Rotor craft voice control method, system and device based on noisy environment

Abstract

The invention discloses a noisy environment-based rotorcraft voice control method, system and device, and the method comprises the steps: synchronously collecting voice and environmental noise through a dual-microphone array, processing a signal through a hybrid noise reduction algorithm, generating a text through a conversion module after the noise reduction, and storing the text through an intelligent helmet and wirelessly transmitting the text to a control module; and the instruction identification unit extracts keywords and matches preset parameters, and finally adjusts the rotating speed of the motor to realize accurate flight control. The system integrates a headset, an intelligent helmet and a control module, and the noise suppression capability is enhanced through the improved sound source positioning and adaptive filtering technology. According to the scheme, the speech recognition robustness in a complex noise environment is remarkably improved, and the method is suitable for the fields of unmanned aerial vehicle inspection, rescue and the like.

Description

Technical Field

[0001] The present invention belongs to the technical field of aircraft control, and particularly relates to a voice control method, system and device for a rotorcraft in a noisy environment. Background Art

[0002] At present, aircraft are widely used in various fields. The traditional flight control of aircraft mainly operates through a remote controller or a mobile control terminal, which has the problems of complex operation and high difficulty in getting started, and has certain requirements for novices. Therefore, aircraft voice control technology has been applied.

[0003] During the flight of an aircraft, due to the influence of various factors such as air flow and propeller rotation, a large amount of noise will be generated. These noises will interfere with the recognition of the operator's voice control commands, affecting flight safety and operation efficiency. Therefore, how to reduce the noise of the aircraft when the operator controls the aircraft by voice and improve the accuracy and clarity of voice recognition is an urgent problem to be solved. Summary of the Invention

[0004] The purpose of the present invention is to provide a voice control method, system and device for a rotorcraft in a noisy environment, which can quickly analyze and generate an anti-phase noise reduction wave through an active noise reduction unit, significantly improving the noise reduction effect.

[0005] To achieve the purpose of the present invention, on the one hand, the present invention provides a voice control method for a rotorcraft in a noisy environment, including the following steps:

[0006] S1: Start the voice control system of the rotorcraft, and synchronously obtain the original voice wave of the operator's voice signal and environmental noise through a dual microphone array;

[0007] S2: Process the signal using a hybrid noise reduction algorithm, and the hybrid noise reduction algorithm includes a dynamic noise modeling algorithm based on the spectral characteristics of environmental noise, a voice enhancement process of a deep voice enhancement network, and an adaptive feedback suppression algorithm;

[0008] The dynamic noise modeling algorithm uses an improved harmonic noise decomposition algorithm to establish a parametric model for the periodic characteristics of rotor noise:

[0009] ;

[0010] Wherein, is the amplitude of the synthesized noise signal at time t, representing the instantaneous sound pressure level of the mixed noise such as aerodynamic noise and mechanical vibration noise generated by the rotor system; K is the number of harmonic components of the noise decomposition, where is the fundamental frequency, ; is the amplitude component of the k-th harmonic at time t, representing the change in noise intensity caused by the rotor speed fluctuation; is the fundamental frequency component of the k-th harmonic, meanwhile, a dynamic compensation mechanism is introduced; is the non-linear phase perturbation of the k-th harmonic, mainly from the frequency modulation caused by the blade vortex shedding effect and the air flow perturbation; t is the time variable, processed by a sliding window, and the window length is adapted to the speech frame length;

[0011] The deep voice enhancement network has a structure of a time-frequency domain processing network based on U-Net, including an encoder with 5 layers of convolutional downsampling, a bottleneck layer for bidirectional LSTM time series modeling, and a decoder with transposed convolutional upsampling;

[0012] The adaptive feedback suppression algorithm establishes a second-order closed-loop control system, and the control system parameters are dynamically updated according to the real-time sound field measurement to ensure the system stability margin. It uses a digital filter transfer function:

[0013] ;

[0014] where, is the forward path gain, representing the phase compensation intensity for controlling the current sampling moment; is the delay compensation term, representing the residual feedback of the previous sampling point; is the recursive coefficient 1, representing suppressing the resonance peak of a specific frequency band; is the recursive coefficient 2, representing the damping characteristic of the control system to prevent oscillation divergence.

