Speech recognition method of intelligent chemical inspection robot

Through voice recognition technology and deep neural network to identify user identity and permissions, and split control instructions, the problem of unsafe remote control of chemical inspection robots is solved, and efficient and safe control is achieved in harsh environments.

CN118968996AInactive Publication Date: 2024-11-15TIANJIN BONUO ZHICHUANG ROBOT TECH CO LTD

Patent Information

Application Number
CN202411440686.6
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-10-16
Publication Date
2024-11-15
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

Existing chemical inspection robots generally adopt remote remote control control methods, which are not very intelligent and safe, making it difficult to effectively control in harsh environments such as high temperature and high pressure.

Method used

Using voice recognition technology, preprocessing and feature extraction by obtaining user voice signals, combining deep neural networks and permission databases, identifying user identity and permissions, splitting and matching control instructions, setting permission levels, ensuring the security and accuracy of voice control sex.

Benefits of technology

It improves the control safety and accuracy of chemical inspection robots in dangerous environments, avoids misoperation of non-professional personnel, and enhances production safety and efficiency.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118968996A_ABST
    Figure CN118968996A_ABST
Patent Text Reader

Abstract

The invention relates to the technical field of voice recognition, in particular to a voice recognition method of an intelligent chemical inspection robot, which comprises the following steps: acquiring a voice signal sent by a user, preprocessing and extracting voice features; calculating the similarity between the voice feature of the user and any voice feature in an authority database, and comparing the similarity with a preset similarity to judge the authority level of the user; recognizing the voice as a text, and optimizing and improving the recognized text; splitting a text containing a plurality of control instructions into a plurality of control texts, and calculating the matching degree with the control instructions in the instruction library; determining a corresponding control instruction and a control instruction permission level, and comparing the control instruction permission level with the user permission level; when the semantics of the user is fuzzy, displaying similar control instructions for the user to select, and updating a personal instruction library of the user; the security of voice control is ensured by setting the authority level, and the practicability is improved by performing personalized setting on the control instruction.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of speech recognition technology, and in particular to a speech recognition method for an intelligent industrial inspection robot. Background Art

[0002] The chemical industry, due to its particularity and dangerousness, will bring many safety issues. It is necessary to conduct regular inspections of the chemical industry to ensure the production safety of the chemical industry. Compared with manual inspection, chemical inspection robots have many advantages, including high safety, high inspection efficiency, strong sustainability, high data accuracy and low cost. Chemical inspection robots can work in harsh environments such as high temperature and high radiation to ensure the safety of workers. They can not only replace manual inspection and monitoring tasks, avoiding personnel working in dangerous environments, but also detect potential problems in advance through real-time monitoring and data analysis, avoid accidents, and improve the safety level of chemical industry production. In addition, chemical inspection robots can realize 24-hour uninterrupted inspection and monitoring, providing more continuous monitoring, which is crucial to ensure the continuous operation of the production line and timely response to potential problems.

[0003] Chinese Patent Publication No.: CN107797561A discloses an explosion-proof robot inspection system and method, which discloses an inspection system and method. The inspection system includes a management platform and an explosion-proof robot. The management platform is connected to the explosion-proof robot. When the explosion-proof robot receives a task from the management platform, the explosion-proof robot can autonomously navigate to the target explosion-proof area for inspection. After establishing a communication connection between the management platform and the explosion-proof robot, the management personnel can operate the explosion-proof robot in the background by controlling the management platform, thereby replacing manual entry into dangerous areas for inspection; it can be seen that the explosion-proof robot inspection system and method have the following problems: Under the current existing technology, inspection robots generally adopt remote control control, and control commands such as forward, backward, left, and right of the inspection robot are realized by operating the control buttons on the remote controller. The intelligence and safety of this remote control method are not high. Summary of the invention

[0004] To this end, the present invention provides a voice recognition method for an intelligent industrial inspection robot, which is used to overcome the problem that the inspection robot in the prior art generally adopts a remote control control method, and the inspection robot is controlled to move forward, backward, left, right, etc. by operating the control buttons on the remote controller. The intelligence and safety of this remote control method are not high.

