An outdoor interview practical training device based on environment self-adaptation

Abstract

The application relates to the technical field of intelligent media practical training equipment, in particular to an outdoor interview practical training device based on environment self-adaptation, which comprises a main control unit, a multi-modal environment sensing module, an audio and video acquisition module and a self-adaptive execution mechanism; the multi-modal environment sensing module comprises an illumination intensity sensor, a wind speed sensor, a noise sensor, a three-axis tilt sensor and a temperature and humidity sensor; the self-adaptive execution mechanism comprises a gravity adjusting mechanism, a camera posture adjusting mechanism, a microphone directivity adjusting mechanism, a light supplement intensity adjusting mechanism and a terrain self-adaptive support foot assembly. The application can realize real-time closed-loop regulation and control of the equipment posture, the sound pickup direction and the light supplement parameters based on the environment parameters, and realizes stable and high-quality audio and video acquisition in a complex outdoor scene.

Description

Technical Field

[0001] This invention belongs to the field of news communication and media education technology, specifically an outdoor interview training device based on environmental adaptation. Background Technology

[0002] With the development of journalism and media education, outdoor interviewing training, as an important part of journalism education, has placed higher demands on the environmental adaptability, ease of operation, and completeness of information collection of equipment. Existing outdoor interviewing equipment often focuses on optimizing single functions, such as microphone wind protection, camera lighting, or equipment portability. However, when facing complex and changing outdoor environments (such as strong light, wind, rain, noise, and uneven terrain), it lacks the ability to comprehensively perceive and adaptively adjust to multiple environmental factors, making it difficult to meet the needs of high-quality and high-efficiency training.

[0003] A search revealed a combined outdoor news gathering device, publication number CN114857459B, published on August 9, 2024. This patent utilizes a housing with a height-adjustable and rotatable camera and audio acquisition unit, employing a rack and pinion mechanism and motor drive to achieve multi-degree-of-freedom adjustment, thereby enhancing the flexibility of information acquisition. However, this device relies solely on preset programs or manual remote control for angle and height adjustments, lacking an integrated environmental sensing module (such as sensors for light intensity, wind speed, and noise levels). It cannot automatically optimize acquisition parameters based on actual environmental changes, requiring manual intervention in situations such as sudden severe weather or strong noise interference, thus reducing the autonomy and real-time response capability of the training process.

[0004] A search revealed a combined news interview device with publication number CN116772063A, published on September 19, 2023. This patent employs a modular design, including a height-adjustable worktable and rotating audio-visual components, and is equipped with a solar power system, making it suitable for outdoor interviews in low-light environments. However, while the device possesses a certain degree of structural adaptability, its adjustment mechanism remains mechanical, manual, or semi-automatic, lacking intelligent decision-making capabilities based on environmental data. For example, it cannot automatically lower the device's center of gravity to enhance stability in strong winds, or it cannot adjust microphone directivity and gain parameters in conjunction with high-noise environments, making it difficult to achieve dynamic coordination among the environment, device, and content.

[0005] The aforementioned problems indicate that existing outdoor interviewing devices have significant shortcomings in environmental perception, intelligent decision-making, and adaptive execution, failing to effectively support the demands for intelligence, robustness, and full-process autonomy in modern news interviewing training. Therefore, this invention proposes an environmentally adaptive outdoor interviewing training device. By integrating multimodal environmental sensors, edge computing units, and adaptive execution mechanisms, it aims to achieve real-time perception of external conditions such as light, wind, noise, and terrain, and automatic optimization of equipment parameters, thereby improving the quality, safety, and teaching efficiency of outdoor interviewing training. Summary of the Invention

[0006] To address the problems in existing technologies, this invention provides an environmentally adaptive outdoor interview training device. It solves the technical shortcomings of existing outdoor interview devices in complex and changing environments (including strong light, wind, rain, noise, and uneven terrain), such as lack of comprehensive perception of multiple environmental factors, inability to automatically optimize and adjust equipment parameters, reliance on manual intervention for angle and height adjustments, lack of intelligent decision-making mechanisms based on environmental data, and difficulty in achieving dynamic coordination among "environment, equipment, and content." This invention aims to improve the autonomy, robustness, and completeness of information collection during outdoor interview training by constructing an integrated structural system that combines multimodal environmental sensors, edge computing units, and adaptive actuators. This system enables real-time acquisition and localized processing of external environmental parameters such as light intensity, wind speed, noise level, and ground tilt, as well as closed-loop control of equipment action commands.