[0015] S3: The voice conversion unit converts the noise-reduced voice sound wave into text information and outputs it;

[0016] S4: The information storage unit in the intelligent helmet temporarily stores the information;

[0017] S5: The information sending unit in the intelligent helmet sends the text information to the control system through a wireless communication protocol;

[0018] S6: The instruction recognition unit in the control system receives the text information and judges whether there is a control instruction keyword in the text information according to the control instruction database. If so, it extracts the control instruction keyword and clarifies the instruction parameters; when there is no clear instruction parameter, the control instruction database will give preset initial parameters;

[0019] S7: The flight state adjustment unit will execute the control instruction with parameters, and change the flight state of the rotorcraft by adjusting the motor speed of the rotorcraft.

[0020] On the other hand, the present invention provides a system for a voice control method of a rotary-wing aircraft in a noisy environment, including the following modules:

[0021] A voice noise reduction module, configured to receive the original voice sound wave and perform active noise reduction processing, and then convert it into text information for output;

[0022] A control module, configured to receive the text information and perform judgment and extraction of control instructions, and execute the control instructions to achieve voice control of the rotary-wing aircraft.

[0023] The voice noise reduction module includes:

[0024] A sound collection unit, configured to collect the original voice sound wave of the instructions issued by the operator and the ambient noise;

[0025] An active noise reduction unit, which performs sound source localization through an improved Fourier beamforming algorithm, separates voice components by applying a time-frequency masking neural network model, adopts a sliding window adaptive filtering technology to update the noise feature library in real time, and generates an anti-phase cancellation sound wave with dynamic phase compensation;

[0026] The sound collection unit and the active noise reduction unit are integrated into a head-mounted earphone;

[0027] A voice conversion unit, configured to convert the voice sound wave after noise reduction processing into text information for output and send it to the control module;

[0028] An intelligent helmet, configured to connect the head-mounted earphone and the voice conversion unit.

[0029] The active noise reduction unit includes:

[0030] An audio processing module, configured to obtain the original voice sound wave collected by the head-mounted earphone;

[0031] The audio processing module includes a dynamic noise modeling sub-module and a deep voice enhancement network sub-module; the dynamic noise modeling sub-module is configured to provide key prior knowledge for subsequent active noise reduction by accurately analyzing the time-varying characteristics of the noise; the deep voice enhancement network sub-module is configured to perform voice enhancement processing;

[0032] An anti-phase sound wave playback module, configured to play the anti-phase noise reduction sound wave; the head-mounted earphone and the anti-phase sound wave playback module are close to the mouth and nose of the sound emitter, and are separated by a sound insulation layer therebetween to prevent the anti-phase noise reduction sound wave from entering the sound transmission microphone;

[0033] The anti-phase sound wave playback module includes an adaptive feedback suppression sub-module, and the adaptive feedback suppression sub-module is configured to establish a second-order closed-loop control system, and the control system is dynamically updated according to real-time sound field measurement to ensure its stability margin.

[0034] The intelligent helmet includes:

[0035] An information storage module, which is used to temporarily store the information converted into control instructions;

[0036] An information sending module, which is used to send the stored control instructions to the control system through a wireless communication protocol;

[0037] The control instructions are sent and received between the intelligent helmet and the control module through a wireless communication protocol.

[0038] The control module includes:

[0039] An instruction recognition unit, which is used to receive text information and determine whether there are control instruction keywords in the text information according to the control instruction database. If there are, it extracts the control instruction keywords and clarifies the instruction parameters; if there are clear instruction parameters, it directly transmits the extracted control instructions to the flight state adjustment unit; if there are no clear instruction parameters, it transmits the extracted control instructions to the control instruction database for keyword matching recognition to generate initial instruction parameters, and then transmits them to the flight state adjustment unit;

[0040] The control instruction database gives default initial parameters to the control instructions without clear instruction parameters;

[0041] A flight state adjustment unit, which is used to execute control instructions and change the flight state of the rotary-wing aircraft by adjusting the motor speed of the rotary-wing aircraft.