[0005] To achieve the above object, the present invention provides a speech recognition method for an intelligent chemical inspection robot, comprising: Step S1, obtaining a voice signal sent by a user, preprocessing the collected voice signal and extracting voice features; Step S2, calculating the similarity between the user's voice feature and any voice feature in the authority database, comparing the similarity with a preset similarity, determining the user's authority level based on the comparison result, and identifying the user's identity; Step S3, recognizing the voice input by the user into text, and optimizing and improving the recognized text; Step S4, if the text contains multiple control texts, split the text into several control texts, extract keywords from each control text and calculate the matching degree with the control instructions in the instruction library; Step S5, determining the control instruction and the control instruction authority level corresponding to the control text according to the matching degree, and comparing the control instruction authority level with the user authority level to determine whether the control instruction exceeds the user authority level; Step S6, if the user's semantics are unclear, similar control instructions are matched for the user to select, and the text data of the control instructions in the user's personal instruction library is updated according to the frequency of selection; Step S7, calculating the frequency of user calling control instructions, dividing the control instructions into primary calling instructions and secondary calling instructions, and dividing the instruction library into primary instruction library and secondary instruction library according to the calling instruction level.

[0006] The collected speech signals are pre-processed by pre-filtering, noise reduction, normalization, etc. Pre-filtering is to filter out noise in some other frequency bands of the input signal. Noise reduction technology helps to remove background noise that may be introduced during the recording process. The normalization process involves adjusting the amplitude of the sound signal to ensure that the signal has a consistent energy level, making it easier to compare sounds from different sources.

[0007] For the preprocessed speech signal, the commonly used feature extraction methods MFCC (Mel Frequency Cepstrum Coefficient) or HMM (Hidden Markov Model) are used to extract sound features. By analyzing the frequency, energy, harmonics and other features of the speech signal, the speech features of the user's speech signal are extracted to provide input for the subsequent speech feature similarity algorithm.

[0008] Numbering a number of user voice signal features in the authority database, where any user voice signal feature in the authority database has a corresponding user authority level; For the extracted user voice signal features, the sound similarity matching algorithm is used to calculate the similarity between the user voice signal features and any user voice signal features in the authority database, and the similarity is compared with the preset similarity standard value. If the similarity between the user's voice feature and any voice feature in the authority database is greater than or equal to the preset similarity standard value, it is determined that the user matches the voice signal feature of any user in the authority database, and the user is determined to have the authority level to voice-control the intelligent inspection robot, and the following method steps are continued; If the similarities between the user's voice features and any voice features in the authority database are all less than the preset similarity standard value, it is determined that the user does not match any user's voice signal features in the authority database, the user does not have the authority level to voice-control the intelligent chemical inspection robot, and the following method steps are stopped.

[0009] For the preprocessed speech signal, the speech signal is input into the trained deep neural network learning model to be recognized as text, and post-processing techniques such as error correction and grammar correction are applied to optimize and improve the recognition results using the trained language model and error correction algorithm; The deep neural network learning model is a speech recognition model, which is obtained by collecting a preset number of speech and inputting the speech into a preset neural network of the deep neural network learning model for deep learning training; The language model is trained based on a large amount of text data and is used to predict the probability of words in the language; the error correction algorithm identifies and corrects errors based on the prediction results of the language model and the contextual information of the words.

[0010] Based on the optimized text of the voice signal recognition input by the user, the NER algorithm is used to perform named entity recognition on the text to identify the entity names in the text, such as object names, place names, and region names; the POS algorithm is used to perform part-of-speech tagging on the text to mark the parts of speech in the text, such as nouns, verbs, adjectives, etc., extract keywords from the text, divide the keywords into general modification keywords, mobile keywords, and functional keywords, and judge whether the recognized text contains multiple control instructions according to the number of mobile keywords and functional keywords; In the case where the text contains multiple control texts, the text is input into a pre-trained segmentation model and split into several control texts. For any control text, the matching degree between it and the control instructions in the control instruction library is calculated; The control instructions in the control instruction library are numbered. Each control instruction in the control instruction library has a corresponding control instruction authority level. Calculate the matching degree between any control text in the text and any control instruction in the control instruction library according to the text keywords, and compare the matching degree with the preset matching degree standard value. If the matching degree between the control text and any control instruction in the control instruction library is greater than or equal to a preset matching degree standard value, then the control instruction is determined to be the control instruction corresponding to the control text in the text signal; Compare the permission level corresponding to the control instruction with the user permission level corresponding to the user, If the control command authority level is within the range of the user authority level, it is determined that the user can control the intelligent inspection robot to execute this control command; If the control command authority level exceeds the user authority level range, it is determined that the user cannot control the intelligent inspection robot to execute this control command; If the matching degree between the control text and any control instruction in the control instruction library is less than the preset matching degree standard value, it is determined that there is no control instruction corresponding to the control text in the control instruction library, and the semantics of the user input is ambiguous or unclear.