[0007] The technical solution adopted by this invention to solve the above-mentioned technical problems is: an outdoor interview training device based on environmental adaptation, comprising a main control unit, a multimodal environmental perception module, an audio and video acquisition module, an adaptive actuator, a power supply module, and a communication module. The main control unit is electrically connected to the multimodal environmental perception module, the audio and video acquisition module, the adaptive actuator, the power supply module, and the communication module respectively. The multimodal environmental perception module includes a light intensity sensor, a wind speed sensor, a noise sensor, a three-axis tilt sensor, and a temperature and humidity sensor. The audio and video acquisition module includes a liftable camera, a rotatable microphone array, and a supplementary lighting group. The adaptive actuator includes a center of gravity adjustment mechanism, a camera posture adjustment mechanism, a microphone directivity adjustment mechanism, and a supplementary lighting intensity adjustment mechanism.

[0008] The main control unit includes an edge computing processor, non-volatile memory, a real-time operating system module, and an instruction scheduler. The edge computing processor is connected to the non-volatile memory via a PCIe bus to store a preset environment-device mapping rule table and historical operating data. The real-time operating system module is embedded inside the edge computing processor to schedule multi-task parallel processing. The instruction scheduler is connected to the edge computing processor via an internal bus to convert the processing results into specific motor control signals or electronic parameter tuning instructions.

[0009] The multimodal environmental sensing module is installed on the top surface of the device's outer shell and at the four corners of the base. The light intensity sensor uses a silicon photodiode structure with a range of 0–120,000 lux and an accuracy of ±2%. The wind speed sensor is an ultrasonic anemometer, installed at the front end of the top bracket of the device, with a measurement range of 0–60 m / s and a resolution of 0.1 m / s. The noise sensor is a MEMS digital microphone with a sampling rate of 48 kHz and a dynamic range of 30–120 dB SPL. The three-axis tilt sensor is a fusion module of MEMS accelerometer and gyroscope, installed at the center of the base, with a measurement range of ±90° and an accuracy of ±0.5°. The temperature and humidity sensor is a capacitive digital sensor, installed inside the ventilation holes on the side wall of the shell, with a temperature measurement range of -20℃ to +70℃ and an accuracy of ±0.3℃, and a humidity measurement range of 0–100% RH and an accuracy of ±2% RH.

[0010] In the audio and video acquisition module, the liftable camera is installed in the inner cavity of the main housing via a lead screw and slide rail mechanism. The lead screw is driven by a stepper motor with a stroke of 0–300 mm and a positioning accuracy of ±0.1 mm. The rotatable microphone array consists of a ring structure composed of four omnidirectional MEMS microphones, fixed on a rotating platform. The rotating platform is driven by a servo motor with a rotation angle of 0–360° and an angular resolution of 0.1°. The supplementary lighting group consists of three sets of LED beads, each set containing red, green, blue, and white LEDs. The brightness is independently controlled by a PWM dimming circuit, with a maximum illuminance of 5000 lux and a color temperature adjustment range of 2700K–6500K.

[0011] The adaptive actuators include a center-of-gravity adjustment mechanism comprising a counterweight slider, a guide rail, and a linear motor. The counterweight slider has a mass of 1.2 kg and is mounted on a transverse guide rail inside the base. The guide rail is 400 mm long. The linear motor has a maximum thrust of 50 N and a response time ≤0.3 s. The camera attitude adjustment mechanism includes a pitch motor and a yaw motor. The pitch motor is mounted at the bottom of the camera gimbal and has an output torque of 0.8 N·m. The yaw motor is mounted on the gimbal base and has an output torque of 1.2 N·m. The microphone directivity adjustment mechanism includes a phase delay controller and a beamforming algorithm module. The phase delay controller is electrically connected to four microphones and uses an FPGA to achieve nanosecond-level delay adjustment. The supplementary lighting intensity adjustment mechanism includes an optical feedback loop, consisting of a photoresistor and an LED driver circuit forming a closed-loop control. The photoresistor is mounted 5 mm in front of the lens and has a response time ≤10 ms.

[0012] Preferably, the adaptive actuator also includes a terrain-adaptive support foot assembly, which includes three independent telescopic legs, a pressure sensor, and a height adjustment motor. Each leg consists of an aluminum alloy sleeve and an inner rod, which is driven by a micro DC motor with a stroke of 0–150 mm. The pressure sensor is installed on the bottom contact surface of the leg, with a range of 0–50 kg and an accuracy of ±0.1 kg. The three legs are distributed in a 120° circumference and are connected to the lower end of the base via ball joints. The ball joints allow the legs to swing freely within a range of ±15° to adapt to uneven ground.

