Method, system and equipment for controlling mechanical arm through voice and medium

Abstract

The invention relates to the field of mechanical control, in particular to a method, system and equipment for controlling a mechanical arm through voice and a medium. The method comprises the following steps: collecting a voice control instruction signal of an electric power production field; noise reduction is carried out on the voice control instruction signal by adopting a spectrogram wavelet adaptive threshold noise reduction technology; converting the de-noised voice signal into a text or an instruction by using a pre-trained improved recurrent neural network model; and the mechanical arm control system executes corresponding mechanical arm operation according to the recognition result. The system comprises a voice module, a visual module and a mechanical arm control module. According to the invention, the quality of the voice signal is effectively improved through the spectrogram wavelet transform and adaptive threshold noise reduction technology, so that the voice instruction can be accurately recognized in a noisy working environment. And the accuracy and the intelligent level of voice recognition are improved.

Description

Technical Field

[0001] The present invention relates to the field of mechanical control, and in particular to a method, system, device and medium for voice-controlled mechanical arms. Background Art

[0002] In the field of intelligent robotic arm control, especially in high-noise environments such as power production sites, accurate voice control technology is crucial to improving operational efficiency and safety. In the prior art, the voice control method for household appliances described in the CN110827811A patent document proposes a method for noise reduction through wavelet transform and threshold processing. This method obtains voice signals, performs noise reduction processing, identifies control instructions, and controls household appliances to perform operations. Although it has a certain effect in indoor environments, its noise reduction effect and adaptability may be limited in complex and noisy environments such as power production sites. This is because the method is mainly aimed at relatively simple indoor noise environments and does not fully consider the special needs of outdoor high-noise environments.

[0003] On the other hand, the operating machinery voice interaction method in the patent document CN114400001A proposes a method for noise reduction processing based on operating environment noise sample data to improve the accuracy of voice recognition. This method takes into account the impact of operating environment noise by acquiring the target voice signal in the operating machinery cab, performing noise reduction processing based on the operating environment noise sample data, performing voice recognition, and controlling the target execution device of the operating machinery. However, it is mainly aimed at heavy machinery such as excavators and cranes, rather than specifically for the application of robotic arms in the power industry. In addition, this method may have limitations in dealing with multi-source noise and speech recognition in complex backgrounds, especially in power operation environments that require precise control. Therefore, in order to improve the accuracy and reliability of voice control of robotic arms, a method, system, device and medium for voice control of robotic arms are urgently needed. Summary of the invention

[0004] The present invention provides a method, system, device and medium for voice-controlled robotic arms, aiming to solve the problems of low recognition accuracy, inaccurate execution of control instructions and inability to effectively cope with complex noise interference when voice-controlled robotic arms are used in high-noise environments in the prior art.

[0005] To achieve the above purpose, the following technical solution is adopted.

[0006] A method for voice controlling a robotic arm, comprising:

[0007] Collect voice control command signals at power production sites;

[0008] Use spectral wavelet transform to decompose the voice control command signal in time and frequency domain to obtain spectral wavelet coefficients;

[0009] Estimate the level and characteristics of noise based on the energy distribution and frequency characteristics of the spectrum wavelet coefficients;

[0010] According to the noise estimation result, the optimal threshold value under different types and intensities of noise signals is adaptively selected through the pre-established threshold dictionary library, the spectrum wavelet coefficients are compared, the coefficients less than the optimal threshold are suppressed or set to zero, and the coefficients greater than the optimal threshold are adjusted to remove noise and enhance signal characteristics, and the processed spectrum wavelet coefficients are obtained;

[0011] Perform inverse transformation on the processed spectrogram wavelet coefficients to reconstruct the denoised speech signal;

[0012] Use the pre-trained improved recurrent neural network model that includes data enhancement, multi-head attention mechanism, and speech recognition model to convert the denoised speech signal into text or instructions;

[0013] The robot control system performs corresponding robot operations based on the recognition results.

