Automatic speech recognition method and device based on echo cancellation

Abstract

The invention provides an automatic voice recognition method and device based on echo cancellation, and the method comprises the steps: extracting an audio part of to-be-recognized voice information, and carrying out the voice recognition through a preset voice recognition technology, and obtaining a voice recognition result; vectorizing the voice recognition result by using a preset vectorization algorithm to obtain a target vector; performing semantic similarity comparison on a pre-constructed vector database and the target vector, and performing echo elimination according to a comparison result to obtain an echo elimination result; performing automatic speech recognition according to the echo elimination result; wherein the vector database comprises massive voice interaction vectors; the voice interaction vector is obtained by performing voice recognition by using a preset voice recognition technology according to voice interaction information output by the intelligent robot or the digital human and then continuing to perform vectorization by using the preset vectorization algorithm. The semantic information in the voice signal is intelligently analyzed and processed through the voice recognition technology, the semantic information is converted into the vector to be compared with the information output by the intelligent robot or the digital human subjected to similar processing, the echo is recognized and eliminated according to the comparison result, echo elimination with high efficiency and good adaptability is achieved, and the user experience is improved. Therefore, the accuracy of speech recognition can be improved on the premise of not sacrificing speech signal quality and semantic integrity.

Description

Technical Field

[0001] The present invention relates to the field of speech recognition technology, and in particular to an automatic speech recognition method and device based on echo cancellation. Background Art

[0002] Automatic speech recognition (ASR) technology is one of the core technologies in the field of modern human-computer interaction, which allows computers and smart devices to receive instructions and provide responses by recognizing and understanding human speech. With the rapid development of artificial intelligence generated content (AIGC) technology, the application scope of automatic natural language processing algorithms continues to expand, especially in the interaction between intelligent robots and digital humans, ASR technology has become the key to achieving smooth conversations.

[0003] However, in actual applications, when intelligent robots and digital humans interact with humans through voice, their voice output often interferes with human voice input, resulting in echo. This echo not only affects the clarity of the voice, but also interferes with the accurate recognition of the ASR system, resulting in increased errors in the recognition results. Traditional solutions include using filtering technology and training large models to process audio with echo.

[0004] Digital filtering technology processes audio signals through specific algorithms, trying to remove the echo part while retaining the original speech content. However, although digital filtering can reduce echo to a certain extent, it is difficult to completely eliminate echo without losing the quality of the original speech. Especially in complex sound environments, the filter may have difficulty distinguishing between echo and original speech, resulting in loss of speech information. The echo path function generated based on this may often be inaccurate. On this basis, some studies use deep learning models to deal with echo problems. Deep learning models train models that can identify and eliminate echoes by learning a large number of speech samples with and without echoes. Although deep learning-based models can effectively deal with echo problems, training these models requires a lot of data and computing resources. In practical applications, especially when interacting with intelligent robots and digital humans in real time, it is crucial to process echoes quickly and accurately, and existing solutions often fail to meet the needs of efficient real-time processing. At the same time, in order to maintain the accuracy of the model, it is necessary to update and retrain the model regularly, which further increases costs and maintenance work. At the same time, traditional solutions may need to readjust filter settings or retrain models when facing different voice environments and interaction scenarios, which will result in poor adaptability of the model to changing interaction environments.

[0005] In summary, the existing technology has the problems of low echo processing efficiency and poor adaptability. Summary of the invention

[0006] The present invention provides an automatic speech recognition method and device based on echo cancellation, which are used to solve the defects of low echo processing efficiency and poor adaptability in the prior art, and realize echo cancellation with high efficiency and good adaptability, thereby improving the accuracy and efficiency of automatic speech recognition.