[0013] Preferably, the multimodal environment perception module also includes an environmental data fusion and correction unit. The environmental data fusion and correction unit includes a Kalman filter and a cross-validation logic module. The Kalman filter performs time synchronization and noise suppression on data from three types of sensors: wind speed, noise, and tilt angle. The cross-validation logic module compares the physical consistency of light intensity and temperature and humidity data, removes outliers, and outputs the corrected environmental state vector.

[0014] Preferably, the main control unit also includes a local training module, which includes a lightweight neural network model and an online learning interface. The lightweight neural network model is a MobileNetV2 structure, with the input layer receiving the environment state vector, the output layer generating device adjustment parameters, and the model weights stored in non-volatile memory. After each training session, the model parameters are updated through the online learning interface.

[0015] Preferably, the center of gravity adjustment mechanism is connected to the base via a guide rail groove, the bottom of the counterweight slider is equipped with a ball bearing to reduce frictional resistance, and the linear motor is connected to the counterweight slider via a coupling; the camera attitude adjustment mechanism is fixedly connected to the liftable camera via a flange, and the output shaft of the pitch motor is coaxially mounted with the camera pan-tilt shaft; the microphone directivity adjustment mechanism is electrically connected to the rotatable microphone array via a flexible ribbon cable, and the phase delay controller is installed inside the main control unit; the fill light intensity adjustment mechanism is connected to the fill light group via a constant current drive chip, and the photoresistor signal in the light feedback loop is fed into the ADC converter and then sent to the edge computing processor.

[0016] The terrain-adaptive support foot assembly is mechanically connected to the base via a ball joint. The pressure sensor signal line is introduced into the main control unit through a waterproof connector. The height adjustment motor is controlled by an H-bridge drive circuit. The environmental data fusion and correction unit is connected to the multimodal environmental perception module via an SPI bus. The corrected data is stored in a circular buffer of non-volatile memory. The local training module shares the NPU acceleration unit of the edge computing processor with the main control unit. The training data comes from the environment-equipment operation logs of each training session.

[0017] The structural composition, implementation method, and operating principle of this invention are as follows:

[0018] When the device is deployed in an outdoor environment, the light intensity sensor, wind speed sensor, noise sensor, three-axis tilt sensor, and temperature and humidity sensor in the multimodal environment perception module synchronously collect environmental parameters. The raw data is transmitted to the main control unit via the SPI bus. The edge computing processor calls the Kalman filter in the environmental data fusion and correction unit to perform time alignment and noise filtering on the wind speed, noise, and tilt data. At the same time, the cross-validation logic module judges whether there is a physical contradiction between light intensity and temperature and humidity (such as a sudden drop in temperature and humidity under strong light). If so, the data frame is marked as abnormal and discarded. The corrected environmental state vector is input to the lightweight neural network model in the local training module. The model outputs the center of gravity position offset, camera pitch angle, microphone beam main lobe direction, supplementary light color temperature and brightness, and extension length of each support leg according to the pre-trained environment-device mapping rules. The instruction scheduler converts the above outputs into specific motor control pulse sequences and electronic parameter adjustment instructions, which are sent to the linear motor of the center of gravity adjustment mechanism, the pitch and yaw motors of the camera attitude adjustment mechanism, and the microphone pointing mechanism, respectively. The system includes a phase delay controller for the dynamic adjustment mechanism, a PWM dimming circuit for the supplementary lighting intensity adjustment mechanism, and a height adjustment motor for the terrain-adaptive support leg assembly. The center of gravity adjustment mechanism calculates the optimal counterweight position based on wind speed and direction vectors, and a linear motor drives the counterweight slider along the guide rail to lower the overall center of gravity. The camera posture adjustment mechanism adjusts the pitch angle based on the light intensity and distance to the subject to avoid overexposure in backlight or shadows in frontlight. The microphone directivity adjustment mechanism adjusts the delay of each microphone channel using a phase delay controller based on the noise source location and the interviewee's position to form a directional beam to suppress background noise. The supplementary lighting intensity adjustment mechanism monitors the actual illuminance in real time through a light feedback loop and dynamically adjusts the LED drive current to maintain the set illuminance value. The terrain-adaptive support leg assembly controls the extension and retraction length of the three legs based on the ground slope detected by a three-axis tilt sensor, restoring the base to a level state. Simultaneously, pressure sensors monitor the force on each leg to prevent local overload and tipping. The entire adjustment process is completed locally on the edge computing processor, without relying on a remote server, with a response latency of less than 200 ms.

