A bilingual voice real-time translation method, system, terminal and medium

Abstract

The invention discloses a method, system, terminal and medium for real-time bilingual speech translation, and relates to the technical field of machine translation. The technical scheme thereof is as follows: in the process of applying machine translation technology to perform bilingual translation, the original sound features of each word in the original speech information of the first language are retained, and the original sound features are preliminarily corrected according to the sound difference vector representing the difference in language expression styles between the first language and the second language under the same tone feature, and the template speech and the corrected original sound information are fused to obtain the final translated speech information, which not only truly restores the fluctuation of emotional expression in a dialogue scene, but also can perform adaptive adjustments to extreme situations such as too fast speech speed, too slow speech speed, too high voice, too low voice, etc. in the dialogue scene, thereby ensuring the stability and reliability of the real-time bilingual speech translation.

Description

Technical Field

[0001] The present invention relates to the technical field of machine translation, and more specifically, to a bilingual speech real-time translation method, system, terminal and medium. Background Art

[0002] Bilingual translation is the translation from one language to another, and is generally performed using efficient machine translation technology, such as: rule-based translation methods, which use linguistic rules and vocabulary databases to translate texts; statistical-based translation methods, which learn translation rules by analyzing large amounts of bilingual text data; neural machine translation methods, which use deep learning models, especially sequence-to-sequence (Seq2Seq) models such as Transformer, to perform translation.

[0003] The machine translation technology in the prior art mainly translates text information or voice information in one language into text information or language information in another language after preprocessing. The translation process uses mutual translation between texts as an intermediate transition. A large amount of information involving emotional expression will be lost in this process, so it is generally used for document translation or one-way dialogue translation; even if the translation involves voice, the text information is converted using template voice, which cannot restore the emotional expression of the real dialogue, so it is difficult to apply to scenarios with interactive attributes, especially in scenarios where dialogue is conducted in two tonal languages, such as Chinese, Korean, Thai and other languages.

[0004] Therefore, how to research and design a bilingual speech real-time translation method, system, terminal and medium that can overcome the above-mentioned defects is a problem that we urgently need to solve. Summary of the invention

[0005] In order to address the deficiencies in the prior art, the purpose of the present invention is to provide a bilingual speech real-time translation method, system, terminal and medium, which not only truly restore the emotional expression fluctuations in the dialogue scene, but also can adaptively adjust the extreme situations such as too fast speaking speed, too slow speaking speed, too high voice, too low voice in the dialogue scene, thereby ensuring the stability and reliability of bilingual speech real-time translation.

[0006] The above technical objectives of the present invention are achieved through the following technical solutions:

[0007] In a first aspect, a bilingual speech real-time translation method is provided, comprising the following steps:

[0008] Receive original speech information in the first language;

[0009] The original speech information of the first language is converted into original text, and the original text is segmented and word-segmented to obtain the word vector of a single sentence;

[0010] Identify and extract the tone features of individual sentences from the original speech information of the first language;

[0011] Extract the original acoustic features of each word from the original speech information of the first language to obtain an acoustic feature sequence corresponding to the word vector;

[0012] According to the tone feature matching, the sound difference vectors of the first language and the second language in different parts of speech are obtained;

[0013] Correcting the corresponding original sound features according to the sound difference vector to obtain reference sound features of each word;

[0014] The word vector is translated into a sentence in the second language using machine translation technology, and the initial speech information of the second language is generated by translating the corresponding sentence into the second language in combination with the template speech of the second language;

[0015] The reference acoustic features are fused with the corresponding initial speech information to obtain the translated speech information of the second language.

[0016] Furthermore, the tone features are identified and classified based on emotional vocabulary, sentence structure, adverbs, modifiers and intonation;

[0017] Among them, tone characteristics include friendly tone, polite tone, commanding tone, interrogative tone, persuasive tone, sarcastic tone, humorous tone, serious tone, sad tone, angry tone, surprised tone, encouraging tone, and warning tone.