[0007] The present invention provides an automatic speech recognition method based on echo cancellation, comprising the following steps: Extracting the audio part of the voice information to be recognized, and performing voice recognition using a preset voice recognition technology to obtain a voice recognition result; Vectorizing the speech recognition result using a preset vectorization algorithm to obtain a target vector; Using a pre-built vector database to compare semantic similarity with the target vector, performing echo cancellation according to the comparison result, and obtaining an echo cancellation result; Performing automatic speech recognition according to the echo cancellation result; Among them, the vector database includes a large number of voice interaction vectors; the voice interaction vectors are obtained by performing voice recognition using a preset voice recognition technology and then continuing to perform vectorization using the preset vectorization algorithm based on the voice interaction information output by the intelligent robot or digital human.

[0008] According to an automatic speech recognition method based on echo cancellation provided by the present invention, the speech recognition result is vectorized by using a preset vectorization algorithm to obtain a target vector, which specifically includes: Vectorizing the sentences in the speech recognition result respectively using a preset vectorization algorithm to obtain sentence sub-vectors; The sentence sub-vectors are aggregated to obtain a target vector.

[0009] According to an automatic speech recognition method based on echo cancellation provided by the present invention, a pre-built vector database is used to compare semantic similarity with the target vector, specifically comprising: The sentence sub-vectors are respectively compared with all speech interaction vectors in the vector database for semantic similarity to obtain comparison results.

[0010] According to an automatic speech recognition method based on echo cancellation provided by the present invention, echo cancellation is performed according to the comparison result to obtain an echo cancellation result, which specifically includes: When the comparison result is that there is vector data in the vector database that is semantically similar to the sentence sub-vector, the vector data with similar semantics is removed from the sentence sub-vector to obtain a removal result; The elimination results corresponding to all the sentence sub-vectors are concatenated to obtain the echo cancellation result.

[0011] According to an automatic speech recognition method based on echo cancellation provided by the present invention, the calculation method of the semantic similarity includes the cosine function of the vector and the large model of AIGC.

[0012] According to an automatic speech recognition method based on echo cancellation provided by the present invention, the preset vectorization algorithm includes an Embedding algorithm.

[0013] The present invention also provides an automatic speech recognition device based on echo cancellation, comprising the following modules: An extraction unit, used to extract the audio part of the voice information to be recognized, and perform voice recognition using a preset voice recognition technology to obtain a voice recognition result; A vector unit, used to vectorize the speech recognition result using a preset vectorization algorithm to obtain a target vector; A comparison unit, used to compare the semantic similarity between the pre-built vector database and the target vector, and perform echo cancellation according to the comparison result to obtain an echo cancellation result; A recognition unit, configured to perform automatic speech recognition according to the echo cancellation result; Among them, the vector database includes a large number of voice interaction vectors; the voice interaction vectors are obtained by performing voice recognition using a preset voice recognition technology and then continuing to perform vectorization using the preset vectorization algorithm based on the voice interaction information output by the intelligent robot or digital human.

[0014] The present invention also provides an electronic device, 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, the automatic speech recognition method based on echo cancellation as described above is implemented.

[0015] The present invention also provides a non-transitory computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the automatic speech recognition method based on echo cancellation as described above is implemented.

[0016] The present invention also provides a computer program product, comprising a computer program, wherein when the computer program is executed by a processor, the automatic speech recognition method based on echo cancellation as described above is implemented.

[0017] The automatic speech recognition method and device based on echo cancellation provided by the present invention extract the audio part of the speech information to be recognized, and use the preset speech recognition technology to perform speech recognition to obtain the speech recognition result; use the preset vectorization algorithm to vectorize the speech recognition result to obtain the target vector; use the pre-constructed vector database to compare the semantic similarity with the target vector, and perform echo cancellation according to the comparison result to obtain the echo cancellation result; perform automatic speech recognition according to the echo cancellation result; wherein the vector database includes a large number of speech interaction vectors; the speech interaction vector is obtained by using the preset speech recognition technology to perform speech recognition according to the speech interaction information output by the intelligent robot or digital human, and then continue to use the preset vectorization algorithm to perform vectorization. The present invention intelligently analyzes and processes the semantic information in the speech signal through the speech recognition technology, and converts the semantic information into a vector to compare with the information output by the intelligent robot or digital human after similar processing, and recognizes and eliminates the echo according to the comparison result, so as to achieve high-efficiency and good-adaptability echo cancellation, thereby being able to improve the accuracy of speech recognition without sacrificing the quality of the speech signal and the semantic integrity. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0019] Figure 1 This is one of the flow charts of the automatic speech recognition method based on echo cancellation provided by the present invention.