[0019] The beneficial effects of this invention are:

[0020] 1. The terrain-adaptive support foot assembly works in conjunction with the center of gravity adjustment mechanism**. Through real-time feedback from the three-axis tilt sensor and pressure sensor, it achieves automatic level correction of the base and dynamic adjustment of the center of gravity, maintaining the stability of the device on uneven ground with a slope of no more than 15°. This is significantly better than the structure in CN116772063A that relies solely on manual adjustment.

[0021] 2. The microphone directivity adjustment mechanism, combined with a phase delay controller and beamforming algorithm**, automatically forms a directional pickup beam in environments with noise levels exceeding 75 dB SPL, suppressing noise from non-target directions. This solves the problem in CN114857459B where the audio acquisition unit cannot adaptively adjust to the noise environment.

[0022] 3. The integration of the environmental data fusion correction unit with the local training module enables the device to optimize adjustment strategies based on historical data during continuous use, avoiding misadjustment caused by the failure of a single sensor or sudden environmental changes, and improving the robustness and long-term adaptability of the system in complex outdoor scenarios. Attached Figure Description

[0023] Figure 1 This is a three-dimensional schematic diagram of the overall structure of the outdoor interview training device based on environmental adaptation according to the present invention.

[0024] Figure 2 This is a schematic diagram showing the connection relationship between the multimodal environment sensing module and the main control unit, as well as the sensor layout of the device of the present invention.

[0025] Figure 3 This is a schematic diagram of the internal structure of the audio and video acquisition module in the device of the present invention, and the assembly relationship of the liftable camera, the rotatable microphone array and the fill light group.

[0026] Figure 4 This is an exploded view of the center-of-gravity adjustment mechanism and the terrain-adaptive support leg assembly of the adaptive actuator in the device of the present invention.

[0027] Figure 5 This is a schematic diagram of the drive and control connection between the camera posture adjustment mechanism and the microphone directivity adjustment mechanism in the device of the present invention;

[0028] Figure 6 This is a flowchart illustrating the internal module composition and data flow interaction of the main control unit of this invention.

[0029] Figure 7 This is a partially enlarged schematic diagram of a single telescopic leg structure of the terrain-adaptive support leg assembly of the present invention and its connection with the ball joint of the base.

[0030] Figure 8 This is a schematic diagram of the electrical connection between the light feedback loop and the LED driving circuit of the supplementary light intensity adjustment mechanism of the present invention.

[0031] In the diagram: 1. Main control unit; 2. Multimodal environment perception module; 3. Audio and video acquisition module; 4. Adaptive actuator; 5. Power supply module; 6. Communication module; 7. Light intensity sensor; 8. Wind speed sensor; 9. Noise sensor; 10. Three-axis tilt sensor; 11. Temperature and humidity sensor; 12. Liftable camera; 13. Rotatable microphone array; 14. Fill light assembly; 15. Center of gravity adjustment mechanism; 16. Camera posture adjustment mechanism; 17. Microphone directivity adjustment mechanism; 18. Fill light intensity adjustment mechanism; 19. Terrain adaptive support foot assembly; 20. Edge computing processor; 21. Non-volatile memory; 22. Real-time operating system module; 23. Command scheduler; 24. Environmental data fusion and correction unit; 25. Local training module; 26. Lead screw and slide rail mechanism; 27. Stepper motor; 28. Rotating platform; 29. Servo motor; 30. LED beads; 31. PWM dimming circuit; 32. Counterweight slider; 33. Guide rail; 34. Linear motor; 35. Pitch motor; 36. Yaw motor; 37. Phase delay controller; 38. Beamforming algorithm module; 39. Photoresistor; 40. Constant current drive chip; 41. Telescopic outrigger; 42. Pressure sensor; 43. Height adjustment motor; 44. Ball joint; 45. Ball bearing; 46. Coupling; 47. Flange; 48. Flexible cabling; 49. ADC converter; 50. H-bridge drive circuit; 51. Waterproof connector; 52. Kalman filter; 53. Cross-validation logic module; 54. Lightweight neural network model; 55. Online learning interface; 56. NPU acceleration unit; 57. Circular buffer; 58. MEMS microphone; 59. FPGA; 60. Main housing. Detailed Implementation

[0032] To make the technical means, creative features, objectives and effects of this invention easier to understand, the invention will be further described below in conjunction with specific embodiments.

