the player plays the voice and the image; the voice plays the result selected by the selector in the form of voice, the image Playback is mainly to play the image; when playing in the voice, the word is played first by voice, and the words of the other similar sounds in the same unit of the word are displayed on the display screen, and the subject determines the displayed A word corresponds to the person expressed by the voice; a receiver is used to receive the correctness of the result replied by the subject; mainly input by the tester; The scorer determines the correctness of the result replied by the subject according to the input received by the receiver and compares the built-in answer, and accumulates the reply result after each reply; the scorer includes one The pass mechanism is used to determine whether the score obtained by the subject is passed. If yes, the test is ended. If not, return to the selector and retest the wrong question. Only an exclusion logic can be applied at this time. Choose different from the same unit (not appeared) The combination of close sounds is tested to avoid affecting the test results due to the memory of the subject. For example, the interactive speech test and training platform of claim 1 of the patent scope, wherein the selection of the near-tones is mainly based on double-words, and the frequency of use in daily life according to the representativeness of the words and units; The degree to which speech is close to each other determines how often the phrase appears. For example, the interactive voice test and training platform of claim 1 includes a library of similar sentences, storing various sentences, and storing a similar sound of the word in a sentence. Words. For example, the interactive voice test and training platform of claim 1 of the patent scope, wherein the database further includes a mouth-shaped database for storing mouth shapes corresponding to various pronunciations. For example, the interactive voice test and training platform of claim 1 or 4, wherein the database further includes an expression database for storing expressions corresponding to various pronunciations. For example, the interactive voice test and training platform of claim 1 of the patent scope, wherein the database still contains a noise database, and various noises are stored, and the volume of the noise can be adjusted as needed. For example, the interactive speech test and training platform of the third application patent scope, in which the test of the sentence is mainly based on the way the tester directly recites. For example, the interactive speech test and training platform of claim 1 of the patent scope, wherein the test results include the correct words and the subjects selected. The sound, medium, rhyme, and tone of a character in a word, and indicate the wrong sound, media, rhyme, and tone, so that the subject can analyze the correctness of each voice, so the training of hearing can be strengthened.