[0014] Optionally, the steps of establishing a threshold dictionary library include:

[0015] In different power field environments, noise signals of different types and intensities are obtained, and according to the needs of robot arm control, the pure version of the command is synthesized with the noise signal to form a command data set;

[0016] Use spectrogram wavelet transform to decompose the speech signal of the instruction data set in time and frequency domain to obtain spectrogram wavelet coefficients;

[0017] Estimate the level and characteristics of noise based on the energy distribution and frequency characteristics of the spectrum wavelet coefficients;

[0018] According to the noise estimation results, multiple threshold function determination methods including hard threshold and soft threshold are selected to compare the wavelet coefficients of the spectrum, suppress or set the coefficients less than the threshold, and adjust the coefficients greater than the threshold to remove noise and enhance signal characteristics;

[0019] Perform inverse transformation on the processed spectrogram wavelet coefficients to reconstruct the denoised speech signal;

[0020] Combined with the pure version of the command, the signal-to-noise ratio (SNR), waveform similarity parameter (NCC) and mean square error (MSE) of the reconstructed signal are calculated, and the optimal threshold function determination method is selected through weighted calculation;

[0021] According to the optimal threshold function determination method, the spectrum wavelet threshold is calculated under different types and intensities of noise signals in the power environment, and a threshold dictionary library is established.

[0022] Optionally, during weighted calculation, the signal-to-noise ratio and waveform similarity parameter are positive weights, and the mean square error is a negative weight.

[0023] Optionally, the training steps of the improved recurrent neural network model including data augmentation, multi-head attention mechanism, and speech recognition model include:

[0024] Build an improved recurrent neural network model that includes data enhancement, multi-head attention mechanism, and speech recognition model;

[0025] Using the clean version of the command and the noise-reduced voice command, a training data set for training the improved recurrent neural network model is constructed;

[0026] Preprocessing the voice commands in the training data set, including removing the silent part and removing the modal particle part, to obtain the preprocessed voice commands;

[0027] Perform data enhancement on the preprocessed voice commands, including speed change, pitch change, and noise addition, to obtain an enhanced voice command data set;

[0028] Use multiple independent attention heads to calculate the attention weights of the enhanced voice command dataset, and concatenate or weighted sum the calculation results to obtain the output of the multi-head attention mechanism;

[0029] The output of the multi-head attention mechanism is used for feature extraction. According to the task requirements and the relevance of the features, the appropriate feature subset is selected to obtain the extracted feature set.

[0030] The extracted feature set is used to train the speech recognition model, and the speech signal after weighted summation of the multi-head attention mechanism is converted into text to obtain the control instructions of the robotic arm.

[0031] A system for controlling a robotic arm by voice, comprising a voice module, a visual module and a robotic arm control module; the voice module comprises a voice signal acquisition unit, a voice signal processing unit and a voice recognition unit;

[0032] The voice signal acquisition unit is used to acquire voice control command signals at the power production site;

[0033] The voice signal processing unit uses a spectrogram wavelet adaptive threshold noise reduction technology to perform noise reduction processing on the voice control command;

[0034] The speech recognition unit is used to convert the noise-reduced voice control instructions into text or instructions using a pre-trained improved recurrent neural network model including data enhancement, a multi-head attention mechanism, and a speech recognition model.

[0035] The visual module is used to collect the target image of the robot arm to obtain the target position information;

[0036] The robot arm control module is used to control the robot arm according to the noise-reduced voice command output by the voice recognition unit and the target position information obtained by the vision module.

[0037] Optionally, the activation method of the voice signal acquisition unit includes:

[0038] Periodic activation mode: The voice signal acquisition unit periodically acquires voice information and continuously performs audio acquisition;

[0039] Wake-up word activation mode: The voice signal acquisition unit starts to receive voice information after being awakened. When the user issues a specific wake-up word, the voice signal acquisition unit starts to obtain voice information;

[0040] Infrared induction activation method: The infrared sensor is used to detect whether there is a staff member approaching, which serves as a trigger mechanism for obtaining voice signals.

[0041] Optionally, the visual module includes:

[0042] A depth camera unit, used to collect target images of the robotic arm;

[0043] An image preprocessing unit is provided with an image preprocessing algorithm to preprocess the collected image, including noise reduction and enhancement;

[0044] Image recognition unit, which uses deep learning algorithms to extract and fuse features of the target object;

[0045] The coordinate acquisition unit acquires the three-dimensional coordinates of the target according to the target recognized by the image recognition unit, so as to guide the movement of the robot arm.