[0033] This invention provides an environmentally adaptive outdoor interview training device, addressing the shortcomings of existing outdoor interview devices in complex and changing environments (including strong light, wind, rain, noise, and uneven terrain). These shortcomings include a lack of comprehensive perception of multiple environmental factors, inability to automatically optimize equipment parameters, reliance on manual intervention for angle and height adjustments, lack of intelligent decision-making mechanisms based on environmental data, and difficulty in achieving dynamic coordination among environment, equipment, and content. This invention aims to improve the autonomy, robustness, and completeness of information collection during outdoor interview training by constructing an integrated structure system that combines multimodal environmental sensors, edge computing units, and adaptive actuators. This system enables real-time acquisition and localized processing of external environmental parameters such as light intensity, wind speed, noise level, and ground tilt, as well as closed-loop control of equipment action commands.

[0034] To better understand the above technical solutions, the following will provide a detailed explanation of the technical solutions in conjunction with the accompanying drawings and specific implementation methods.

[0035] See Figures 1 to 8 An outdoor interview training device based on environmental adaptation includes a main control unit 1, a multimodal environmental perception module 2, an audio and video acquisition module 3, an adaptive actuator 4, a power supply module 5, and a communication module 6. The main control unit 1 is electrically connected to the multimodal environmental perception module 2, the audio and video acquisition module 3, the adaptive actuator 4, the power supply module 5, and the communication module 6. The multimodal environmental perception module 2 includes a light intensity sensor 7, a wind speed sensor 8, a noise sensor 9, a three-axis tilt sensor 10, and a temperature and humidity sensor 11. The audio and video acquisition module 3 includes a liftable camera 12, a rotatable microphone array 13, and a supplementary light group 14. The adaptive actuator 4 includes a center of gravity adjustment mechanism 15, a camera posture adjustment mechanism 16, a microphone directivity adjustment mechanism 17, and a supplementary light intensity adjustment mechanism 18.

[0036] See Figure 1 and Figure 2 The multimodal environmental sensing module 2 is installed on the top surface of the outer shell and at the four corners of the base. Among them, the light intensity sensor 7 adopts a silicon photodiode structure with a range of 0–120000 lux and an accuracy of ±2%, and is installed at the top center of the main shell 60; the wind speed sensor 8 is an ultrasonic anemometer, installed at the front end of the top bracket of the device, with a measurement range of 0–60 m / s and a resolution of 0.1 m / s; the noise sensor 9 is a MEMS digital microphone 58 with a sampling rate of 48 kHz and a dynamic range of 30–120 dB SPL, and is installed at the top edge of the main shell 60; the three-axis tilt sensor 10 is a MEMS accelerometer and gyroscope fusion module, installed at the center of the base, with a measurement range of ±90° and an accuracy of ±0.5°; the temperature and humidity sensor 11 is a capacitive digital sensor, installed inside the ventilation hole on the side wall of the shell, with a temperature measurement range of -20℃ to +70℃ and an accuracy of ±0.3℃, and a humidity measurement range of 0–100% RH and an accuracy of ±2% RH.

[0037] See Figure 3In the audio and video acquisition module 3, the liftable camera 12 is installed in the inner cavity of the main housing 60 via a lead screw and slide rail mechanism 26. The lead screw is driven by a stepper motor 27, with a stroke of 0–300 mm and a positioning accuracy of ±0.1 mm. The rotatable microphone array 13 consists of a ring structure composed of four omnidirectional MEMS microphones 58, which is fixed on the rotating platform 28. The rotating platform 28 is driven by a servo motor 29, with a rotation angle of 0–360° and an angular resolution of 0.1°. The supplementary lighting group 14 consists of three groups of LED beads 30, each containing red, green, blue, and white LEDs. The brightness is independently controlled by a PWM dimming circuit 31, with a maximum illuminance of 5000 lux and a color temperature adjustment range of 2700K–6500K.