[0011] In the case where the semantics of the user input is ambiguous or the recognition is unclear, the control instruction with the greatest similarity to the voice signal input by the user is retrieved from the control instruction library and displayed on the screen or display device for the user to select. The replacement frequency of the control instruction is calculated according to the number of times the user selects a similar control instruction displayed on the display device within a certain period. If the replacement frequency is greater than or equal to the preset replacement frequency, it is determined that the frequency of the user selecting similar instructions is high, and the text data of the control instruction selected by the user in the personal instruction library is updated according to the user's selection result; If the replacement frequency is less than the preset replacement frequency, it is determined that the user's frequency of selecting similar instructions is low, and there is no need to update the user's personal instruction library.

[0012] Set up a user's personal instruction library corresponding to the user, calculate the frequency of the user calling any control instruction, If the calling frequency is greater than or equal to the preset calling frequency, it is determined that the user calls this control instruction at a high frequency, and this control instruction is set as a first-level calling instruction, and the instruction library containing all first-level calling instructions is divided into a first-level instruction library; If the calling frequency is less than the preset calling frequency, it is determined that the user calls this control instruction at a low frequency, and this control instruction is set as a secondary calling instruction, and the instruction library containing all secondary calling instructions is divided into a secondary instruction library.

[0013] Compared with the prior art, the beneficial effect of the present invention is that the intelligent inspection robot works in a dangerous environment of high temperature, high pressure, toxicity and harm. The voice-controlled intelligent inspection robot requires technical personnel to perform professional control command voice input. By setting user authority levels for users of the voice-controlled intelligent inspection robot, non-professionals are prohibited from voice-controlling the intelligent inspection robot, and unintentional voice input to the intelligent inspection robot by irrelevant personnel is avoided.

[0014] Furthermore, the traditional speech recognition process involves multiple steps, such as feature extraction, acoustic model training and decoding, while the end-to-end training method starts directly from the original speech signal and realizes the end-to-end conversion from acoustic features to text through a deep learning model, which simplifies the traditional process and inputs a large amount of speech into the model for training in advance, thereby improving the accuracy and robustness of speech recognition.

[0015] Furthermore, since the recognition accuracy of existing speech recognition algorithms is not absolutely 100%, and the staff's Mandarin pronunciation may not be standard, and even worse, the staff controls the intelligent chemical inspection robot in dialect, which can lead to speech recognition errors. Therefore, when the inspection robot's motion voice control instructions are recognized as text data, recognition errors may occur, and text signal post-processing is required to optimize and improve the recognition results using the trained language model and error correction algorithm.

[0016] Furthermore, inspection robots have multiple functions in the chemical industry, including personnel safety monitoring, infrared temperature measurement, smoke and open flame detection, meter reading, equipment status identification, noise detection, switch identification, intrusion detection, and leakage detection, etc. These functions are crucial to ensuring production safety and improving efficiency. Any inaccurate or inappropriate operation instructions input to the intelligent inspection robot will have unimaginable consequences for the chemical production inspection process and the inspection robot itself, seriously affecting the production safety and production efficiency of chemical production inspection. By setting control instruction authority levels for the intelligent inspection robot, users with insufficient authority levels are restricted from selecting control actions for the intelligent inspection robot, thereby avoiding property losses and damaging the production safety and production efficiency of chemical production inspection.