[0042] The initial parameters include:

[0043] Flight altitude, the relative altitude with respect to the initial position of the rotary-wing aircraft when receiving the control instruction;

[0044] Flight speed, the relative speed with respect to the initial speed of the rotary-wing aircraft when receiving the control instruction;

[0045] Flight distance, the relative distance with respect to the initial position of the rotary-wing aircraft when receiving the control instruction;

[0046] Flight direction, the relative direction with respect to the initial position of the rotary-wing aircraft when receiving the control instruction;

[0047] Takeoff, climb from the ground to a certain initial flight altitude at a certain initial speed;

[0048] Return, fly from the current position to the initial position at takeoff.

[0049] The above-mentioned head-mounted earphone uses a bone conduction microphone and a common microphone in combination; the processing module built in the head-mounted earphone uses machine learning to compare, analyze, and deeply learn the sound data collected by both, establish a data processing model for a specific user, and use this data processing model to supplement and restore the sound data obtained from the bone conduction microphone and the common microphone.

[0050] The present invention also provides a device for a voice control method of a rotary-wing aircraft in a noisy environment, including a head-mounted earphone, an intelligent helmet, and a control module; the head-mounted earphone is used to integrate a sound collection unit and an active noise reduction unit, the active noise reduction unit is connected to the intelligent helmet, the intelligent helmet includes a voice conversion unit, and can temporarily store the converted text information and send the text information to the control module; the control module and the intelligent helmet send and receive text information and gesture images through a wireless communication protocol.

[0051] Compared with the prior art, the significant progress of the present invention lies in: (1) The present invention applies the principle of inverse active noise reduction to the noise reduction of the voice commands of aircraft operators, improving the robustness; (2) The noise contained in the control voice commands issued by the operator of the present invention is mainly the noise of the propeller, which has the characteristic of periodicity, facilitating the active noise reduction unit to quickly analyze and generate an inverse noise reduction sound wave; (3) The voice sound wave signal after noise reduction processing of the present invention is clearer, improving the accuracy of control command recognition, providing a strong guarantee for operation efficiency and flight safety; (4) The present invention introduces a head-mounted earphone, which can cooperate with the active noise reduction unit to significantly improve the noise reduction effect; (5) The establishment of the control command database of the present invention shortens the thinking time for the operator to issue control commands, improving the operation efficiency.

[0052] To more clearly illustrate the functional characteristics and structural parameters of the present invention, the following further explains in combination with the drawings and specific embodiments. Description of the Drawings

[0053] The drawings described herein are used to provide a further understanding of the present invention, form a part of this application, and the schematic embodiments and descriptions of the present invention are used to explain the present invention and do not constitute an improper limitation to the present invention. In the drawings:

[0054] Figure 1 is the flowchart of the method steps of the present invention;

[0055] Figure 2 is the system structure diagram of the present invention;

[0056] Figure 3 is the structure diagram of the active noise reduction unit of the present invention;

[0057] Figure 4This is the schematic diagram of the anti-noise principle of the present invention. Detailed implementation manners

[0058] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments; based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0059] Combined with Figure 1 , the present invention provides a voice control method for a rotorcraft in a noisy environment, including the following steps:

[0060] S1: Start the voice control system of the rotorcraft, and synchronously acquire the original voice sound waves of the operator's voice signal and environmental noise through a dual microphone array;

[0061] S2: Process the signal using a hybrid noise reduction algorithm, and the hybrid noise reduction algorithm includes a dynamic noise modeling algorithm based on the spectral characteristics of environmental noise, a voice enhancement process of a deep voice enhancement network, and an adaptive feedback suppression algorithm;

[0062] For the dynamic noise modeling algorithm, an improved harmonic noise decomposition algorithm is adopted to establish a parametric model for the periodic characteristics of rotor noise:

[0063] ;