[0038] See Figure 4 and Figure 7 In the adaptive actuator 4, the center of gravity adjustment mechanism 15 includes a counterweight slider 32, a guide rail 33, and a linear motor 34. The counterweight slider 32 has a mass of 1.2 kg and is installed on the transverse guide rail 33 inside the base. The guide rail 33 has a length of 400 mm. The linear motor 34 has a maximum thrust of 50 N and a response time of ≤0.3 s. The terrain adaptive support leg assembly 19 includes three independent telescopic legs 41, a pressure sensor 42, and a height adjustment motor 43. Each telescopic leg 41 is composed of an aluminum alloy sleeve and an inner rod. The inner rod is driven by a micro DC motor with a stroke of 0–150 mm. The pressure sensor 42 is installed on the bottom contact surface of the telescopic leg 41 with a range of 0–50 kg and an accuracy of ±0.1 kg. The three telescopic legs 41 are distributed in a 120° circle and are connected to the lower end of the base through ball joints 44. The ball joints 44 allow the telescopic legs 41 to swing freely within a range of ±15° to adapt to uneven ground.

[0039] See Figure 5 The camera attitude adjustment mechanism 16 includes a pitch motor 35 and a yaw motor 36. The pitch motor 35 is mounted on the bottom of the camera gimbal and has an output torque of 0.8 N·m. The yaw motor 36 is mounted on the gimbal base and has an output torque of 1.2 N·m. The microphone directivity adjustment mechanism 17 includes a phase delay controller 37 and a beamforming algorithm module 38. The phase delay controller 37 is electrically connected to four MEMS microphones 58 and achieves nanosecond-level time delay adjustment through FPGA 59.

[0040] See Figure 6The main control unit 1 includes an edge computing processor 20, a non-volatile memory 21, a real-time operating system module 22, and an instruction scheduler 23. The edge computing processor 20 is connected to the non-volatile memory 21 via a PCIe bus and is used to store a preset environment-device mapping rule table and historical operating data. The real-time operating system module 22 is embedded inside the edge computing processor 20 and is used to schedule multi-task parallel processing. The instruction scheduler 23 is connected to the edge computing processor 20 via an internal bus and is used to convert the processing results into specific motor control signals or electronic parameter tuning instructions. The main control unit 1 also includes an environmental data fusion and correction unit 24 and a local training module 25. The environmental data fusion and correction unit 24 includes a Kalman filter 52 and a cross-validation logic module 53. The local training module 25 includes a lightweight neural network model 54 and an online learning interface 55. The lightweight neural network model 54 is a MobileNetV2 structure. The input layer receives the environmental state vector, the output layer generates device adjustment parameters, and the model weights are stored in the non-volatile memory 21.

[0041] See Figure 8 The supplementary light intensity adjustment mechanism 18 includes a light feedback loop, which is a closed-loop control composed of a photoresistor 39 and an LED driving circuit. The photoresistor 39 is installed 5 mm in front of the lens and has a response time of ≤10 ms. The signal from the photoresistor 39 is sent to the edge computing processor 20 after being connected to the ADC converter 49. The LED driving circuit includes a constant current driving chip 40 and a PWM dimming circuit 31.

[0042] See Figure 4 The center of gravity adjustment mechanism 15 is connected to the base via a guide rail groove. A ball bearing 45 is provided at the bottom of the counterweight slider 32 to reduce frictional resistance. The linear motor 34 is connected to the counterweight slider 32 via a coupling 46. The camera attitude adjustment mechanism 16 is fixedly connected to the liftable camera 12 via a flange 47. The output shaft of the pitch motor 35 is coaxially mounted with the camera pan-tilt head shaft. The microphone directivity adjustment mechanism 17 is electrically connected to the rotatable microphone array 13 via a flexible cable 48. The phase delay controller 37 is installed inside the main control unit 1. Terrain adaptive... The support foot assembly 19 is mechanically connected to the base via a ball joint 44. The signal line of the pressure sensor 42 is introduced into the main control unit 1 through a waterproof connector 51. The height adjustment motor 43 is controlled by the H-bridge drive circuit 50. The environmental data fusion correction unit 24 is connected to the multimodal environmental perception module 2 via an SPI bus. The corrected data is stored in the ring buffer 57 of the non-volatile memory 21. The local training module 25 shares the NPU acceleration unit 56 of the edge computing processor 20 with the main control unit 1. The training data comes from the environment-equipment operation log of each training session.

