A method, device and storage medium for voice translation based on native voiceprint features

Abstract

The present invention discloses a method, device and storage medium for voice pickup translation based on native voiceprint features, and relates to the technical field of speech recognition. The present invention includes obtaining native audio; segmenting the native audio to obtain multiple original phonemes and corresponding sequences; obtaining several types of voiceprint features of each original phoneme; performing semantic recognition on the native audio to obtain the original language text; translating the original language text into translated semantic text; performing phoneme fitting on the translated semantic text to obtain multiple translated phonemes and corresponding sequences; and correcting the translated phonemes according to the original phonemes and the corresponding sequences and the corresponding several types of voiceprint features to obtain translated audio. The present invention realizes the correction of speech translation results and the polishing of translation results by identifying and extracting the speaker's native voiceprint features.

Description

Technical Field

[0001] The present invention belongs to the technical field of speech recognition, and in particular to a method, device and storage medium for voice pickup translation based on native voiceprint features. Background Art

[0002] As globalization deepens, cross-language communication has become a daily need. To meet this need, speech translation technology has developed rapidly. The traditional speech translation process generally includes four steps: speech signal acquisition, speech recognition (converting speech to text), text translation, and speech synthesis (converting the translated text back to speech). Although existing technologies have made significant progress in the accuracy of speech recognition and machine translation, there are still some limitations.

[0003] Existing speech translation systems mostly focus on the text content of speech and often ignore the rich non-verbal information contained in the sound. This not only causes the translation results to lack the speaker's voice emotion, but also reduces the accuracy of recognition translation. Summary of the invention

[0004] The purpose of the present invention is to provide a method, device and storage medium for voice translation based on native voiceprint features, which can correct the speech translation results and polish the translation results by identifying and extracting the native voiceprint features of the speaker.

[0005] In order to solve the above technical problems, the present invention is achieved through the following technical solutions:

[0006] The present invention provides a method for translating sound based on native voiceprint features, comprising:

[0007] Get native audio;

[0008] Segmenting the native audio to obtain a plurality of original phonemes and corresponding sequences;

[0009] Acquire several types of voiceprint features of each of the original phonemes, wherein the types of the voiceprint features include spectrum features, formant features and / or sound intensity features;

[0010] Performing semantic recognition on the native audio to obtain the original language text;

[0011] Translating the original language text into a translated semantic text;

[0012] Performing phoneme fitting on the translated semantic text to obtain a plurality of translated phonemes and corresponding sequences;

[0013] The translated phonemes are modified according to the original phonemes and the corresponding sequence as well as the corresponding types of voiceprint features to obtain translated audio.

[0014] The present invention also discloses a method for translating sound based on native voiceprint features, comprising:

[0015] Acquire and store audio streams in real time;

[0016] Performing noise reduction filtering on the audio stream to obtain a human voice stream;

[0017] Obtaining a blank period in the vocal stream;

[0018] intercepting the vocal stream between the blank periods as native audio;

[0019] The native audio is modified to obtain the translated audio.

[0020] The present invention also discloses a method for translating sound based on native voiceprint features, comprising:

[0021] receiving the translated audio;

[0022] The translated audio is played.

[0023] The present invention also discloses a device, characterized in that it comprises:

[0024] Microphone, used to record native audio;

[0025] A translation unit that translates the native audio into translated audio

[0026] A speaker is used to play the translated audio.

[0027] The present invention also discloses a storage medium, comprising:

[0028] The storage medium stores at least one command, at least one program, code set or instruction set, and the at least one command, at least one program, code set or instruction set is loaded and executed by the processor to implement a method for picking up and translating sound based on native voiceprint features.

[0029] The present invention records the speaker's original audio through a microphone, then uses a translation unit to identify and extract the speaker's original voiceprint features, and finally plays the translated audio through a speaker. In this process, the speaker's personal voice features can be added to the translated and synthesized voice, which not only corrects the voice translation results, but also polishes the translation results.