[0064] Among them, is the amplitude of the synthesized noise signal at time t, representing the instantaneous sound pressure level of the mixed noise such as aerodynamic noise and mechanical vibration noise generated by the rotor system; K is the number of harmonic components of the noise decomposition, , where is the fundamental frequency, ; is the amplitude component of the k-th harmonic at time t, representing the change in noise intensity caused by the rotor speed fluctuation; is the fundamental frequency component of the k-th harmonic, , and at the same time, a dynamic compensation mechanism is introduced; is the non-linear phase perturbation of the k-th harmonic, mainly from the frequency modulation caused by the blade vortex shedding effect and air flow disturbance; t is the time variable, and a sliding window is used for processing, and the window length is adapted to the voice frame length;

[0065] For the deep voice enhancement network, its structure is a time-frequency domain processing network based on U-Net, including an encoder with 5 layers of convolutional downsampling, a bottleneck layer for bidirectional LSTM time series modeling, and a decoder with transposed convolutional upsampling;

[0066] The adaptive feedback suppression algorithm establishes a second-order closed-loop control system, and the control system parameters are dynamically updated according to real-time sound field measurements to ensure the system stability margin. It uses a digital filter transfer function:

[0067] ;

[0068] where, is the forward path gain, representing the phase compensation intensity at the current sampling moment; is the delay compensation term, representing the residual feedback of the previous sampling point; is the recursion coefficient 1, representing the suppression of resonance peaks in a specific frequency band; is the recursion coefficient 2, representing the damping characteristic of the control system to prevent oscillation divergence.

[0069] S3: The voice conversion unit converts the noise-reduced voice sound wave into text information and outputs it;

[0070] S4: The information storage unit in the intelligent helmet temporarily stores the information;

[0071] S5: The information sending unit in the intelligent helmet sends the text information to the control system through a wireless communication protocol;

[0072] S6: The instruction recognition unit in the control system receives the text information and judges whether there is a control instruction keyword in the text information according to the control instruction database. If so, it extracts the control instruction keyword and clarifies the instruction parameters; when there are no clear instruction parameters, the control instruction database will give preset initial parameters;

[0073] S7: The flight state adjustment unit will execute the control instruction with parameters to change the flight state of the rotorcraft by adjusting the motor speed of the rotorcraft.

[0074] Combined with Figure 2 , the present invention provides a system for a rotorcraft voice control method based on a noisy environment, including the following modules:

[0075] A voice noise reduction module, configured to receive the original voice sound wave and perform active noise reduction processing, and then convert it into text information for output;

[0076] A control module, configured to receive the text information and perform judgment and extraction of control instructions, and execute the control instructions to achieve voice control of the rotorcraft.

[0077] The voice noise reduction module includes:

[0078] A sound collection unit, configured to collect the original voice sound wave of the operator's instruction and environmental noise;

[0079] The active noise reduction unit locates the sound source through an improved Fourier beamforming algorithm, separates the speech components using a time-frequency masking neural network model, and updates the noise feature library in real time using a sliding window adaptive filtering technique to generate an inverted cancellation sound wave with dynamic phase compensation;

[0080] The sound collection unit and the active noise reduction unit are integrated into a head-mounted earphone;

[0081] The speech conversion unit is used to convert the noise-reduced voice sound wave into text information and output it, and send it to the control module;

[0082] The intelligent helmet is used to connect the head-mounted earphone and the speech conversion unit.

[0083] Combined Figure 3 , the active noise reduction unit includes:

[0084] The audio processing module is used to obtain the original voice sound wave collected by the head-mounted earphone;

[0085] The audio processing module includes a dynamic noise modeling sub-module and a deep voice enhancement network sub-module; the dynamic noise modeling sub-module is used to provide key prior knowledge for subsequent active noise reduction by accurately analyzing the time-varying characteristics of the noise; the deep voice enhancement network sub-module is used for voice enhancement processing;

[0086] The inverted sound wave playback module is used to play the inverted noise reduction sound wave; the head-mounted earphone and the inverted sound wave playback module are close to the mouth and nose of the speaker, and the two are separated by a sound insulation layer to prevent the inverted noise reduction sound wave from entering the sound transmission microphone;

[0087] The inverted sound wave playback module includes an adaptive feedback suppression sub-module, and the adaptive feedback suppression sub-module is used to establish a second-order closed-loop control system, and the control system is dynamically updated according to real-time sound field measurement to ensure its stability margin.