[0043] In practical operation, when the device is deployed in an outdoor environment, the light intensity sensor 7, wind speed sensor 8, noise sensor 9, triaxial tilt sensor 10, and temperature and humidity sensor 11 in the multimodal environmental sensing module 2 synchronously collect environmental parameters. The raw data is transmitted to the main control unit 1 via the SPI bus. The edge computing processor 20 calls the Kalman filter 52 in the environmental data fusion and correction unit 24 to perform time alignment and noise filtering on the wind speed, noise, and tilt data. At the same time, the cross-validation logic module 53 judges whether there is a physical contradiction between light intensity and temperature and humidity (such as a sudden drop in temperature and humidity under strong light). If so, The data frame is marked as abnormal and discarded; the corrected environment state vector is input into the lightweight neural network model 54 in the local training module 25. The model outputs the center of gravity position offset, camera pitch angle, microphone beam main lobe direction, fill light color temperature and brightness, and the extension length of each telescopic leg 41 according to the pre-trained environment-device mapping rules; the command scheduler 23 converts the above outputs into specific motor control pulse sequences and electronic parameter adjustment commands, which are sent to the linear motor 34 of the center of gravity adjustment mechanism 15, the pitch motor 35 and yaw motor 36 of the camera attitude adjustment mechanism 16, and the microphone directivity adjustment, respectively. The mechanism 17 includes a phase delay controller 37, a PWM dimming circuit 31 for the fill light intensity adjustment mechanism 18, and a height adjustment motor 43 for the terrain adaptive support foot assembly 19; the center of gravity adjustment mechanism 15 calculates the optimal counterweight position based on wind speed and wind direction vector, and a linear motor 34 drives the counterweight slider 32 to move along the guide rail 33 to lower the overall center of gravity; the camera posture adjustment mechanism 16 adjusts the pitch angle based on the light intensity and the distance to the subject to avoid backlighting overexposure or front lighting shadows; the microphone directivity adjustment mechanism 17 adjusts each MEM through the phase delay controller 37 based on the noise source location and the interviewee's position. The delay of the 58 channels of the S microphone forms a directional beam to suppress background noise; the supplementary light intensity adjustment mechanism 18 monitors the actual illuminance in real time through the light feedback loop and dynamically adjusts the LED drive current to maintain the set illuminance value; the terrain adaptive support foot assembly 19 controls the extension length of the three telescopic legs 41 according to the ground slope detected by the triaxial tilt sensor 10, so that the base returns to a horizontal state, while the pressure sensor 42 monitors the force on each telescopic leg 41 to prevent local overload from causing overturning; the entire adjustment process is completed locally in the edge computing processor 20 without relying on a remote server, and the response latency is less than 200 ms.

[0044] For example, during an interview training exercise in an outdoor mountainous environment with a slope of 12°, the three-axis tilt sensor 10 detected that the base tilt angle was 12°. The main control unit 1, through the command scheduler 23, controlled the three height adjustment motors 43 to drive the three telescopic legs 41 to extend to different lengths, so that the base returned to a horizontal position. At the same time, the wind speed sensor 8 detected a wind speed of 8 m / s. After calculating the wind direction vector, the main control unit 1 drove the linear motor 34 to move the counterweight slider 32 150 mm towards the windward side, lowering the center of gravity and improving stability. The light intensity sensor 7 detected an ambient illuminance of 90,000 lux. The main control unit 1 controlled the supplementary lighting group 14 to turn off and adjusted the tilt angle of the liftable camera 12 to -15° to avoid backlighting. The noise sensor 9 detected a background noise of 80 dBSPL. The main control unit 1, through the phase delay controller 37, adjusted the time delay of the four MEMS microphones 58 to form a pickup beam pointing towards the interviewee, effectively suppressing background noise. The entire process was completed within 200 ms, realizing the device's adaptive and stable operation and high-quality audio and video acquisition in complex environments.

Claims

1. An outdoor interview training device based on environmental adaptation, characterized in that, The system includes a main control unit (1), a multimodal environment perception module (2), an audio and video acquisition module (3), an adaptive actuator (4), a power supply module (5), and a communication module (6). The main control unit (1) is electrically connected to the multimodal environment perception module (2), the audio and video acquisition module (3), the adaptive actuator (4), the power supply module (5), and the communication module (6), respectively. The multimodal environment perception module (2) includes a light intensity sensor (7), a wind speed sensor (8), a noise sensor (9), a three-axis tilt sensor (10), and a temperature and humidity sensor (11). The audio and video acquisition module (3) includes a liftable camera (12), a rotatable microphone array (13), and a fill light group (14). The adaptive actuator (4) includes a center of gravity adjustment mechanism (15), a camera posture adjustment mechanism (16), a microphone directivity adjustment mechanism (17), and a fill light intensity adjustment mechanism (18).