[0046] Optionally, the robotic arm control module includes:

[0047] A robot arm communication unit, used to establish communication between the robot arm and the voice module and the vision module;

[0048] The robot arm computing unit is used to calculate the movement distance and angle of each joint of the robot arm according to voice commands and kinematic algorithms, and to plan the optimal path based on obstacle information;

[0049] The robot drive unit is used to call the robot motion function library to drive the robot to complete various actions;

[0050] The gripper control unit is used to control the opening and closing of the gripper.

[0051] An electronic device comprises a processor and a memory, wherein the processor is used to execute a computer program stored in the memory to implement a method for controlling a robotic arm by voice.

[0052] A computer-readable storage medium stores at least one instruction, and when the at least one instruction is executed by a processor, a method for voice-controlled robotic arm is implemented.

[0053] Compared with the prior art, the present invention has the following beneficial effects:

[0054] The present invention effectively improves the quality of voice signals through spectral wavelet transform and adaptive threshold noise reduction technology, so that voice commands can be accurately recognized even in a noisy working environment. This not only improves the safety and accuracy of the robot arm operation, but also enhances the robustness and adaptability of the system.

[0055] The present invention establishes a threshold dictionary library, so that the system can adaptively select the optimal threshold according to the noise type and intensity in different power field environments, further improving the noise reduction effect and the accuracy of speech recognition. This method makes the system more flexible, can adapt to changing working environments, and improves the practicality and reliability of the system.

[0056] The present invention optimizes the threshold selection process by using a specific weighted calculation method, so that the signal-to-noise ratio and waveform similarity parameters are used as positive weights, and the mean square error is used as a negative weight. This optimization helps to more accurately evaluate and select the best threshold, thereby improving the quality of speech signal processing.

[0057] The present invention improves the accuracy and intelligence level of speech recognition through an improved recurrent neural network model, including data enhancement, multi-head attention mechanism and speech recognition model. This method enables the system to better understand and execute complex voice commands, enhancing the control ability and operational flexibility of the robotic arm.

[0058] The system of the present invention can combine voice commands and visual information to achieve more accurate robot arm control. This integrated method improves the automation level of the system, reduces dependence on manual operation, and improves work efficiency and operational accuracy.

[0059] The present invention provides a variety of activation modes for the voice signal acquisition unit, so that the system can flexibly select the most appropriate activation strategy according to actual needs. This not only improves the response speed and efficiency of the system, but also helps save energy and protect privacy.

[0060] The present invention describes the function of the visual module in detail, so that the system can more accurately identify the target and obtain its three-dimensional coordinates, thereby guiding the robot arm to perform precise operations. This visual control capability significantly improves the operating accuracy and work efficiency of the robot arm.

[0061] The present invention describes in detail the components of the robot control module, so that the system can achieve more complex robot control tasks. This modular design improves the flexibility and scalability of the system, so that the robot can perform more diverse operations. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] Figure 1 The present invention is a control flow diagram of a method for voice controlling a robotic arm.

[0063] Figure 2 This is a schematic diagram of an improved recurrent neural network data processing flow for a method for voice controlling a robotic arm of the present invention.

[0064] Figure 3 The present invention is a schematic diagram of the system structure of a voice-controlled robotic arm.

[0065] Figure 4 The present invention is a structural block diagram of an electronic device. DETAILED DESCRIPTION

[0066] The present invention will be described in detail below with reference to the accompanying drawings and in combination with embodiments. It should be noted that the embodiments and features in the embodiments of the present application can be combined with each other without conflict.

[0067] The following detailed description is an exemplary description, which is intended to provide further detailed description of the present invention. Unless otherwise specified, all technical terms used in the present invention have the same meaning as those generally understood by those skilled in the art to which the present application belongs. The terms used in the present invention are only for describing specific embodiments, and are not intended to limit exemplary embodiments according to the present invention.

[0068] Example 1

[0069] like Figure 1 As shown, a method for voice-controlled robotic arm, the method includes spectral wavelet adaptive threshold noise reduction technology

[0070] S11: In different power field environments, noise signals of different types and intensities are obtained, and according to the needs of robot arm control, the pure version of the command is synthesized with the noise signal to form a command data set.