[0088] Combined Figure 4 , the principle of reverse sound cancellation, that is, the principle of active noise reduction, works based on the original sound waveform, the reverse sound waveform, and the noise reduction sound waveform. The original sound waveform represents the noise in the environment, which is irregular and without pattern, composed of different frequency components, and is transmitted to the human ear through the vibration of air molecules and perceived as noise; the active noise reduction unit analyzes the characteristics such as the frequency, amplitude, and phase of the original noise, generates a reverse sound waveform with a completely opposite phase, and the reverse sound waveform and the original sound waveform meet and overlap in space. According to the principle of wave interference, in an ideal state, the wave peaks and wave troughs of the two cancel each other out, forming a noise reduction sound waveform, so that the amplitude of the synthesized waveform decreases or even approaches zero.

[0089] The intelligent helmet includes:

[0090] An information storage module for temporarily storing information converted into control instructions;

[0091] An information sending module for sending the stored control instructions to the control system via a wireless communication protocol;

[0092] Control instructions are sent and received between the intelligent helmet and the control module via a wireless communication protocol.

[0093] The control module includes:

[0094] An instruction recognition unit for receiving text information and judging whether there are control instruction keywords in the text information according to the control instruction database. If there are, extract the control instruction keywords and clarify the instruction parameters; if there are clear instruction parameters, directly transmit the extracted control instructions to the flight state adjustment unit; if there are no clear instruction parameters, transmit the extracted control instructions to the control instruction database for keyword matching recognition to generate initial instruction parameters, and then transmit them to the flight state adjustment unit;

[0095] The control instruction database assigns default initial parameters to control instructions without clear instruction parameters;

[0096] A flight state adjustment unit for executing control instructions and changing the flight state of the rotorcraft by adjusting the motor speed of the rotorcraft.

[0097] The initial parameters include:

[0098] Flight altitude, the relative altitude with respect to the initial position of the rotorcraft when receiving the control instruction;

[0099] Flight speed, the relative speed with respect to the initial speed of the rotorcraft when receiving the control instruction;

[0100] Flight distance, the relative distance with respect to the initial position of the rotorcraft when receiving the control instruction;

[0101] Flight direction, the relative direction with respect to the initial position of the rotorcraft when receiving the control instruction;

[0102] Takeoff, climbing from the ground to a certain initial flight altitude at a certain initial speed;

[0103] Return flight, flying from the current position to the initial position at takeoff.

[0104] In the form of lift height - 2 meters; lift speed - 2 m / s.

[0105] The head-mounted headset uses both a bone conduction microphone and a regular microphone in combination; the processing module built into the head-mounted headset uses machine learning to compare, analyze, and perform in-depth learning on the sound data collected by both, establishing a data processing model for a specific user, and using this data processing model to supplement and restore the sound data obtained from the bone conduction microphone and the regular microphone.

[0106] The bone conduction microphone is not easily affected by reverse sound waves, but has low sound fidelity; the regular microphone has high sound fidelity, but is easily affected by reverse sound waves. Using them in combination results in the best noise reduction effect.

[0107] After the intelligent helmet is powered on, it enters the sleep standby state; the intelligent helmet includes two methods to enter the working mode:

[0108] Mechanical button method: A mechanical switch button is set on the right side of the helmet. In the sleep standby state, when the button is clicked once, the voice noise reduction system enters the working mode; when the button is clicked continuously twice, the voice noise reduction system enters the sleep standby state;

[0109] Voice wake-up method: In the sleep standby state, when the aircraft operator says a specific wake-up word, the voice noise reduction system will be awakened and switched to the working state to wait for the operator's next instruction; in the working state, when the aircraft operator says a specific sleep word, the voice noise reduction system will switch to the sleep standby state.