[0030] Of course, any product implementing the present invention does not necessarily need to achieve all of the advantages described above at the same time. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for describing the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0032] Figure 1 A schematic diagram of functional modules and information flow of a sound pickup translation device according to an embodiment of the present invention;

[0033] Figure 2 A schematic diagram of the steps of an embodiment of a method for voice translation based on native voiceprint features according to the present invention Figure 1 ;

[0034] Figure 3 A schematic diagram of the steps of an embodiment of a method for voice translation based on native voiceprint features according to the present invention Figure 2 ;

[0035] Figure 4 A schematic diagram of the steps of an embodiment of a method for voice translation based on native voiceprint features according to the present invention Figure 3 ;

[0036] Figure 5 This is a schematic diagram of the step flow of step S7 in one embodiment of the present invention;

[0037] Figure 6 FIG. 1 is a schematic diagram of the step flow of step S76 in an embodiment of the present invention. Figure 1 ;

[0038] Figure 7 A schematic diagram of a step flow chart of step S762 in an embodiment of the present invention;

[0039] Figure 8 FIG. 1 is a schematic diagram of the step flow of step S76 in an embodiment of the present invention. Figure 2 ;

[0040] Fig. 9 A schematic diagram of a step flow chart of step S77 of the present invention in an embodiment;

[0041] In the accompanying drawings, the components represented by the reference numerals are listed as follows:

[0042] 1- Microphone, 2- Translation unit, 3- Speaker. DETAILED DESCRIPTION

[0043] In order to make the objectives, technical solutions and advantages of the present application clearer, the implementation methods of the present application will be further described in detail below with reference to the accompanying drawings.

[0044] It should be noted that the terms "first", "second", etc. in this application are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present application. On the contrary, they are merely examples of devices and methods consistent with some aspects of the present application as detailed in the appended claims.

[0045] Speech translation is the process of converting oral speech input in one language into oral speech output in another language. It combines speech recognition and machine translation technology, allowing people to communicate and exchange across languages ​​through speech. However, the speech of simultaneous interpretation in the prior art uses electronic fitting speech, which cannot restore the tone of the real speaker. In view of this, the present invention provides the following solution.

[0046] See also Figures 1 to 4 As shown, the present invention provides a voice translation device, which functionally includes a microphone 1, a translation unit 2 and a speaker 3. During use, the microphone 1 is used to record native audio, the translation unit 2 is used to translate the native audio into translated audio, and the speaker 3 is used to play the translated audio. Of course, this is just a brief introduction to each functional module, and each functional module is described in detail below.

[0047] First, the microphone 1 can execute step S011 to obtain and store the audio stream in real time. Of course, it can also be stored in a dedicated storage module without recording. The translation unit 2 can then execute step S012 to perform noise reduction filtering on the audio stream to obtain a vocal stream. Next, step S013 can be executed to obtain the blank period in the vocal stream. Next, step S014 can be executed to intercept the vocal stream between the blank periods as native audio. Next, step S015 can be executed to correct the native audio to obtain the translated audio.

[0048] The speaker 3 can be used as an integral device with the microphone 1 and the translation unit 2, and the speaker 3 can also be set separately. For example, multiple speakers 3 can be connected to the translation unit 2 by wire or wirelessly. The speaker 3 can then perform step S021 to receive the translated audio in the method for picking up and translating based on native voiceprint features according to any one of claims 1 to 6, and finally perform step S022 to play the translated audio.

[0049] In the process of translating the native audio to obtain the translated audio, the translation unit 2 can first execute step S1 to obtain the native audio. Next, step S2 can be executed to segment the native audio to obtain multiple original phonemes and the corresponding order. Next, step S3 can be executed to obtain several types of voiceprint features for each original phoneme. Voiceprint features refer to the unique biometric features of an individual in a speech signal, similar to fingerprints or irises. Each person's voice has unique sound features, where the types of voiceprint features include spectral features, resonance peak features and / or sound intensity features.

[0050] Spectral features are characteristic representations of sound signals in the frequency domain, which are used to describe the energy distribution and spectral shape characteristics of sound signals at different frequencies. Spectral features are characteristic representations of sound signals in the frequency domain, which are used to describe the energy distribution and spectral shape characteristics of sound signals at different frequencies. The spectral features of voiceprints reflect the energy distribution and spectral shape characteristics of sound signals at different frequencies. Each person's voiceprint spectral features are unique because they are affected by the shape of the vocal tract, such as the throat, lips, and nasal cavity.

[0051] Resonance peak feature is an important acoustic feature in voiceprint analysis, which is used to describe the distribution of resonance peaks of sound signals in the frequency domain. Resonance peak is a prominent peak with high spectrum intensity in the sound signal, which reflects the resonance frequency of the sound signal when it passes through the resonance cavity such as vocal cords, pharynx and oral cavity in the vocal tract system.