[0018] Furthermore, the sub-features of the original sound features include pitch, volume, speaking speed, stress, pause and sound quality.

[0019] Furthermore, the expression of the acoustic difference vector is:

[0020] ;

[0021] in, Represents the sound difference vector of the first language A and the second language B in different parts of speech; Represents the subvector of the acoustic difference between the first language A and the second language B at the nth part of speech; The feature difference value corresponding to the sub-feature m in the sound difference sub-vector representing the nth part of speech; Indicates the feature value corresponding to the nth part of speech in sub-feature m in the second language B; Indicates the feature value corresponding to the nth part of speech in the first language A in the sub-feature m.

[0022] Furthermore, the calculation formula of the reference sound feature is specifically:

[0023] ;

[0024] in, Indicates the corrected feature value corresponding to the nth part of speech in the reference sound feature in the sub-feature m; Indicates the actual feature value corresponding to the nth part of speech in the original sound feature in the sub-feature m; Represents the feature difference value corresponding to sub-feature m in the sound difference sub-vector of the nth part of speech.

[0025] Furthermore, the expression for fusing the reference acoustic feature with the corresponding initial voice information is specifically:

[0026] ;

[0027] in, Represents the average feature value corresponding to the n parts of speech in the sub-feature m in the reference sound feature; Indicates the number of parts of speech in the reference sound feature; Indicates the corrected feature value corresponding to the i-th part of speech in the reference sound feature in the sub-feature m; Indicates the final feature value corresponding to the nth part of speech in the translated speech information in sub-feature m; Indicates the corrected feature value corresponding to the nth part of speech in the reference sound feature in the sub-feature m; Indicates the template feature value corresponding to the nth part of speech in the initial speech information in the sub-feature m.

[0028] Furthermore, the method also includes:

[0029] Acquiring original voice information of a second language according to the original voice information of the first language, wherein the original voice information of the second language is a conversational voice of the original voice information of the first language, and a timestamp of the original voice information of the second language is earlier than a timestamp of the original voice information of the first language;

[0030] Identify and extract associated tones associated with tone features from original speech information of a second language;

[0031] When matching the sound difference vector, the matching is performed in combination with the tone feature and at least one associated tone.

[0032] In a second aspect, a bilingual speech real-time translation system is provided, the system being used to implement a bilingual speech real-time translation method as described in any one of the first aspects, comprising:

[0033] An information receiving module, used for receiving original speech information in a first language;

[0034] The text processing module is used to convert the original speech information of the first language into original text, and perform sentence segmentation and word segmentation on the original text to obtain the word vector of a single sentence;

[0035] A tone recognition module, used to recognize and extract tone features of a single sentence from the original speech information of the first language;

[0036] A feature extraction module is used to extract the original acoustic features of each word from the original speech information of the first language to obtain an acoustic feature sequence corresponding to the word vector;

[0037] A feature matching module, used to obtain the sound difference vectors of the first language and the second language in different parts of speech according to the tone feature matching;

[0038] A feature correction module is used to correct the corresponding original sound features according to the sound difference vector to obtain the reference sound features of each word;

[0039] An initial translation module, which is used to translate the word vector into a sentence in the second language using machine translation technology, and generate initial speech information for translating the corresponding sentence into the second language in combination with the template speech in the second language;

[0040] The feature fusion module is used to fuse the reference acoustic features with the corresponding initial voice information to obtain the translated voice information of the second language.

[0041] In a third aspect, a computer terminal is provided, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, a method for real-time bilingual speech translation as described in any one of the first aspects is implemented.

[0042] In a fourth aspect, a computer-readable medium is provided, on which a computer program is stored, and the computer program is executed by a processor to implement a bilingual speech real-time translation method as described in any one of the first aspects.