[0020] Figure 2 This is the second flow chart of the automatic speech recognition method based on echo cancellation provided by the present invention.

[0021] Figure 3 It is a structural schematic diagram of an automatic speech recognition device based on echo cancellation provided by the present invention.

[0022] Figure 4 It is a structural schematic diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION

[0023] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0024] Combine the following Figure 1-Figure 2 Describe the automatic speech recognition method based on echo cancellation of the present invention, Figure 1 FIG. 1 is one of the flow charts of the automatic speech recognition method based on echo cancellation provided by the present invention, such as Figure 1 As shown, the method comprises the following steps: Step 110: extract the audio portion of the voice information to be recognized, and perform voice recognition using a preset voice recognition technology to obtain a voice recognition result.

[0025] It should be noted that the voice information to be recognized is sound information of human voice input that may have an echo. The voice information to be recognized is recognized using a pre-selected voice recognition technology and a recognized voice recognition result is obtained.

[0026] Among them, it should be noted that the present invention does not impose specific restrictions on speech recognition technology, and it can be selected according to actual conditions. In some embodiments, speech recognition technology includes: natural language processing algorithm (NLP), template matching-based speech recognition (Template Matching ASR), hidden Markov model (HMM), dynamic time warping (DTW), artificial neural network (ANN), convolutional neural network (CNN), recurrent neural network (RNN) and its variants long short-term memory network (LSTM) and gated recurrent unit (GRU). The specific implementation method is not described in detail in the present invention. It should be emphasized that the speech recognition result obtained by using the existing speech recognition technology is a text result, and due to the presence of echo in the speech information to be recognized, the speech recognition result will contain more sentences than the original human voice input.

[0027] Furthermore, in some embodiments, the preset vectorization algorithm includes an Embedding algorithm.

[0028] Step 120: vectorize the speech recognition result using a preset vectorization algorithm to obtain a target vector.

[0029] In step 120, the sentences in the speech recognition result obtained in step 110 are vectorized to obtain a target vector. It should be noted that the purpose of vectorization is to compare the data in the vector database in the form of a vector.

[0030] Step 130: perform semantic similarity comparison between the pre-built vector database and the target vector, perform echo cancellation based on the comparison result, and obtain an echo cancellation result.

[0031] In step 130, the target vector obtained in step 120 is compared with the vector stored in the vector database to find out whether there are sentences with similar pronunciation.

[0032] It should be emphasized that since the speech recognition result with echo may have errors with the audio recognition result of the original intelligent robot or digital human, the present invention uses vectors for semantic similarity comparison instead of complete string matching for semantic similarity comparison.

[0033] Furthermore, the vector database includes a large number of voice interaction vectors; the voice interaction vectors are obtained by performing voice recognition using a preset voice recognition technology and then continuing to perform vectorization using the preset vectorization algorithm based on the voice interaction information output by the intelligent robot or digital human.

[0034] It should be explained that the vector database stores at least the output of the intelligent robot or digital human in the previous interaction, that is, the output of the intelligent robot or digital human in the interaction targeted by the sound information of the human voice input in the current voice information to be recognized.

[0035] Based on this, in the specific implementation process, Figure 2 As shown, before step 110, step 100 is also included: in response to the intelligent robot or digital human outputting voice interaction information, voice recognition is performed on the voice interaction information using a preset voice recognition technology to obtain an interaction recognition result, and the interaction recognition result is vectorized using the preset vectorization algorithm to obtain a voice interaction vector and store it in a vector database.