[0071] S12: Preprocess the speech instruction data set and perform frame cutting to obtain short-time frame segmented speech signals, apply a windowing function to each frame, and perform spectrogram wavelet transform on each windowed frame.

[0072] S13: Decompose the speech signal into time-frequency domain representation using spectrogram wavelet transform to obtain spectrogram wavelet coefficients.

[0073]

[0074] ψ t，n (m) represents the value of the spectral wavelet at scale, position and time, t represents the scale, and g(*) is defined as the spectral wavelet operator corresponding to scale t. x 1 (m) is the value of the signal at n and m.

[0075] By analyzing the energy distribution and frequency characteristics of the signal, the level and characteristics of the noise can be estimated.

[0076] S14: According to the result of noise estimation, hard threshold, soft threshold and multiple threshold function determination methods are applied. The spectrum wavelet coefficients are compared with the determined threshold, and the coefficients less than the threshold are suppressed or set to zero to remove the influence of noise. At the same time, the coefficients greater than the threshold are appropriately adjusted to enhance the important features of the signal.

[0077] S15: Perform inverse transformation on the processed spectrogram wavelet coefficients to reconstruct the denoised speech signal.

[0078] S16: Combined with the pure version of the command, calculate the signal-to-noise ratio (SNR), waveform similarity parameter (NCC), and mean square error (MSE) of the reconstructed signal, and through weighted calculation, find the indicator results corresponding to different threshold determination methods, and select the optimal threshold determination method. In weighted calculation, the signal-to-noise ratio (SNR) and waveform similarity parameter (NCC) are positive weights, and the mean square error (MSE) is a negative weight.

[0079] The wavelet thresholds of the spectrum under different types and intensities of noise signals in the power environment are calculated respectively, and a threshold dictionary is formed. When in a certain power environment, the optimal threshold is adaptively selected by analyzing the type and intensity of the noise.

[0080] Example 2

[0081] like Figure 2 As shown, a method for voice-controlled robotic arm, the method comprising an improved recurrent neural network technology:

[0082] An improved recurrent neural network model including data enhancement, multi-head attention mechanism and speech recognition model is established, and a training data set is constructed based on the pure version of the instructions and the speech instructions after denoising by the spectrogram wavelet adaptive threshold.

[0083] The voice command training data set includes basic commands for controlling the robot arm and advanced commands for multiple directions or angles of motion. The improved recurrent neural network model is initially trained using the basic commands, and then the improved recurrent neural network model is retrained using the advanced commands.

[0084] The basic instructions include moving forward, moving backward, moving up, moving down, moving left, moving right, rotating clockwise, rotating counterclockwise, grabbing, placing, starting, stopping, resetting, etc. Advanced instructions include moving to the specified coordinates, grabbing the target object, etc.

[0085] The data enhancement model performs preprocessing and data enhancement on the input voice commands. Data preprocessing includes removing silent parts and modal particles to compress the size of voice commands and improve the quality of feature extraction. Data enhancement includes operations such as speed change, pitch change, and noise addition to expand the amount and diversity of data and improve the robustness and generalization ability of the model.

[0086] The multi-head attention mechanism uses multiple independent attention heads to calculate the attention weights separately, and concatenates or weighted sums their results to obtain a richer representation.

[0087] When applying the attention mechanism for feature extraction, the appropriate feature subset is selected according to the task requirements and the relevance of the features to reduce the dimension, reduce redundant information, and improve the efficiency and performance of the model.

[0088] The speech recognition model converts the speech signal after the weighted summation of the multi-head attention mechanism into text, and the robotic arm control system performs the specified operations of the robotic arm based on the recognition results.

[0089] Example 3

[0090] like Figure 3 As shown, a system for voice-controlled robotic arm, the system comprising:

[0091] Speech module: includes a speech signal acquisition unit, which is responsible for collecting speech control command signals at the power production site; a speech signal processing unit, which uses spectrogram wavelet adaptive threshold noise reduction technology to perform noise reduction on speech control commands; a speech recognition unit, including data enhancement, multi-head attention mechanism, and speech recognition model, which learns the characteristics of speech signals through training and is used to convert speech signals into text or commands.