[0052] The sound intensity feature is a feature that describes the intensity (or volume) of a sound signal. It reflects the energy or amplitude level of the sound signal. The sound intensity feature is often used in fields such as sound processing, audio analysis, and voiceprint recognition.

[0053] Next, step S4 can be executed to perform semantic recognition on the native audio to obtain the original language text. Next, step S5 can be executed to translate the original language text into the translated semantic text. Rule-based Machine Translation (RBMT) can be used in this process: this method relies on manually written translation rules and grammar rules. It analyzes the source language text into grammatical structures, and then generates the target language text according to predefined rules. However, this method requires a lot of manual work and expertise, and may not be flexible enough for complex language structures and expressions. Neural Machine Translation (NMT) can also be used: this method uses a deep neural network model for translation. It directly maps the source language text to the target language text by training an end-to-end neural network model. The NMT method performs better when processing long sentences and complex grammatical structures, and can better capture contextual information. It usually requires a large amount of training data and computing resources for training and inference.

[0054] Next, step S6 can be executed to perform phoneme fitting on the translated semantic text to obtain multiple translated phonemes and corresponding orders. Text-to-Speech (TTS) technology can be used. TTS technology converts text input into audible speech output. There are many online and offline TTS tools and services available, including open source libraries and commercial products. These tools usually provide a variety of speech synthesis models and speech styles, which can generate corresponding speech based on the input text.

[0055] Finally, step S7 is performed to correct the translated phonemes according to the original phonemes and the corresponding order and the corresponding voiceprint features to obtain the translated audio. This process needs to be implemented in combination with the speaker's voiceprint features, which is described in detail below.

[0056] See also Figure 5 As shown, in order to achieve the purpose of correcting the translated phonemes, the above-mentioned step S7 can firstly perform step S71 to semantically segment the original language text to obtain several original text segments and corresponding sequences during the specific implementation process. Next, step S72 can be performed to semantically segment the translated semantic text to obtain several translated text segments and corresponding sequences. Next, step S73 can be performed to obtain several original phonemes corresponding to each original text segment and the corresponding sequence. Next, step S74 can be performed to obtain several translated phonemes corresponding to each translated text segment and the corresponding sequence. Next, step S75 can be performed to semantically match the translated text segments according to the translation comparison relationship between the original language text and the translated semantic text to obtain several pairs of original text segments and translated text segments with the same semantics. Next, step S76 can be performed to correct several translated phonemes corresponding to the translated text segments according to several types of voiceprint features of several original phonemes corresponding to the original text segments to obtain translated text segments corresponding to the translated text segments. Finally, step S77 can be performed to merge the translated text segments according to the order of the translated text segments to obtain translated audio.

[0057] See also Fig. 9 As shown, in order to improve the fluency of the merged translated audio, the above-mentioned step S77 can first execute step S771 to obtain the tone features of the native audio during the specific implementation process. The tone feature refers to the pitch change pattern in the speech. It reflects the pitch difference between different syllables or phonemes in the speech signal. Tone is an important speech feature in language, which can convey information such as word meaning, tone and emotion. Next, step S772 can be executed to assign the tone features of the native audio to multiple translated segments merged in the order of the translated text segments to obtain the translated audio. The fluency of the translated audio is improved by adjusting the tone features, which is more in line with the speaker's actual speaking state and tone.

[0058] In order to supplement the implementation process of the above steps S71 to S77, the source code of some functional modules is provided, and the explanation is compared in the comment section. In order to avoid the leakage of data involving commercial secrets, some data that does not affect the implementation of the solution are desensitized, the same below.

[0059]

[0060]

[0061]

[0062]

[0063]

[0064] The above code is a high-level framework that describes the entire process from text to translated audio. The process is as follows:

[0065] First, the original text and the translated text are semantically segmented to obtain text segments and their order, and then the corresponding phoneme sequence is obtained for each text segment. Next, the original text segment and the translated text segment are matched to ensure semantic consistency. The translated phonemes are then corrected using the voiceprint features of the original phonemes, and the corrected translated phoneme sequences are sequentially merged to generate the translated audio.

[0066] The implementation of this process requires specific algorithms for multiple complex steps such as semantic segmentation, phoneme extraction, text matching, and sound feature adjustment, which are simplified into framework functions in the above code. In actual applications, the implementation of each function will involve complex algorithms and possible deep learning models.