[0043] Compared with the prior art, the present invention has the following beneficial effects:

[0044] 1. A method for real-time bilingual speech translation provided by the present invention retains the original sound features of each word in the original speech information of the first language in the process of bilingual translation using machine translation technology, and performs preliminary correction on the original sound features according to the sound difference vector representing the difference in language expression styles between the first language and the second language under the same tone features, and fuses the template speech and the corrected original sound information to obtain the final translated speech information, which not only truly restores the fluctuation of emotional expression in the dialogue scene, but also can adaptively adjust the extreme situations such as too fast speech speed, too slow speech speed, too high voice, too low voice, etc. in the dialogue scene, thereby ensuring the stability and reliability of real-time bilingual speech translation;

[0045] 2. The present invention refines the specific features in the sound difference vector to different parts of speech, and adaptively combines the original sound features according to the different translation objects, which can be applied to the translation of different languages ​​with different grammatical structures;

[0046] 3. When the reference acoustic feature is fused with the corresponding initial voice information, the present invention only fuses the fluctuation of the corrected feature value relative to the average feature value of all corrected feature values ​​with the acoustic feature in the initial voice information, thereby ensuring the smooth connection between words after the acoustic feature combination and the stability of the translated voice;

[0047] 4. When matching the sound difference vector, the present invention combines the tone feature and at least one associated tone for matching, so that the translated speech information can better restore the context of the corresponding dialogue, and at the same time can reduce the difficulty of selecting polysemous words. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] The drawings described herein are used to provide a further understanding of the embodiments of the present invention, constitute a part of this application, and do not constitute a limitation of the embodiments of the present invention. In the drawings:

[0049] Figure 1 is a flow chart of Embodiment 1 of the present invention;

[0050] Figure 2 It is a system block diagram in Example 2 of the present invention. DETAILED DESCRIPTION

[0051] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with embodiments and drawings. The exemplary embodiments of the present invention and their description are only used to explain the present invention and are not intended to limit the present invention.

[0052] Embodiment 1: A bilingual speech real-time translation method, as shown in FIG1, comprises the following steps:

[0053] S1: receiving original speech information in the first language;

[0054] S2: Convert the original speech information of the first language into original text, and perform sentence segmentation and word segmentation on the original text to obtain the word vector of a single sentence;

[0055] S3: Identify and extract the tone features of individual sentences from the original speech information of the first language;

[0056] S4: extracting the original acoustic features of each word from the original speech information of the first language to obtain an acoustic feature sequence corresponding to the word vector;

[0057] S5: Obtain the sound difference vectors of the first language and the second language in different parts of speech according to the matching of tone features;

[0058] S6: Correcting the corresponding original acoustic features according to the acoustic difference vector to obtain reference acoustic features of each word;

[0059] S7: Using machine translation technology to translate the word vector into a sentence in the second language, and combining the template speech in the second language to generate the initial speech information of the corresponding sentence translated into the second language;

[0060] S8: Fusing the reference acoustic feature with the corresponding initial voice information to obtain the translated voice information of the second language.

[0061] In step S1, the present invention can be applied to a dialogue scenario with two terminals, the first terminal collects original voice information in a first language, and the second terminal collects original voice information in a second language. When the first terminal collects a segment of voice, it can be transmitted to the second terminal, and the second terminal translates the original voice information in the first language into translated voice information in the second language; similarly, when the second terminal collects a segment of voice, it can be transmitted to the first terminal, and the first terminal translates the original voice information in the second language into translated voice information in the first language, thereby realizing real-time bilingual voice translation.

[0062] In addition, the present invention can also integrate the functions of the first terminal and the second terminal into one device.

[0063] In step S2, after receiving the original speech information in the first language, the original speech information needs to be preprocessed, including but not limited to filtering, framing, pre-emphasis, windowing and fast Fourier transform, etc., to extract the characteristics of the sound signal.

[0064] Speech-to-text conversion can be achieved through deep learning models, such as Transformer-based models, CNN models, Conformer or Paraformer, which can directly predict the corresponding text from the input audio or the features corresponding to the audio, without any restrictions.