[0017] Furthermore, each user who controls the intelligent inspection robot has different pronunciation and word usage habits, so there may be a situation where the voice signal input by the user does not match the control instructions in the control instruction library. A user personal instruction library is established based on the user's voice and word usage habits, and the text data of the control instructions in the personal instruction library is updated in real time based on the user's selection results, thereby improving the efficiency of user voice control of the intelligent inspection robot and increasing the practicality and accuracy of the method. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 This is a step diagram of a voice recognition method for an intelligent industrial inspection robot in an embodiment of the present invention; Figure 2 This is a flow chart of determining the user authority level according to the user voice characteristics in an embodiment of the present invention; Figure 3 The present invention is a flowchart for determining whether a user control instruction exceeds the user authority level in an embodiment of the present invention. DETAILED DESCRIPTION

[0019] In order to make the objects and advantages of the present invention more clearly understood, the present invention is further described below in conjunction with embodiments; it should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0020] The preferred embodiments of the present invention are described below with reference to the accompanying drawings. It should be understood by those skilled in the art that these embodiments are only used to explain the technical principles of the present invention and are not intended to limit the protection scope of the present invention.

[0021] It should be noted that, in the description of the present invention, terms such as "up", "down", "left", "right", "inside" and "outside" indicating directions or positional relationships are based on the directions or positional relationships shown in the drawings. This is merely for the convenience of description and does not indicate or imply that the device or element must have a specific orientation, be constructed and operated in a specific orientation. Therefore, it cannot be understood as a limitation on the present invention.

[0022] In addition, it should be noted that in the description of the present invention, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For those skilled in the art, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0023] See also Figure 1-Figure 3 As shown, Figure 1 This is a step diagram of a voice recognition method for an intelligent industrial inspection robot in an embodiment of the present invention; Figure 2 This is a flow chart of determining the user authority level according to the user voice characteristics in an embodiment of the present invention; Figure 3 The present invention is a flowchart for determining whether a user control instruction exceeds the user authority level in an embodiment of the present invention.

[0024] The present invention provides a speech recognition method for an intelligent chemical inspection robot, comprising: Step S1, obtaining a voice signal sent by a user after the user turns on the input device, and preprocessing the collected voice signal; Step S2, extracting speech features from the preprocessed speech signal using a feature extraction method; Step S3, calculating the similarity between the user's voice feature and any voice feature in the authority database, comparing the similarity with a preset similarity, determining the user's authority level based on the comparison result, and identifying the user's identity; Step S4, recognizing the voice input by the user into text, and optimizing and improving the recognized text; Step S5, if the text contains multiple control instructions, split the text into several control texts, extract keywords from the control texts and calculate the matching degree with the control instructions in the instruction library; Step S6, determining the control instruction and the control instruction authority level corresponding to the control text according to the matching degree, and comparing the control instruction authority level with the user authority level to determine whether the control instruction exceeds the user authority level; Step S7, in the case where the user's semantics are ambiguous and unclear, similar control instructions are displayed for the user to choose, and the text data of the control instructions in the user's personal instruction library is updated.

[0025] The beneficial effects of the present invention are that it can recognize multiple control instructions contained in a voice signal at one time, and ensure the security of voice control by setting authority levels, and improve practicality by personalizing the control instructions.

[0026] The collected speech signals are pre-processed by pre-filtering, noise reduction, normalization, etc. Pre-filtering is to filter out noise in some other frequency bands of the input signal. Noise reduction technology helps to remove background noise that may be introduced during the recording process. The normalization process involves adjusting the amplitude of the sound signal to ensure that the signal has a consistent energy level, making it easier to compare sounds from different sources.

[0027] For the preprocessed speech signal, the commonly used feature extraction methods MFCC (Mel Frequency Cepstral Coefficient) or HMM (Hidden Markov Model) are used to extract sound features. By analyzing the frequency, energy, harmonics and other features of the speech signal, the voiceprint features of the user's speech signal are extracted to provide input for the subsequent speech feature similarity algorithm.

[0028] The user voice signal features in the authority database are numbered and recorded as user authority voice signal feature A1, user authority voice signal feature A2, ..., user authority voice signal feature An. The user authority level corresponding to any user voice signal feature Ai in the authority database is αi, and the user authority level α=1, 2, ..., n, For the extracted user voice signal features, the sound similarity matching algorithm is used to calculate the similarity Fi between the user voice signal features and any user voice signal features Ai, i=1,2,...,n, in the authority database, and the similarity Fi is compared with the preset similarity standard value Fie, where Fie is a preset value. In this embodiment, the preset similarity standard value Fie=90%. If Fi ≥ Fie, it is determined that the user matches the voice signal feature of any user in the authority database, the authority level of the voice-controlled intelligent chemical inspection robot is αi, and the following method steps are continued; If Fi<Fie, it is determined that the user does not match the voice signal characteristics of any user in the authority database, and does not have the authority level to voice-control the intelligent chemical inspection robot, and the following method steps are stopped.