[0036] It should be noted that the voice of an intelligent robot or digital person often comes from an audio or video file or a real-time audio or video stream. Regardless of the form, the audio part can be extracted and the existing speech recognition technology can be used for speech recognition to obtain interactive recognition results. In addition, in order to facilitate subsequent semantic similarity matching, the interactive recognition results are vectorized and stored in a vector database.

[0037] Afterwards, echo cancellation is performed according to the comparison result to obtain an echo cancellation result.

[0038] It can be understood that, unlike the traditional methods that rely on filters or large machine learning models, the present invention identifies and removes echoes by intelligently analyzing and processing the semantic information in the voice signal, and only removes the echo components that match the output of the robot or digital human, retaining the integrity of human speech, and thus can significantly improve the accuracy of voice recognition. At the same time, the application of voice recognition technology reduces the demand for hardware resources, and can significantly reduce the demand for computing resources and the energy consumption of the system. In addition, voice recognition technology has good applicability, avoids long model training and adjustment processes, and can flexibly respond to changing interactive environments and voice inputs without frequent configuration adjustments or retraining, thereby improving the applicability and flexibility of the system.

[0039] Step 140: Perform automatic speech recognition according to the echo cancellation result.

[0040] In step 140, the automatic speech recognition is continued based on the echo cancellation result obtained in step 130 as input for subsequent interaction with the intelligent robot and digital human.

[0041] The present invention aims to solve the challenges faced by the prior art when processing voice signals containing echoes, especially in the context of the widespread application of intelligent robots and digital humans, how to effectively eliminate echo interference and maintain the integrity of voice information and the naturalness of interaction without adding additional costs and complex configurations. In the implementation process, the present invention uses speech recognition technology to perform echo elimination to improve the accuracy and efficiency of speech recognition in human-computer interaction of intelligent robots and digital humans, while reducing implementation costs and complexity, and providing users with a better interactive experience. The present invention not only provides a new solution for human-computer interaction of intelligent robots and digital humans, but also opens up a new path for further research and optimization of ASR technology.

[0042] The vectorized speech recognition result is further described below. In some embodiments, the speech recognition result is vectorized using a preset vectorization algorithm to obtain a target vector, specifically including: Vectorizing the sentences in the speech recognition result respectively using a preset vectorization algorithm to obtain sentence sub-vectors; The sentence sub-vectors are aggregated to obtain a target vector.

[0043] Specifically, each sentence in the speech recognition result is vectorized to obtain the sentence sub-vector corresponding to each sentence, and then the target vector is obtained by summarizing them.

[0044] The semantic similarity comparison is further explained below. In some embodiments, using a pre-built vector database to perform semantic similarity comparison with the target vector specifically includes: The sentence sub-vectors are respectively compared with all speech interaction vectors in the vector database for semantic similarity to obtain comparison results.

[0045] Specifically, in the process of semantic similarity comparison, the vectorization result (ie, sentence sub-vector) of each sentence in the speech recognition result can be compared with the vector stored in the vector database to find out whether there are phonetically similar sentences as the comparison result.

[0046] The echo cancellation is further explained below. In some embodiments, the echo cancellation is performed according to the comparison result to obtain the echo cancellation result, which specifically includes: When the comparison result is that there is vector data in the vector database that is semantically similar to the sentence sub-vector, the vector data with similar semantics is removed from the sentence sub-vector to obtain a removal result; The elimination results corresponding to all the sentence sub-vectors are concatenated to obtain the echo cancellation result.

[0047] Specifically, based on the above embodiment, since the comparison is performed on each sentence sub-vector respectively, it is necessary to analyze the comparison results corresponding to the sentence sub-vectors respectively.