[0092] The above-mentioned voice recognition is performed through a voice signal acquisition unit, and there are three main ways to obtain voice information. The first is that the voice signal acquisition unit periodically acquires voice information and continuously collects audio. The second is that the voice signal acquisition unit starts receiving voice information after being awakened. When the user issues a specific wake-up word, the voice signal acquisition unit starts to acquire voice information. The third is to use an infrared sensor to detect whether there is a staff member approaching as a trigger mechanism for obtaining voice signals.

[0093] Preferably, the second method of obtaining voice information is adopted, that is, the voice signal acquisition unit starts to obtain voice information after being awakened, which is a more appropriate choice. The voice signal acquisition unit does not need to work continuously, and only starts to receive voice information when it is awakened, which can save energy and extend the service life of the equipment. By using the wake-up word to trigger the voice signal acquisition unit to obtain voice information, useless processing and transmission can be reduced, and the response speed of the system can be improved. Since the voice signal acquisition unit only receives voice information when it is awakened, the potential risk of privacy leakage and misoperation can be reduced.

[0094] Vision module: includes a depth camera for collecting target images; an image preprocessing unit, which is equipped with an image preprocessing algorithm to preprocess the image, such as noise reduction and enhancement; an image recognition unit, which uses a deep learning algorithm to extract and fuse the features of the target object to achieve target recognition; a coordinate acquisition unit, which obtains the three-dimensional coordinates of the target based on the identified target to guide the movement of the robotic arm.

[0095] Robotic arm control module: controls the robotic arm according to the voice commands output by the voice recognition unit and the target position information obtained by the vision module.

[0096] Furthermore, the robot arm control module includes:

[0097] The robot arm communication unit establishes communication between the robot arm and the voice module and vision module;

[0098] The robot arm computing unit calculates the movement distance and angle of each joint of the robot arm according to voice commands and kinematic algorithms, and performs optimal path planning based on obstacle information;

[0099] The robot drive unit calls the robot motion function library to drive the robot to complete various actions;

[0100] The gripper control unit is used to control the opening and closing of the gripper.

[0101] The overall process of the system is as follows:

[0102] First, the voice signal acquisition unit collects voice commands, and the voice signal processing unit uses the spectral wavelet adaptive threshold noise reduction technology to reduce the noise of the voice control commands. The denoised voice signal is input into the voice recognition unit. Through data enhancement, multi-head attention mechanism, and voice recognition model, the voice recognition unit converts the voice signal into text commands and transmits it to the robot computing unit through the robot communication unit.

[0103] According to the instruction requirements, the depth camera in the visual module collects the scene image containing the target, and after preprocessing by the image preprocessing unit, the image recognition unit accurately identifies the target object, and the coordinate acquisition unit obtains the three-dimensional coordinates of the target. The robot arm computing unit obtains the three-dimensional coordinates of the target through the robot arm communication unit.

[0104] The robot arm computing unit calculates the movement distance and angle of each joint of the robot arm according to voice commands and kinematic algorithms, and plans the optimal path based on obstacle information. The robot arm driving unit drives the robot arm to move and turn, and the gripper control unit controls the opening and closing of the gripper, so that the robot arm moves to the target position and operates on the target.

[0105] The robot arm calculation unit calculates the movement distance and angle of each joint of the robot arm. The forward kinematics method of the robot arm can be used to solve the joint angle through the DH parameters and transformation matrix. First, the robot arm motion coordinate system is established to obtain the target position. The transformation matrix is ​​determined according to the DH parameters of the robot arm to solve the joint angle. After solving the joint angle, each joint angle is converted into radians. Whether to continue the operation is determined by judging whether the robot arm has reached the target position.

[0106] Example 4

[0107] like Figure 4 As shown, an electronic device includes a processor and a memory, wherein the processor is used to execute a computer program stored in the memory to implement a method for voice controlling a robotic arm.