2. The outdoor interview training device based on environmental adaptation according to claim 1, characterized in that, The main control unit (1) includes an edge computing processor (20), a non-volatile memory (21), a real-time operating system module (22), and an instruction scheduler (23). The edge computing processor (20) is connected to the non-volatile memory (21) via a PCIe bus. The real-time operating system module (22) is embedded inside the edge computing processor (20). The instruction scheduler (23) is connected to the edge computing processor (20) via an internal bus.

3. The outdoor interview training device based on environmental adaptation according to claim 1, characterized in that, The multimodal environment perception module (2) further includes an environment data fusion correction unit (24), which includes a Kalman filter (52) and a cross-validation logic module (53). The environment data fusion correction unit (24) is connected to the multimodal environment perception module (2) via an SPI bus.

4. The outdoor interview training device based on environmental adaptation according to claim 1, characterized in that, In the audio and video acquisition module (3), the liftable camera (12) is installed in the inner cavity of the main housing (60) through the lead screw and slide rail mechanism (26), which is driven by the stepper motor (27); the rotatable microphone array (13) consists of four omnidirectional MEMS microphones (58) forming a ring structure and is fixed on the rotating platform (28), which is driven by the servo motor (29); the fill light group (14) consists of three groups of LED beads (30), each group containing red, green, blue and white LEDs, and the brightness is independently controlled by the PWM dimming circuit (31).

5. The outdoor interview training device based on environmental adaptation according to claim 1, characterized in that, The adaptive actuator (4) also includes a terrain-adaptive support foot assembly (19), which includes three independent telescopic legs (41), a pressure sensor (42), and a height adjustment motor (43). Each telescopic leg (41) is composed of an aluminum alloy sleeve and an inner rod, which is driven by a micro DC motor. The pressure sensor (42) is installed on the bottom contact surface of the telescopic leg (41). The three telescopic legs (41) are distributed in a 120° circle and connected to the lower end of the base through a ball joint (44).

6. The outdoor interview training device based on environmental adaptation according to claim 1, characterized in that, The center of gravity adjustment mechanism (15) includes a counterweight slider (32), a guide rail (33) and a linear motor (34). The counterweight slider (32) has a mass of 1.2 kg and is installed on the transverse guide rail (33) inside the base. The guide rail (33) has a length of 400 mm. The linear motor (34) has a maximum thrust of 50 N. The bottom of the counterweight slider (32) is provided with a ball bearing (45). The linear motor (34) is connected to the counterweight slider (32) through a coupling (46).

7. The outdoor interview training device based on environmental adaptation according to claim 1, characterized in that, The camera attitude adjustment mechanism (16) includes a pitch motor (35) and a yaw motor (36). The pitch motor (35) is installed at the bottom of the camera gimbal and has an output torque of 0.8 N·m. The yaw motor (36) is installed on the gimbal base and has an output torque of 1.2 N·m. The camera attitude adjustment mechanism (16) is fixedly connected to the liftable camera (12) through a flange (47). The output shaft of the pitch motor (35) is coaxially installed with the camera gimbal shaft.

8. The outdoor interview training device based on environmental adaptation according to claim 1, characterized in that, The microphone directivity adjustment mechanism (17) includes a phase delay controller (37) and a beamforming algorithm module (38). The phase delay controller (37) is electrically connected to four MEMS microphones (58) in the rotatable microphone array (13) via a flexible ribbon cable (48). The phase delay controller (37) is installed inside the main control unit (1) and the nanosecond-level time delay adjustment is achieved through the FPGA (59).

9. The outdoor interview training device based on environmental adaptation according to claim 1, characterized in that, The supplementary light intensity adjustment mechanism (18) includes a light feedback loop composed of a photoresistor (39) and an LED driving circuit. The photoresistor (39) is installed 5 mm in front of the lens, and its signal is connected to the edge computing processor (20) via an ADC converter (49). The LED driving circuit includes a constant current driving chip (40) and a PWM dimming circuit (31).

10. The outdoor interview training device based on environmental adaptation according to claim 2, characterized in that, The main control unit (1) also includes a local training module (25), which includes a lightweight neural network model (54) and an online learning interface (55). The lightweight neural network model (54) is a MobileNetV2 structure. Its input layer receives the environmental state vector, and its output layer generates device adjustment parameters. The model weights are stored in a non-volatile memory (21). The local training module (25) shares an NPU acceleration unit (56) with the edge computing processor (20).

Images