[0048] In the comparison results, if the comparison result shows that the sentence sub-vector being compared has semantic similarity with the necklace in the vector database, it is considered that the sentence corresponding to the sentence sub-vector is the echo generated by the output of the intelligent robot or digital human, and it needs to be discarded and eliminated. Otherwise, it is retained, and finally the elimination result corresponding to the sentence sub-vector is obtained.

[0049] The elimination results corresponding to all the sentence sub-vectors are concatenated to form the recognition result, i.e., the echo cancellation result.

[0050] It should be noted that the embodiments of the present invention do not limit the calculation method of semantic similarity, and the calculation method can be selected according to actual conditions. Further, in some embodiments, the calculation method of semantic similarity includes a cosine function of a vector and a large model of AIGC (Artificial Intelligence Generated Content).

[0051] The embodiments of the present invention can effectively distinguish the voice input of a human user from the echo generated by the machine by accurately identifying and comparing the semantic content of the voice output by the intelligent robot or digital human and the received voice signal containing the echo, thereby ensuring that only the user's voice content is transmitted to the intelligent robot or digital human for subsequent interaction.

[0052] The automatic speech recognition method based on echo cancellation provided by the present invention extracts the audio part of the speech information to be recognized, and performs speech recognition using a preset speech recognition technology to obtain a speech recognition result; the speech recognition result is vectorized using a preset vectorization algorithm to obtain a target vector; a semantic similarity comparison is performed between a pre-constructed vector database and the target vector, and echo cancellation is performed according to the comparison result to obtain an echo cancellation result; automatic speech recognition is performed according to the echo cancellation result; wherein the vector database includes a large number of speech interaction vectors; the speech interaction vector is obtained by performing speech recognition using a preset speech recognition technology and then continuing to perform vectorization using the preset vectorization algorithm based on the speech interaction information output by an intelligent robot or a digital human. The present invention intelligently analyzes and processes the semantic information in the speech signal through speech recognition technology, and converts the semantic information into a vector and compares it with the information output by the intelligent robot or digital human that has been similarly processed, and recognizes and eliminates the echo according to the comparison result, thereby realizing highly efficient and adaptable echo cancellation, thereby being able to improve the accuracy of speech recognition without sacrificing the quality of the speech signal and the semantic integrity.

[0053] The automatic speech recognition device based on echo cancellation provided by the present invention is described below. The automatic speech recognition device based on echo cancellation described below and the automatic speech recognition method based on echo cancellation described above can be referred to each other. Figure 3 As shown, the device includes the following modules: The extraction unit 310 is used to extract the audio part of the voice information to be recognized, and perform voice recognition using a preset voice recognition technology to obtain a voice recognition result; A vector unit 320, configured to vectorize the speech recognition result using a preset vectorization algorithm to obtain a target vector; A comparison unit 330 is used to compare the semantic similarity between the pre-built vector database and the target vector, and perform echo cancellation according to the comparison result to obtain an echo cancellation result; A recognition unit 340, configured to perform automatic speech recognition according to the echo cancellation result; Among them, the vector database includes a large number of voice interaction vectors; the voice interaction vectors are obtained by performing voice recognition using a preset voice recognition technology and then continuing to perform vectorization using the preset vectorization algorithm based on the voice interaction information output by the intelligent robot or digital human.

[0054] According to an automatic speech recognition device based on echo cancellation provided by the present invention, the speech recognition result is vectorized by using a preset vectorization algorithm to obtain a target vector, which specifically includes: Vectorizing the sentences in the speech recognition result respectively using a preset vectorization algorithm to obtain sentence sub-vectors; The sentence sub-vectors are aggregated to obtain a target vector.

[0055] According to an automatic speech recognition device based on echo cancellation provided by the present invention, a semantic similarity comparison is performed between a pre-built vector database and the target vector, specifically comprising: The sentence sub-vectors are respectively compared with all speech interaction vectors in the vector database for semantic similarity to obtain comparison results.