[0108] The present invention also provides an electronic device 100 for implementing a method for voice-controlled robotic arm in the above embodiment; the electronic device 100 includes a memory 101, at least one processor 102, a computer program 103 stored in the memory 101 and executable on the at least one processor 102, and at least one communication bus 104. The memory 101 can be used to store the computer program 103, and the processor 102 implements the method steps of voice-controlled robotic arm in embodiment 1 by running or executing the computer program stored in the memory 101 and calling the data stored in the memory 101. The memory 101 can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, an application required for at least one function (such as a sound playback function, an image playback function, etc.), etc.; the data storage area can store data (such as audio data) created according to the use of the electronic device 100, etc. In addition, the memory 101 may include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, at least one disk storage device, a flash memory device, or other non-volatile solid-state storage devices. At least one processor 102 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. The processor 102 may be a microprocessor or the processor 102 may also be any conventional processor, etc. The processor 102 is the control center of the electronic device 100, and various interfaces and lines are used to connect various parts of the entire electronic device 100. The memory 101 in the electronic device 100 stores multiple instructions to implement a method for voice-controlled robotic arms.

[0109] Example 5

[0110] A computer-readable storage medium stores at least one instruction, and when the at least one instruction is executed by a processor, a method for voice-controlled robotic arm is implemented.

[0111] If the module / unit integrated in the electronic device 100 is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present invention implements all or part of the process in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium. When the computer program is executed by the processor, the steps of the above-mentioned various method embodiments can be implemented. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may include: any entity or device that can carry computer program code, recording medium, U disk, mobile hard disk, disk, optical disk, computer memory and read-only memory (ROM, Read-Only Memory). It should be understood by those skilled in the art that the embodiments of the present invention can be provided as a method, system, or computer program product. Therefore, the present invention can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code. The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 These computer program instructions can also be stored in a computer-readable memory that can guide a computer or other programmable data processing device to work in a specific way, so that the instructions stored in the computer-readable memory produce a product including an instruction device, which implements the functions specified in the process. Figure 1 A process or multiple processes and / or boxes Figure 1 These computer program instructions can also be loaded onto a computer or other programmable data processing device, so that a series of operation steps are executed on the computer or other programmable device to produce a computer-implemented process, so that the instructions executed on the computer or other programmable device provide for implementing the process in the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0112] It is known from common technical knowledge that the present invention can be implemented by other embodiments that do not deviate from its spirit or essential features. Therefore, the above disclosed embodiments are only illustrative in all respects and are not exclusive. All changes within the scope of the present invention or within the scope equivalent to the present invention are included in the present invention.

Claims

1. A method for controlling a robotic arm by voice, characterized in that: include, Collect voice control command signals at power production sites; Use spectral wavelet transform to decompose the voice control command signal in time and frequency domain to obtain spectral wavelet coefficients; Estimate the level and characteristics of noise based on the energy distribution and frequency characteristics of the spectrum wavelet coefficients; According to the noise estimation result, the optimal threshold value under different types and intensities of noise signals is adaptively selected through the pre-established threshold dictionary library, the spectrum wavelet coefficients are compared, the coefficients less than the optimal threshold are suppressed or set to zero, and the coefficients greater than the optimal threshold are adjusted to remove noise and enhance signal characteristics, and the processed spectrum wavelet coefficients are obtained; Perform inverse transformation on the processed spectrogram wavelet coefficients to reconstruct the denoised speech signal; Use the pre-trained improved recurrent neural network model that includes data enhancement, multi-head attention mechanism, and speech recognition model to convert the denoised speech signal into text or instructions; The robot control system performs corresponding robot operations based on the recognition results.

2. The method for voice controlling a robotic arm according to claim 1, characterized in that: The steps to establish the threshold dictionary library include: In different power field environments, noise signals of different types and intensities are obtained, and according to the needs of robot arm control, the pure version of the command is synthesized with the noise signal to form a command data set; Use spectrogram wavelet transform to decompose the speech signal of the instruction data set in time and frequency domain to obtain spectrogram wavelet coefficients; Estimate the level and characteristics of noise based on the energy distribution and frequency characteristics of the spectrum wavelet coefficients; According to the noise estimation results, multiple threshold function determination methods including hard threshold and soft threshold are selected to compare the wavelet coefficients of the spectrum, suppress or set the coefficients less than the threshold, and adjust the coefficients greater than the threshold to remove noise and enhance signal characteristics; Perform inverse transformation on the processed spectrogram wavelet coefficients to reconstruct the denoised speech signal; Combined with the pure version of the command, the signal-to-noise ratio, waveform similarity parameters and mean square error index of the reconstructed signal are calculated, and the optimal threshold function determination method is selected through weighted calculation; According to the optimal threshold function determination method, the spectrum wavelet threshold is calculated under different types and intensities of noise signals in the power environment, and a threshold dictionary library is established.