[0067] See also Figure 6As shown, since the duration of a conversation of a speaker is relatively long, a conversation can be divided into multiple pairs of original text segments and translated text segments with correlation matching, and their corresponding voiceprint features are also correlated. In view of this, for each pair of semantically matched original text segments and translated text segments, the above-mentioned step S76 can firstly perform step S761 to vectorize the voiceprint features of each type of each original phoneme corresponding to the original text segment to obtain the voiceprint feature vector of each original phoneme corresponding to the original text segment. Next, step S762 can be performed to select a number of characteristic original phones according to the voiceprint feature vector of each original phoneme corresponding to the original text segment, and obtain the duration ratio coefficient of each characteristic original phoneme. Next, step S763 can be performed to obtain the order of the characteristic original phonemes corresponding to the original text segment according to the order of all original phones corresponding to the original text segment and the characteristic original phones corresponding to the original text segment. Next, step S764 can be performed to obtain the total duration of all translated phonemes corresponding to the translated text segment. Next, step S765 can be executed to arrange the duration ratio coefficient of each characteristic original phoneme in the order of the characteristic original phonemes and then divide the total duration of the transliterated phonemes corresponding to the translated text segment to obtain a number of transliterated phonemes corresponding to each characteristic original phoneme. Finally, step S766 can be executed to assign a number of types of voiceprint features of the characteristic original phonemes to the corresponding number of transliterated phonemes and then combine them to obtain a translated text segment corresponding to the translated text segment.

[0068] See also Figure 8 As shown, of course, not every translated text segment has the same semantic original text segment, so it is necessary to execute step S767 to determine whether there is a translated text segment that has no corresponding original text segment for semantic matching. If not, no processing is performed. If yes, then step S768 can be executed next. For the translated text segment that has no corresponding original text segment for semantic matching, the adjacent original text segment corresponding to the translated text segment with the corresponding original text segment for semantic matching is obtained as the semantically matched translated text segment and the corresponding original text segment. Then continue to execute steps S762 to S766.

[0069] In order to provide additional explanation for the implementation process of the above-mentioned steps S761 to S766, source code of some functional modules is provided, and comparative explanations are provided in the comment section.

[0070]

[0071]

[0072]

[0073]

[0074] The above code implements a process of applying the phoneme features of the original text segment to the phonemes of the translated text segment. First, the phoneme features of the original text segment are vectorized, then the characteristic phonemes are selected and their duration ratio coefficients are calculated, and then these ratio coefficients are applied to the phoneme duration distribution of the translated text segment, and the voiceprint features of the characteristic phonemes of the original text segment are assigned to the corresponding translated phonemes. This code is an example of a framework nature. Some functions such as the vectorization of voiceprint features and the assignment of phoneme features need to be filled in with implementation details based on the actual voiceprint feature extraction and processing algorithms in actual applications.

[0075] See also Figure 7 As shown, since the number of original phonemes corresponding to the original text segment is large, in order to increase the speed of translation without significantly reducing the translation sound effect, for each original text segment, the above step S762 can first perform step S7621 to select several marked voiceprint feature vectors from all voiceprint feature vectors during the specific implementation process. Next, step S7622 can be performed to calculate the vector difference between the marked voiceprint feature vector and the unmarked voiceprint feature vector. Next, step S7623 can be performed to classify the marked voiceprint feature vector with the smallest modulus of the vector difference between the unmarked voiceprint feature vector and the vector into the same marked group. Next, step S7624 can be performed to calculate the mean vector of all marked voiceprint feature vectors and unmarked voiceprint feature vectors in the marked group. Next, step S7625 can be performed to calculate the marked voiceprint feature vector or the unmarked voiceprint feature vector with the smallest modulus of the vector difference with the mean vector in the marked group as the updated marked voiceprint feature vector. Next, step S7626 can be performed to determine whether the marked voiceprint feature vector of the marked group has changed. If yes, then the steps S7622 to S7626 can be executed to return to continuously update the marked group and the marked voiceprint feature vector. If no, then the step S7627 can be executed to use the original phoneme corresponding to the marked voiceprint feature vector as the characteristic original phoneme. Finally, the step S7628 can be executed to use the ratio coefficient between the accumulated durations of the original phonemes corresponding to the marked voiceprint feature vector or the unmarked voiceprint feature vector in the marked group as the duration ratio coefficient of each corresponding characteristic original phoneme.