[0065] Sentence segmentation of the original text can be performed using rule-based, machine learning-based, and other methods. For example, the rule-based method relies on specific punctuation marks and separators to identify sentence boundaries, such as identifying common sentence-ending punctuation marks, including but not limited to periods, question marks, exclamation marks, and the like.

[0066] The segmented sentences can be processed by existing segmentation tools. For example, Chinese can be processed by jieba, SnowNLP, THULAC, NLPIR and other segmentation tools. For Korean, KoNLPY, Mecab-ko, Khaiii and other segmentation tools can be used.

[0067] The multiple words after a sentence is split into two parts form a word vector.

[0068] In step S3, the segmented sentences can be trained to identify the tone features by using machine learning models such as support vector machines and naive Bayes, and the tone features are identified and classified based on emotional vocabulary, sentence structure, adverbs, modifiers and intonation. The tone features include but are not limited to friendly tone, polite tone, command tone, interrogative tone, persuasive tone, sarcastic tone, humorous tone, serious tone, sad tone, angry tone, surprised tone, encouraging tone and warning tone.

[0069] In step S4, the sub-features in the original sound features include but are not limited to pitch, volume, speaking speed, stress, pause and sound quality. Since there are multiple words in the word vector, the sound feature sequence can also be represented in matrix form.

[0070] In step S5, the classification of parts of speech includes but is not limited to nouns, verbs, adjectives, adverbs, prepositions, conjunctions, articles, pronouns, numerals, interjections, auxiliary words, modal words and participles.

[0071] The sound difference vector can be matched only by the parts of speech contained in the word vector to reduce the amount of calculation, so the sound difference vector corresponds one-to-one to the sound feature sequence and can be represented in a matrix form.

[0072] In addition, the sound difference vector can also be expressed by a sequence combination, such as the expression of the sound difference vector is:

[0073] ;

[0074] in, Represents the sound difference vector of the first language A and the second language B in different parts of speech; Represents the subvector of the acoustic difference between the first language A and the second language B at the nth part of speech; The feature difference value corresponding to the sub-feature m in the sound difference sub-vector representing the nth part of speech; Indicates the feature value corresponding to the nth part of speech in sub-feature m in the second language B; Indicates the feature value corresponding to the nth part of speech in the first language A in the sub-feature m.

[0075] The characteristic values ​​in the sound difference vector can be classified by a classifier, and the classification basis is the tone feature.

[0076] In addition, the classification basis can also be: obtaining original voice information of a second language based on original voice information of a first language, the original voice information of the second language is the conversational voice of the original voice information of the first language, and the timestamp of the original voice information of the second language is earlier than the timestamp of the original voice information of the first language; identifying and extracting associated tones associated with the tone features from the original voice information of the second language; and matching in combination with the tone features and at least one associated tone when matching the sound difference vector.

[0077] The present invention combines the tone feature and at least one associated tone for matching when matching the sound difference vector, so that the translated speech information can better restore the corresponding dialogue context and reduce the difficulty of selecting polysemous words.

[0078] In step S6, the calculation formula of the reference sound feature is specifically:

[0079] ;

[0080] in, Indicates the corrected feature value corresponding to the nth part of speech in the reference sound feature in the sub-feature m; Indicates the actual feature value corresponding to the nth part of speech in the original sound feature in the sub-feature m; Represents the feature difference value corresponding to sub-feature m in the sound difference sub-vector of the nth part of speech.

[0081] In step S7, the machine translation technology may apply existing translation technology, which will not be described in detail here.

[0082] In step S8, the expression for fusing the reference acoustic feature with the corresponding initial voice information is specifically:

[0083] ;

[0084] in, Represents the average feature value corresponding to the n parts of speech in the sub-feature m in the reference sound feature; Indicates the number of parts of speech in the reference sound feature; Indicates the corrected feature value corresponding to the i-th part of speech in the reference sound feature in the sub-feature m; Indicates the final feature value corresponding to the nth part of speech in the translated speech information in sub-feature m; Indicates the corrected feature value corresponding to the nth part of speech in the reference sound feature in the sub-feature m; Indicates the template feature value corresponding to the nth part of speech in the initial speech information in the sub-feature m.