3. The method for voice controlling a robotic arm according to claim 2, characterized in that: In weighted calculation, the signal-to-noise ratio and waveform similarity parameter are positive weights, and the mean square error is a negative weight.

4. The method for voice controlling a robotic arm according to claim 1, characterized in that: The training steps of the improved recurrent neural network model including data enhancement, multi-head attention mechanism, and speech recognition model include: Build an improved recurrent neural network model that includes data enhancement, multi-head attention mechanism, and speech recognition model; Using the clean version of the command and the noise-reduced voice command, a training data set for training the improved recurrent neural network model is constructed; Preprocessing the voice commands in the training data set, including removing the silent part and removing the modal particle part, to obtain the preprocessed voice commands; Perform data enhancement on the preprocessed voice commands, including speed change, pitch change, and noise addition, to obtain an enhanced voice command data set; Use multiple independent attention heads to calculate the attention weights of the enhanced voice command dataset, and concatenate or weighted sum the calculation results to obtain the output of the multi-head attention mechanism; The output of the multi-head attention mechanism is used for feature extraction. According to the task requirements and the relevance of the features, the appropriate feature subset is selected to obtain the extracted feature set. The extracted feature set is used to train the speech recognition model, and the speech signal after weighted summation of the multi-head attention mechanism is converted into text to obtain the control instructions of the robotic arm.

5. A system for controlling a robotic arm by voice, based on a method for controlling a robotic arm by voice according to any one of claims 1 to 4, characterized in that: It includes a voice module, a visual module and a robotic arm control module; the voice module includes a voice signal acquisition unit, a voice signal processing unit and a voice recognition unit; The voice signal acquisition unit is used to acquire voice control command signals at the power production site; The voice signal processing unit uses a spectrogram wavelet adaptive threshold noise reduction technology to perform noise reduction processing on the voice control command; The speech recognition unit is used to convert the noise-reduced voice control instructions into text or instructions using a pre-trained improved recurrent neural network model including data enhancement, a multi-head attention mechanism, and a speech recognition model. The visual module is used to collect the target image of the robot arm to obtain the target position information; The robot arm control module is used to control the robot arm according to the noise-reduced voice command output by the voice recognition unit and the target position information obtained by the vision module.

6. The voice-controlled robotic arm system according to claim 5, characterized in that: The activation method of the voice signal acquisition unit includes: Periodic activation mode: The voice signal acquisition unit periodically acquires voice information and continuously performs audio acquisition; Wake-up word activation mode: The voice signal acquisition unit starts to receive voice information after being awakened. When the user issues a specific wake-up word, the voice signal acquisition unit starts to obtain voice information; Infrared induction activation method: The infrared sensor is used to detect whether there is a staff member approaching, which serves as a trigger mechanism for obtaining voice signals.

7. The system for controlling a robotic arm by voice according to claim 5, characterized in that: The visual module comprises: A depth camera unit, used to collect target images of the robotic arm; An image preprocessing unit is provided with an image preprocessing algorithm to preprocess the collected image, including noise reduction and enhancement; Image recognition unit, which uses deep learning algorithms to extract and fuse features of the target object; The coordinate acquisition unit acquires the three-dimensional coordinates of the target according to the target recognized by the image recognition unit, so as to guide the movement of the robot arm.

8. The system for controlling a robotic arm by voice according to claim 5, characterized in that: The robotic arm control module comprises: A robot arm communication unit, used to establish communication between the robot arm and the voice module and the vision module; The robot arm computing unit is used to calculate the movement distance and angle of each joint of the robot arm according to voice commands and kinematic algorithms, and to plan the optimal path based on obstacle information; The robot drive unit is used to call the robot motion function library to drive the robot to complete various actions; The gripper control unit is used to control the opening and closing of the gripper.

9. An electronic device, characterized in that: It comprises a processor and a memory, wherein the processor is used to execute a computer program stored in the memory to implement a method for voice-controlled robotic arm as claimed in any one of claims 1 to 4.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores at least one instruction, and when the at least one instruction is executed by the processor, a method for voice-controlled robotic arm according to any one of claims 1 to 4 is implemented.

Images