[0056] According to an automatic speech recognition device based on echo cancellation provided by the present invention, echo cancellation is performed according to the comparison result to obtain an echo cancellation result, which specifically includes: When the comparison result is that there is vector data in the vector database that is semantically similar to the sentence sub-vector, the vector data with similar semantics is removed from the sentence sub-vector to obtain a removal result; The elimination results corresponding to all the sentence sub-vectors are concatenated to obtain the echo cancellation result.

[0057] According to an automatic speech recognition device based on echo cancellation provided by the present invention, the calculation method of the semantic similarity includes the cosine function of the vector and the large model of AIGC.

[0058] According to an automatic speech recognition device based on echo cancellation provided by the present invention, the preset vectorization algorithm includes an Embedding algorithm.

[0059] The automatic speech recognition device based on echo cancellation provided by the present invention extracts the audio part of the speech information to be recognized, and performs speech recognition using a preset speech recognition technology to obtain a speech recognition result; vectorizes the speech recognition result using a preset vectorization algorithm to obtain a target vector; compares the semantic similarity between the pre-constructed vector database and the target vector, performs echo cancellation according to the comparison result, and obtains an echo cancellation result; performs automatic speech recognition according to the echo cancellation result; wherein the vector database includes a large number of speech interaction vectors; the speech interaction vector is obtained by performing speech recognition using a preset speech recognition technology and then continuing to perform vectorization using the preset vectorization algorithm based on the speech interaction information output by an intelligent robot or digital human. The present invention intelligently analyzes and processes the semantic information in the speech signal through speech recognition technology, and converts the semantic information into a vector and compares it with the information output by the intelligent robot or digital human that has been similarly processed, identifies and eliminates the echo according to the comparison result, and realizes high-efficiency and adaptable echo cancellation, thereby being able to improve the accuracy of speech recognition without sacrificing the quality of the speech signal and the semantic integrity.

[0060] Figure 4 An example of a physical structure diagram of an electronic device is shown in FIG. Figure 4 As shown, the electronic device may include: a processor 410, a communication interface 420, a memory 430 and a communication bus 440, wherein the processor 410, the communication interface 420 and the memory 430 communicate with each other through the communication bus 440. The processor 410 may call the logic instructions in the memory 430 to execute an automatic speech recognition method based on echo cancellation, the method comprising: extracting the audio part of the speech information to be recognized, and performing speech recognition using a preset speech recognition technology to obtain a speech recognition result; vectorizing the speech recognition result using a preset vectorization algorithm to obtain a target vector; performing semantic similarity comparison with the target vector using a pre-constructed vector database, performing echo cancellation according to the comparison result, and obtaining an echo cancellation result; performing automatic speech recognition according to the echo cancellation result; wherein the vector database includes a large number of speech interaction vectors; the speech interaction vector is obtained by performing speech recognition using a preset speech recognition technology and then continuing to perform vectorization using the preset vectorization algorithm according to the speech interaction information output by the intelligent robot or digital human.

[0061] In addition, the logic instructions in the above-mentioned memory 430 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.

[0062] On the other hand, the present invention also provides a computer program product, which includes a computer program. The computer program can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the automatic speech recognition method based on echo cancellation provided by the above-mentioned methods, which includes: extracting the audio part of the speech information to be recognized, and performing speech recognition using a preset speech recognition technology to obtain a speech recognition result; vectorizing the speech recognition result using a preset vectorization algorithm to obtain a target vector; performing semantic similarity comparison with the target vector using a pre-constructed vector database, and performing echo cancellation based on the comparison result to obtain an echo cancellation result; performing automatic speech recognition based on the echo cancellation result; wherein the vector database includes a large number of speech interaction vectors; the speech interaction vector is obtained by performing speech recognition using a preset speech recognition technology based on the speech interaction information output by the intelligent robot or digital human, and then continuing to perform vectorization using the preset vectorization algorithm.