[0085] When the present invention fuses the reference acoustic features with the corresponding initial voice information, only the fluctuation of the corrected feature value relative to the average feature value of all corrected feature values ​​is fused with the acoustic features in the initial voice information, thereby ensuring the smooth connection between words after the acoustic feature combination and the stability of the translated voice.

[0086] Embodiment 2: A bilingual speech real-time translation system, which is used to implement a bilingual speech real-time translation method as described in Embodiment 1, as shown in FIG2, including an information receiving module, a text processing module, a tone recognition module, a feature extraction module, a feature matching module, a feature correction module, an initial translation module and a feature fusion module.

[0087] Among them, the information receiving module is used to receive the original voice information of the first language; the text processing module is used to convert the original voice information of the first language into the original text, and perform sentence segmentation and word segmentation on the original text to obtain the word vector of a single sentence; the tone recognition module is used to identify and extract the tone features of a single sentence from the original voice information of the first language; the feature extraction module is used to extract the original sound features of each word from the original voice information of the first language to obtain the sound feature sequence corresponding to the word vector; the feature matching module is used to obtain the sound difference vector of the first language and the second language in different parts of speech according to the matching of the tone features; the feature correction module is used to correct the corresponding original sound features according to the sound difference vector to obtain the reference sound features of each word; the initial translation module is used to translate the word vector into a sentence in the second language by using the machine translation technology, and generate the initial voice information of the corresponding sentence translated into the second language in combination with the template voice of the second language; the feature fusion module is used to fuse the reference sound features with the corresponding initial voice information to obtain the translated voice information in the second language.

[0088] The present invention also describes a computer terminal, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, a bilingual speech real-time translation method as described in Example 1 is implemented.

[0089] The present invention also describes a computer-readable medium on which a computer program is stored. The computer program is executed by a processor to implement a bilingual speech real-time translation method as described in Example 1.

[0090] Working principle: In the process of bilingual translation using machine translation technology, the present invention retains the original sound features of each word in the original voice information of the first language, and performs preliminary correction on the original sound features based on the sound difference vector that characterizes the difference in language expression style between the first language and the second language under the same tone features, and fuses the template voice and the corrected original sound information to obtain the final translated voice information. This not only truly restores the fluctuations in emotional expression in the dialogue scene, but also can adaptively adjust extreme situations such as too fast speaking speed, too slow speaking speed, too high voice, too low voice, etc. in the dialogue scene, thereby ensuring the stability and reliability of real-time bilingual voice translation.

[0091] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present application may adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.

[0092] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0093] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0094] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1A process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0095] The specific implementation methods described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation method of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A bilingual speech real-time translation method, characterized in that: The following steps are involved: Receive original speech information in the first language; The original speech information of the first language is converted into original text, and the original text is segmented and word-segmented to obtain the word vector of a single sentence; Identify and extract the tone features of individual sentences from the original speech information of the first language; Extract the original acoustic features of each word from the original speech information of the first language to obtain an acoustic feature sequence corresponding to the word vector; According to the tone feature matching, the sound difference vectors of the first language and the second language in different parts of speech are obtained; Correcting the corresponding original sound features according to the sound difference vector to obtain reference sound features of each word; The word vector is translated into a sentence in the second language using machine translation technology, and the initial speech information of the second language is generated by translating the corresponding sentence into the second language in combination with the template speech of the second language; The reference acoustic features are fused with the corresponding initial speech information to obtain the translated speech information of the second language.

2. A bilingual speech real-time translation method according to claim 1, characterized in that: The tone features are identified and classified based on emotional vocabulary, sentence structure, adverbs, modifiers and intonation; Among them, tone characteristics include friendly tone, polite tone, commanding tone, interrogative tone, persuasive tone, sarcastic tone, humorous tone, serious tone, sad tone, angry tone, surprised tone, encouraging tone, and warning tone.