[0110] The present invention also provides a device for a voice control method of a rotary-wing aircraft in a noisy environment, including a head-mounted headset, an intelligent helmet, and a control module; the head-mounted headset is used to integrate a sound collection unit and an active noise reduction unit, the active noise reduction unit is connected to the intelligent helmet, the intelligent helmet includes a voice conversion unit, and can temporarily store the converted text information and send the text information to the control module; the control module and the intelligent helmet send and receive text information and gesture images through a wireless communication protocol.

[0111] It should be noted that in this article, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variant thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements not only includes those elements, but also includes other elements not expressly listed, or also includes elements inherent to such process, method, article or device.

[0112] Although embodiments of the present invention have been shown and described, it will be understood by those of ordinary skill in the art that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the present invention, and the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for voice control of a rotorcraft in a noisy environment, characterized in that: The following steps are involved: S1: Start the voice control system of the rotorcraft and synchronously obtain the original voice sound waves of the operator's voice signal and environmental noise through the dual microphone array; S2: using a hybrid noise reduction algorithm to process the signal, the hybrid noise reduction algorithm includes a dynamic noise modeling algorithm based on the spectral characteristics of environmental noise, a speech enhancement processing of a deep speech enhancement network, and an adaptive feedback suppression algorithm; S3: The speech conversion unit converts the noise-reduced speech sound waves into text information and outputs it; S4: The information storage unit in the smart helmet temporarily stores the information; S5: The information sending unit in the smart helmet sends the text information to the control system through the wireless communication protocol; S6: The instruction recognition unit in the control system receives the text information, and determines whether there are control instruction keywords in the text information according to the control instruction database. If there are control instruction keywords, the control instruction keywords are extracted and instruction parameters are clarified. If there are no clear instruction parameters, the control instruction database will give preset initial parameters. S7: The flight state adjustment unit will execute the control instruction containing the parameters, and change the flight state of the rotorcraft by adjusting the rotation speed of the rotorcraft motor.

2. The method for voice control of a rotorcraft in a noisy environment according to claim 1, characterized in that: The dynamic noise modeling algorithm, deep speech enhancement network, and adaptive feedback suppression algorithm in step S2: The dynamic noise modeling algorithm adopts an improved harmonic noise decomposition algorithm to establish a parameterized model for the periodic characteristics of rotor noise: ； in, is the amplitude of the synthetic noise signal at time t, representing the instantaneous sound pressure level of the mixed noise such as aerodynamic noise and mechanical vibration noise generated by the rotor system; K is the number of harmonic components of the noise decomposition, ,in is the fundamental frequency, ; is the amplitude component of the kth harmonic at time t, representing the change in noise intensity caused by the rotor speed fluctuation; is the fundamental frequency component of the kth order harmonic, , and introduce dynamic compensation mechanism; is the nonlinear phase disturbance of the kth order harmonic, which mainly comes from the frequency modulation caused by the blade vortex shedding effect and airflow disturbance; t is the time variable, which is processed by sliding window, and the window length is adapted to the speech frame length; The deep speech enhancement network is a time-frequency domain processing network based on U-Net, including a 5-layer convolution downsampling encoder, a bottleneck layer of bidirectional LSTM time series modeling, and a transposed convolution upsampling decoder; The adaptive feedback suppression algorithm establishes a second-order closed-loop control system. The control system data is dynamically updated according to the real-time sound field measurement to ensure the system stability margin. It adopts the digital filter transfer function: ； in, is the forward path gain, which represents the phase compensation strength controlling the current sampling moment; is the delay compensation term, representing the residual feedback of the previous sampling point; is the recursive coefficient 1, which means suppressing the resonance peak in a specific frequency band; is the recursive coefficient 2, which represents the damping characteristics of the control system to prevent oscillation divergence.

3. A system based on a rotorcraft voice control method in a noisy environment according to any one of claims 1-2, characterized in that: Includes the following modules: A speech noise reduction module is used to receive the original speech sound waves and perform active noise reduction processing, and then convert them into text information output; The control module is used to receive text information and judge and extract control instructions, and execute control instructions to realize voice control of the rotorcraft.