[0063] On the other hand, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to execute the automatic speech recognition method based on echo cancellation provided by the above-mentioned methods, the method comprising: extracting the audio part of the speech information to be recognized, and performing speech recognition using a preset speech recognition technology to obtain a speech recognition result; vectorizing the speech recognition result using a preset vectorization algorithm to obtain a target vector; performing semantic similarity comparison with the target vector using a pre-constructed vector database, performing echo cancellation based on the comparison result, and obtaining an echo cancellation result; performing automatic speech recognition based on the echo cancellation result; wherein the vector database includes a large number of speech interaction vectors; the speech interaction vector is obtained by performing speech recognition using a preset speech recognition technology and then vectorizing using the preset vectorization algorithm based on the speech interaction information output by the intelligent robot or digital human.

[0064] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0065] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0066] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. An automatic speech recognition method based on echo cancellation, characterized in that: include: Extracting the audio part of the voice information to be recognized, and performing voice recognition using a preset voice recognition technology to obtain a voice recognition result; Vectorizing the speech recognition result using a preset vectorization algorithm to obtain a target vector; Using a pre-built vector database to compare semantic similarity with the target vector, performing echo cancellation according to the comparison result, and obtaining an echo cancellation result; Performing automatic speech recognition according to the echo cancellation result; Among them, the vector database includes a large number of voice interaction vectors; the voice interaction vectors are obtained by performing voice recognition using a preset voice recognition technology and then continuing to perform vectorization using the preset vectorization algorithm based on the voice interaction information output by the intelligent robot or digital human.

2. The automatic speech recognition method based on echo cancellation according to claim 1, characterized in that: The speech recognition result is vectorized using a preset vectorization algorithm to obtain a target vector, specifically including: Vectorizing the sentences in the speech recognition result respectively using a preset vectorization algorithm to obtain sentence sub-vectors; The sentence sub-vectors are aggregated to obtain a target vector.

3. The automatic speech recognition method based on echo cancellation according to claim 2, characterized in that: Using a pre-built vector database to compare semantic similarity with the target vector, specifically including: The sentence sub-vectors are respectively compared with all speech interaction vectors in the vector database for semantic similarity to obtain comparison results.

4. The automatic speech recognition method based on echo cancellation according to claim 3, characterized in that: Echo cancellation is performed according to the comparison result to obtain an echo cancellation result, which specifically includes: When the comparison result is that there is vector data in the vector database that is semantically similar to the sentence sub-vector, the vector data with similar semantics is removed from the sentence sub-vector to obtain a removal result; The elimination results corresponding to all the sentence sub-vectors are concatenated to obtain the echo cancellation result.

5. The automatic speech recognition method based on echo cancellation according to claim 1, characterized in that: The semantic similarity calculation method includes the cosine function of the vector and the large model of AIGC.

6. The automatic speech recognition method based on echo cancellation according to claim 1, characterized in that: The preset vectorization algorithm includes an Embedding algorithm.

7. An automatic speech recognition device based on echo cancellation, characterized in that: include: An extraction unit, used to extract the audio part of the voice information to be recognized, and perform voice recognition using a preset voice recognition technology to obtain a voice recognition result; A vector unit, used to vectorize the speech recognition result using a preset vectorization algorithm to obtain a target vector; A comparison unit, used to compare the semantic similarity between the pre-built vector database and the target vector, and perform echo cancellation according to the comparison result to obtain an echo cancellation result; A recognition unit, configured to perform automatic speech recognition according to the echo cancellation result; Among them, the vector database includes a large number of voice interaction vectors; the voice interaction vectors are obtained by performing voice recognition using a preset voice recognition technology and then continuing to perform vectorization using the preset vectorization algorithm based on the voice interaction information output by the intelligent robot or digital human.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the automatic speech recognition method based on echo cancellation as claimed in any one of claims 1 to 6 is implemented.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the automatic speech recognition method based on echo cancellation as claimed in any one of claims 1 to 6 is implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the automatic speech recognition method based on echo cancellation as claimed in any one of claims 1 to 6 is implemented.

Images