[0076] In order to provide additional explanation for the implementation process of the above-mentioned steps S7621 to S7628, the source code of some functional modules is provided, and a comparative explanation is provided in the comment section.

[0077]

[0078]

[0079]

[0080]

[0081]

[0082] This code uses an algorithm to group a group of phonemes, then selects the phoneme closest to the mean vector in each group as the characteristic original phoneme, and calculates the duration ratio coefficient of each phoneme in its group. In this way, a group of representative phonemes and corresponding duration ratios can be obtained, which is helpful for subsequent sound analysis and synthesis.

[0083] The present invention also discloses a storage medium, in which at least one command, at least one program, code set or instruction set is stored. The at least one command, at least one program, code set or instruction set is loaded and executed by a processor to implement the above-mentioned method of picking up and translating based on native voiceprint features.

[0084] The flow chart and block diagram in the accompanying drawings show the possible architecture, function and operation of the device, system, method and computer program product according to multiple embodiments of the present application. In this regard, each square frame in the flow chart or block diagram can represent a part of a module, program segment or instruction, and a part of the module, program segment or instruction includes one or more executable instructions for realizing the logical function of the specification. In some alternative implementations, the functions marked in the square frame can also occur in a sequence different from that marked in the accompanying drawings. For example, two continuous square frames can actually be executed substantially in parallel, and they can also be executed in reverse order sometimes, depending on the functions involved.

[0085] It should also be noted that each box in the block diagram and / or flowchart, and the combination of boxes in the block diagram and / or flowchart, can be implemented by hardware that performs the corresponding function or action, such as a circuit or ASIC (Application Specific Integrated Circuit), or can be implemented by a combination of hardware and software, such as firmware.

[0086] Although the present invention is described herein in conjunction with various embodiments, in the process of implementing the claimed invention, those skilled in the art may understand and implement other variations of the disclosed embodiments by viewing the drawings, the disclosure, and the appended claims. In the claims, the word "comprising" does not exclude other components or steps, and "one" or "an" does not exclude multiple situations. A single processor or other unit may implement several functions listed in the claims. Certain measures are recorded in different dependent claims, but this does not mean that these measures cannot be combined to produce good results.

[0087] The embodiments of the present application have been described above, and the above description is exemplary, not exhaustive, and is not limited to the disclosed embodiments. Many modifications and changes are obvious to those of ordinary skill in the art without departing from the scope of the described embodiments. The selection of terms used herein is intended to best explain the principles of the embodiments, practical applications, or improvements to the technology in the market, or to enable other persons of ordinary skill in the art to understand the embodiments disclosed herein.

Claims

1. A method for voice translation based on native voiceprint features, characterized in that: include, Get native audio; Segmenting the native audio to obtain a plurality of original phonemes and corresponding sequences; Acquire several types of voiceprint features of each of the original phonemes, wherein the types of the voiceprint features include spectrum features, formant features and / or sound intensity features; Performing semantic recognition on the native audio to obtain the original language text; Translating the original language text into a translated semantic text; Performing phoneme fitting on the translated semantic text to obtain a plurality of translated phonemes and corresponding sequences; Modifying the translated phonemes according to the original phonemes and the corresponding sequence and the corresponding voiceprint features of several types to obtain translated audio; in, The step of correcting the translated phoneme to obtain the translated audio according to the original phoneme and the corresponding sequence and the corresponding several types of voiceprint features includes: Performing semantic segmentation on the original language text to obtain a plurality of original text segments and corresponding sequences; Performing semantic segmentation on the translated semantic text to obtain a plurality of translated text segments and corresponding sequences; Obtaining a number of original phonemes corresponding to each original text segment and the corresponding order; Obtaining a plurality of the translated phonemes corresponding to each of the translated text segments and the corresponding order; According to the translation comparison relationship between the original language text and the translated semantic text, semantic matching is performed on the translated text segment to obtain a plurality of pairs of the original text segment and the translated text segment with the same semantics; According to the voiceprint features of the original phonemes corresponding to the original text segment, the translated phonemes corresponding to the translated text segment are modified to obtain a translated text segment corresponding to the translated text segment; Merging the translated text segments according to the order of the translated text segments to obtain translated audio; in, The step of correcting the plurality of translated phonemes corresponding to the translated text segment according to the plurality of types of voiceprint features of the plurality of original phonemes corresponding to the original text segment to obtain a translated language segment corresponding to the translated text segment includes: For each pair of semantically matched original text segments and translated text segments, Vectorizing the voiceprint feature of each type of each original phoneme corresponding to the original text segment to obtain a voiceprint feature vector of each original phoneme corresponding to the original text segment, According to the voiceprint feature vector of each original phoneme corresponding to the original text segment, several characteristic original phonemes are selected, and the duration ratio coefficient of each characteristic original phoneme is obtained. According to the order of all the original phonemes corresponding to the original text segment and the characteristic original phonemes corresponding to the original text segment, the order of the characteristic original phonemes corresponding to the original text segment is obtained, Obtaining the total duration of all the translated phonemes corresponding to the translated text segment, After arranging the duration ratio coefficient of each characteristic original phoneme in the order of the characteristic original phonemes, the total duration of the translated phonemes corresponding to the translated text segment is divided to obtain a plurality of the translated phonemes corresponding to each characteristic original phoneme. The voiceprint features of several types of characteristic original phonemes are assigned to several corresponding translated phonemes and then combined to obtain the translated text segment corresponding to the translated text segment.