4. The voice control system for a rotorcraft in a noisy environment according to claim 3, characterized in that: The speech noise reduction module comprises: A sound collection unit, used to collect the original voice sound waves of the operator's instructions and environmental noise; The active noise reduction unit uses an improved Fourier beam forming algorithm to locate the sound source, applies a time-frequency masking neural network model to separate speech components, uses a sliding window adaptive filtering technique to update the noise feature library in real time, and generates an anti-phase anechoic wave with dynamic phase compensation; The sound collection unit and the active noise reduction unit are integrated into a head-mounted headset; A speech conversion unit, used to convert the speech sound waves after noise reduction processing into text information and output it to the control module; A smart helmet, used to connect a headset and a voice conversion unit.

5. A voice control system for a rotorcraft in a noisy environment according to claim 4, characterized in that: The active noise reduction unit comprises: An audio processing module, used for acquiring original voice sound waves collected by the headset; The audio processing module includes a dynamic noise modeling submodule and a deep speech enhancement network submodule; the dynamic noise modeling submodule is used to provide key prior knowledge for subsequent active noise reduction by accurately analyzing the time-varying characteristics of noise; the deep speech enhancement network submodule is used for speech enhancement processing; The reverse phase sound wave playing module is used to play the reverse phase noise reduction wave; the headset and the reverse phase sound wave playing module are close to the mouth and nose of the speaker, and a sound insulation layer is used to block the reverse phase noise reduction wave from entering the microphone; The anti-phase sound wave playing module includes an adaptive feedback suppression submodule, which is used to establish a second-order closed-loop control system. The control system is dynamically updated according to real-time sound field measurement to ensure its stability margin.

6. The voice control system for a rotorcraft in a noisy environment according to claim 4, characterized in that: The smart helmet comprises: An information storage module, used for temporarily storing information converted into control instructions; An information sending module, used for sending the stored control instructions to the control system via a wireless communication protocol; The smart helmet and the control module send and receive control instructions via a wireless communication protocol.

7. The voice control system for a rotorcraft in a noisy environment according to claim 3, characterized in that: The control module comprises: The instruction recognition unit is used to receive text information and determine whether there are control instruction keywords in the text information according to the control instruction database. If there are, the control instruction keywords are extracted and instruction parameters are clarified; if there are clear instruction parameters, the extracted control instruction is directly transmitted to the flight state adjustment unit; if there are no clear instruction parameters, the extracted control instruction is transmitted to the control instruction database, keyword matching and identification are performed to generate initial instruction parameters, and then transmitted to the flight state adjustment unit; The control instruction database gives default initial parameters to control instructions without explicit instruction parameters; The flight state adjustment unit is used to execute control instructions and change the flight state of the rotorcraft by adjusting the rotation speed of the rotorcraft motor.

8. The voice control system for a rotorcraft in a noisy environment according to claim 7, characterized in that: The initial parameters include: Flight altitude, relative to the initial position of the rotorcraft when it receives the control command; Flight speed, relative to the initial speed of the rotorcraft when it receives the control command; Flight distance, relative to the initial position of the rotorcraft when it receives the control command; Flight direction, relative to the initial position of the rotorcraft when it receives the control command; Take off, climb from the ground at a certain initial speed to a certain initial flight altitude; Return, fly from the current position to the initial position when taking off.

9. The voice control system for a rotorcraft in a noisy environment according to claim 4, characterized in that: The headset uses a bone conduction microphone and an ordinary microphone in combination; the built-in processing module of the headset uses machine learning to compare and analyze the sound data collected by the two, and deep learning to establish a data processing model for a specific user. The data processing model is used to supplement and restore the sound data obtained from the bone conduction microphone and the ordinary sound transmission microphone.

10. A device for a rotorcraft voice control method in a noisy environment according to any one of claims 1-2, characterized in that: It comprises a head-mounted headset, a smart helmet, and a control module; the head-mounted headset is used to integrate a sound collection unit and an active noise reduction unit, the active noise reduction unit is connected to the smart helmet, the smart helmet comprises a voice conversion unit, and can temporarily store converted text information and send text information to the control module; the control module and the smart helmet send and receive text information and gesture images through a wireless communication protocol.

Images