[0029] Specifically, the intelligent inspection robot works in a dangerous environment with high temperature, high pressure, toxicity and harmful substances. The voice-controlled intelligent inspection robot requires technical personnel to input professional control commands by voice. By setting user authority levels for users of the voice-controlled intelligent inspection robot, non-professionals are prohibited from voice-controlling the intelligent inspection robot, and unintentional voice input to the intelligent inspection robot by unrelated personnel is also avoided.

[0030] For the preprocessed speech signal, the speech signal is input into the trained deep neural network learning model to be recognized as a text signal, and post-processing techniques such as error correction and grammar correction are applied. The trained language model and error correction algorithm are used to optimize and improve the recognition results.

[0031] Among them, the deep neural network learning model is a speech recognition model, which is obtained by collecting a preset number of speech and inputting the speech into a preset neural network of the deep neural network learning model for deep learning training.

[0032] The traditional speech recognition process involves multiple steps, such as feature extraction, acoustic model training and decoding. The end-to-end training method starts directly from the original speech signal and realizes the end-to-end conversion from acoustic features to text through a deep learning model. This simplifies the traditional process and inputs a large amount of speech into the model for training in advance, thus improving the accuracy and robustness of speech recognition. Since the recognition accuracy of existing speech recognition algorithms is not absolutely 100%, and the staff's Mandarin pronunciation may not be standard, and even worse, the staff may control the intelligent chemical inspection robot in dialect, which can lead to speech recognition errors. Therefore, when the inspection robot's motion voice control instructions are recognized as text data, recognition errors may occur, and text signal post-processing is required to optimize and improve the recognition results using the trained language model and error correction algorithm; The language model is trained based on a large amount of text data and is used to predict the probability of words in the language; the error correction algorithm identifies and corrects errors based on the prediction results of the language model and the contextual information of the words.

[0033] Based on the optimized text signal after the voice signal recognition input by the user, the NER algorithm is used to perform named entity recognition on the text to identify the entity names in the text, such as object names, place names, and region names; the POS algorithm is used to perform part-of-speech tagging on the text to mark the parts of speech in the text, such as nouns, verbs, adjectives, etc., extract the keywords in the text, divide the keywords into general modification keywords, mobile keywords, and functional keywords, and judge whether the recognized text contains multiple control instructions according to the number of mobile keywords and functional keywords; In the case where the text signal contains multiple control texts, the text signal is input into a pre-trained segmentation model to be split into several control texts, and numbered, respectively, as the first control text C1, the second control text C2, ..., the nth control text Cn. For any control text Ci, i=1,2,...,n, its matching degree with the control instructions in the control instruction library is calculated. The control instructions in the control instruction library are numbered and recorded as control instruction B1, control instruction B2, ..., control instruction Bn. Any control instruction Bi in the control instruction library, i=1,2,...,n, corresponds to a control instruction authority level βi, β=1,2,...,n. The matching degree Pi between any control text Ci in the text signal and any control instruction Bi in the control instruction library is calculated according to the text keywords, and the matching degree Pi is compared with a preset matching degree standard value Pie, where Pie is a preset value. In this embodiment, the preset matching degree standard value Pie=95%. If Pi ≥ Pie, then the control instruction corresponding to the control text Ci in the text signal is determined to be Bi. Compare the authority level βi corresponding to the control instruction Bi with the user authority level αi corresponding to the user, If βi≤αi, it is determined that the user can control the intelligent inspection robot to execute this control instruction. If βi>αi, it is judged that the control instruction authority level exceeds the user authority level, and the user cannot control the intelligent inspection robot to execute this control instruction. If Pi<Pie, it is determined that there is no control instruction corresponding to the control text Ci in the control instruction library, and the semantics of the user input is ambiguous or unclear.