3. A bilingual speech real-time translation method according to claim 1, characterized in that: The sub-features of the original sound features include pitch, volume, speaking speed, stress, pause and sound quality.

4. A bilingual speech real-time translation method according to claim 1, characterized in that: The expression of the acoustic difference vector is: S A,B ={a1,a2,…,a n } a n ={b n,1 ,b n,2 ,…,b n,m }; Among them, S A,B Represents the sound difference vector of the first language A and the second language B in different parts of speech; a n represents the subvector of the difference in the first language A and the second language B at the nth part of speech; b n,m The feature difference value corresponding to the sub-feature m in the sound difference sub-vector representing the nth part of speech; Indicates the feature value corresponding to the nth part of speech in sub-feature m in the second language B; Indicates the feature value corresponding to the nth part of speech in the first language A in the sub-feature m.

5. A bilingual speech real-time translation method according to claim 1, characterized in that: The calculation formula of the reference sound feature is specifically: c n,m =d n,m +b n,m ; Among them, c n,m represents the corrected feature value corresponding to the nth part of speech in the reference sound feature in sub-feature m; d n,m Indicates the actual feature value corresponding to the nth part of speech in the original sound feature in the sub-feature m; b n,m Represents the feature difference value corresponding to sub-feature m in the sound difference sub-vector of the nth part of speech.

6. A bilingual speech real-time translation method according to claim 1, characterized in that: The expression for fusing the reference acoustic feature with the corresponding initial voice information is specifically: in, represents the average feature value of n parts of speech in the reference sound feature corresponding to sub-feature m; n represents the number of parts of speech in the reference sound feature; c i,m F represents the corrected feature value corresponding to the i-th part of speech in the reference sound feature in sub-feature m; n,m represents the final feature value corresponding to the nth part of speech in the translated speech information in sub-feature m; c n,m represents the corrected feature value corresponding to the nth part of speech in the reference sound feature in sub-feature m; f n,m Indicates the template feature value corresponding to the nth part of speech in the initial speech information in the sub-feature m.

7. A bilingual speech real-time translation method according to claim 1, characterized in that: The method further includes: Acquiring original voice information of a second language according to the original voice information of the first language, wherein the original voice information of the second language is a conversational voice of the original voice information of the first language, and a timestamp of the original voice information of the second language is earlier than a timestamp of the original voice information of the first language; Identify and extract associated tones associated with tone features from original speech information of a second language; When matching the sound difference vector, the matching is performed in combination with the tone feature and at least one associated tone.

8. A bilingual speech real-time translation system, characterized in that: The system is used to implement a bilingual speech real-time translation method as described in any one of claims 1 to 7, comprising: An information receiving module, used for receiving original speech information in a first language; The text processing module is used to convert the original speech information of the first language into original text, and perform sentence segmentation and word segmentation on the original text to obtain the word vector of a single sentence; A tone recognition module, used to recognize and extract tone features of a single sentence from the original speech information of the first language; A feature extraction module is used to extract the original acoustic features of each word from the original speech information of the first language to obtain an acoustic feature sequence corresponding to the word vector; A feature matching module, used to obtain the sound difference vectors of the first language and the second language in different parts of speech according to the tone feature matching; A feature correction module is used to correct the corresponding original sound features according to the sound difference vector to obtain the reference sound features of each word; An initial translation module, which is used to translate the word vector into a sentence in the second language using machine translation technology, and generate initial speech information for translating the corresponding sentence into the second language in combination with the template speech in the second language; The feature fusion module is used to fuse the reference acoustic features with the corresponding initial voice information to obtain the translated voice information of the second language.

9. A computer terminal comprising a memory, a processor and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, a bilingual speech real-time translation method as described in any one of claims 1 to 7 is implemented.

10. A computer readable medium having a computer program stored thereon, characterized in that: The computer program is executed by a processor to implement a bilingual speech real-time translation method as described in any one of claims 1 to 7.

Images