2. The method according to claim 1, characterized in that The step of selecting a plurality of characteristic original phonemes according to the voiceprint feature vector of each original phoneme corresponding to the original text segment and obtaining the duration ratio coefficient of each characteristic original phoneme includes: For each original text segment, Selecting a number of marked voiceprint feature vectors from all voiceprint feature vectors; Calculate and obtain the vector difference between the marked voiceprint feature vector and the unmarked voiceprint feature vector; Classify the unmarked voiceprint feature vector and the marked voiceprint feature vector with the smallest modulus of vector difference into the same marked group; Calculate the mean vector of all the marked voiceprint feature vectors and the unmarked voiceprint feature vectors in the marked group; Calculate and obtain the marked voiceprint feature vector or the unmarked voiceprint feature vector with the smallest modulus of the vector difference between the marked voiceprint feature vector and the mean vector in the marked group as the updated marked voiceprint feature vector; Determine whether the marked voiceprint feature vector of the marked group has changed; If so, return to continuously update the marked group and the marked voiceprint feature vector; If not, the original phoneme corresponding to the marked voiceprint feature vector is used as the feature original phoneme; The proportional coefficient between the accumulated durations of the original phonemes corresponding to the marked voiceprint feature vectors or the unmarked voiceprint feature vectors in the marked group is used as the duration proportional coefficient of each corresponding characteristic original phoneme.

3. The method according to claim 1, characterized in that The step of obtaining a plurality of original phonemes by segmenting the native audio and correcting the translated phonemes in a corresponding order to obtain the translated audio further includes: Determining whether there is a translated text segment that has no corresponding original text segment for semantic matching; If not, no processing is performed; If so, for the translated text segment that has no corresponding original text segment for semantic matching, obtain the original text segment corresponding to the adjacent translated text segment that has a corresponding original text segment for semantic matching as the semantically matched translated text segment and the corresponding original text segment.

4. The method according to claim 1, characterized in that The step of merging the translated text segments according to the order of the translated text segments to obtain the translated audio, include, Obtaining the tone features of the native audio; The translated audio is obtained by assigning the tone features of the native audio to the plurality of translated language segments merged in the order of the translated text segments.

5. A method for voice translation based on native voiceprint features, characterized in that: include, Acquire and store audio streams in real time; Performing noise reduction filtering on the audio stream to obtain a human voice stream; Obtaining a blank period in the vocal stream; intercepting the vocal stream between the blank periods as native audio; The native audio is corrected according to the method for picking up and translating based on native voiceprint features as described in any one of claims 1 to 4 to obtain translated audio.

6. A method for voice translation based on native voiceprint features, characterized in that: include, Receiving the translated audio in the method for voice pickup and translation based on native voiceprint features as described in any one of claims 1 to 4; The translated audio is played.

7. A voice translation device, characterized in that: include, Microphone, used to record native audio; A translation unit, configured to translate the native audio into translated audio according to the method for translation based on native voiceprint features according to any one of claims 1 to 4; A speaker is used to play the translated audio.

8. A storage medium, characterized in that: include, The storage medium stores at least one command, at least one program, a code set or an instruction set, and the at least one command, the at least one program, the code set or the instruction set is loaded and executed by the processor to implement a method for picking up and translating based on native voiceprint features as described in any one of claims 1 to 4.

Images