[0034] Specifically, inspection robots have multiple functions in the chemical industry, including personnel safety monitoring, infrared temperature measurement, smoke and open flame detection, meter reading, equipment status identification, noise detection, switch identification, intrusion detection, and leakage detection. These functions are crucial to ensuring production safety and improving efficiency. Any inaccurate or inappropriate operation instructions input to the intelligent inspection robot will have unimaginable consequences for the chemical production inspection process and the inspection robot itself, seriously affecting the production safety and production efficiency of chemical production inspection. By setting control instruction authority levels for the intelligent inspection robot, users with insufficient authority levels are restricted from selecting control actions for the intelligent inspection robot, thereby avoiding property losses and damaging the production safety and production efficiency of chemical production inspection.

[0035] Wherein, in case of ambiguity or unclear recognition of the semantics of the user input, a control instruction similar to the text information after the recognition of the voice signal input by the user is retrieved from the control instruction library and displayed on the display device for the user to select; According to the number of times m1 that the user selects similar control instructions displayed on the display device within a certain period t1, the user's replacement frequency f1 is calculated, f1=m1÷t1, and a preset replacement frequency fe1 is set. If f1≥fe1, it is determined that the frequency of the user selecting similar instructions is high, and the text data of the control instruction selected by the user in the personal instruction library is updated according to the user's selection result; If f1<fe1, it is determined that the frequency of the user selecting similar instructions is low, and there is no need to update the user's personal instruction library.

[0036] Set up a personal instruction library corresponding to the user. The personal instruction library contains the text data of all control instructions under the user's authority level. According to the number m2 of times the user calls any control instruction Bi to the inspection robot within a certain period t1, the calling frequency f2 of the user calling the control instruction Bi is calculated, f2=m2÷t1, and the preset calling frequency fe2 is set. If f2 ≥ fe2, it is determined that the frequency of the user calling the control instruction Bi is high, the control instruction Bi is set as a first-level calling instruction, and the instruction library containing all first-level calling instructions is divided into a first-level instruction library; If f2<fe2, it is determined that the frequency of the user calling the control instruction Bi is low, the control instruction Bi is set as a secondary calling instruction, and the instruction library containing all secondary calling instructions is divided into a secondary instruction library.

[0037] Specifically, each user who controls the intelligent inspection robot has different pronunciation and word usage habits, so there may be a situation where the voice signal input by the user does not match the control instructions in the control instruction library. A user personal instruction library is established based on the user's voice and word usage habits, and the text data of the control instructions in the personal instruction library is updated in real time based on the user's selection results, thereby improving the efficiency of user voice control of the intelligent inspection robot and increasing the practicality and accuracy of the method.

[0038] So far, the technical solutions of the present invention have been described in conjunction with the preferred embodiments shown in the accompanying drawings. However, it is easy for those skilled in the art to understand that the protection scope of the present invention is obviously not limited to these specific embodiments. Without departing from the principle of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will fall within the protection scope of the present invention.

[0039] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A speech recognition method for an intelligent chemical inspection robot. It is characterized in that include: Step S1, obtaining a voice signal sent by a user, preprocessing the collected voice signal and extracting voice features; Step S2, calculating the similarity between the user's voice feature and any voice feature in the authority database, comparing the similarity with a preset similarity, determining the user's authority level based on the comparison result, and identifying the user's identity; Step S3, recognizing the voice input by the user into text, and optimizing and improving the recognized text; Step S4, if the text contains multiple control texts, split the text into several control texts, extract keywords from each control text and calculate the matching degree with the control instructions in the instruction library; Step S5, determining the control instruction and the control instruction authority level corresponding to the control text according to the matching degree, and comparing the control instruction authority level with the user authority level to determine whether the control instruction exceeds the user authority level; Step S6, if the user's semantics are unclear, similar control instructions are matched for the user to select, and the text data of the control instructions in the user's personal instruction library is updated according to the frequency of selection; Step S7, calculating the frequency of user calling control instructions, dividing the control instructions into primary calling instructions and secondary calling instructions, and dividing the instruction library into primary instruction library and secondary instruction library according to the calling instruction level.

2. The speech recognition method of the intelligent chemical inspection robot according to claim 1, It is characterized in that The collected voice signal is pre-filtered, denoised, and normalized, and the voice features of the user's voice signal are extracted by analyzing the frequency, energy, and harmonic characteristics of the voice signal.

3. The speech recognition method of the intelligent chemical inspection robot according to claim 2, It is characterized in that When the similarity between the user's voice features and the voice features in the authority database is compared with the preset similarity standard value, If the similarity between the user's voice feature and any voice feature in the authority database is greater than or equal to the preset similarity standard value, it is determined that the user matches the corresponding user voice signal feature in the authority database, and it is determined that the user has the authority level for voice-controlled intelligent inspection robots, and step S3 is continued; If the similarities between the user's voice features and any voice features in the authority database are all less than the preset similarity standard value, it is determined that the user's voice signal features do not match any user in the authority database, the user does not have the authority level to voice-control the intelligent chemical inspection robot, and step S3 is stopped.

4. The speech recognition method of the intelligent chemical inspection robot according to claim 3, It is characterized in that The preprocessed speech signal is input into the trained deep neural network learning model to be recognized as text, and the language model and error correction algorithm are applied to post-process the text, and the recognition results are corrected for text errors and grammar.

5. The speech recognition method of the intelligent chemical inspection robot according to claim 4, It is characterized in that For the recognized text information, the entity names in the text are identified, the text is tagged with parts of speech, and the keywords in the text are extracted, and the keywords are divided into general modification keywords, mobile keywords, and functional keywords.

6. The speech recognition method of the intelligent chemical inspection robot according to claim 5, It is characterized in that When the matching degree between the control text in the text signal and any control instruction in the control instruction library is compared with the preset matching degree standard value, If the matching degree between the control text and any control instruction in the control instruction library is greater than or equal to a preset matching degree standard value, then the control instruction is determined to be the control instruction corresponding to the control text in the text signal; If the matching degree between the control text and any control instruction in the control instruction library is less than the preset matching degree standard value, it is determined that there is no control instruction corresponding to the control text in the control instruction library, and the semantics of the user input is ambiguous or unclear.

7. The speech recognition method of the intelligent chemical inspection robot according to claim 6, It is characterized in that The number of control instructions in the recognized text is determined according to the number of mobile keywords and functional keywords, and the text signal is input into a pre-trained segmentation model and split into several control texts corresponding to a single control instruction.

8. The speech recognition method of the intelligent chemical inspection robot according to claim 7, It is characterized in that Compare the permission level corresponding to the control instruction with the user permission level corresponding to the user, If the control instruction authority level is within the user authority level range, it is determined that the user can control the intelligent inspection robot to execute this control instruction; If the control instruction authority level exceeds the user authority level, it is determined that the user cannot control the intelligent inspection robot to execute this control instruction.

9. The speech recognition method of the intelligent chemical inspection robot according to claim 8, It is characterized in that If the user's semantics is ambiguous or the recognition is unclear, the control instruction with the greatest similarity to the voice signal input by the user is retrieved from the control instruction library and displayed on the screen or display device for the user to select. The replacement frequency of the control instruction is calculated according to the number of times the user selects a similar control instruction displayed on the display device within a certain period. If the replacement frequency is greater than or equal to the preset replacement frequency, it is determined that the frequency of the user selecting similar instructions is high, and the text data of the control instruction selected by the user in the personal instruction library is updated according to the user's selection result; If the replacement frequency is less than the preset replacement frequency, it is determined that the user's frequency of selecting similar instructions is low, and there is no need to update the user's personal instruction library.

10. The speech recognition method of the intelligent chemical inspection robot according to claim 9, It is characterized in that Set up a user's personal instruction library corresponding to the user, calculate the frequency of the user calling any control instruction, If the calling frequency is greater than or equal to the preset calling frequency, it is determined that the user calls this control instruction at a high frequency, and this control instruction is set as a first-level calling instruction, and the instruction library containing all first-level calling instructions is divided into a first-level instruction library; If the calling frequency is less than the preset calling frequency, it is determined that the user calls this control instruction at a low frequency, and this control instruction is set as a secondary calling instruction, and the instruction library containing all secondary calling instructions is divided into a secondary instruction library.

Citation Information

Patent Citations

  • Explosion-proof robot inspection system and method

    CN107797561A

  • Method for remotely controlling television with voice and remote voice control

    CN102036033A

  • Information processing method, terminal and compute readable medium

    CN108227564A

  • Voice recognition method and system based on information interaction, and storage medium

    CN112562665A

  • Robot and control method thereof

    CN114310909A

Cited By

  • Voice interaction method based on artificial intelligence, medium and display terminal

    CN120412572A