Vehicle-mounted noise reduction system

Abstract

The invention provides a vehicle-mounted noise reduction system. The vehicle-mounted noise reduction system comprises a multi-mode sensor array, at least one vibration sensor, a plurality of high-precision acoustic sensors, an IMU unit and a temperature and humidity sensor. And the signal processing module is used for executing noise feature extraction, multi-sensor data fusion and dynamic weight distribution. The self-adaptive control module comprises a filtering module and a local mute area control module and is used for being connected with a vehicle-mounted CAN bus and adjusting noise reduction parameters according to the vehicle state. And the self-calibration unit is used for periodically detecting the health state of the sensor, implementing temperature compensation and redundancy switching, filtering the monitored sound wave, analyzing based on the filtered sound wave information, comparing the sound wave information with the to-be-denoised parameter to calculate a sound wave information difference value, and transmitting the sound wave information difference value to the self-adaptive control module. The situation that noise data errors are large due to road noise and wind noise is avoided, the situation that the noise reduction effect fails due to collection errors is avoided, and the mute performance in the vehicle is guaranteed.

Description

Technical Field

[0001] The present invention relates to the technical field of automobiles, and particularly relates to an in-vehicle noise reduction system. Background Art

[0002] With the increasing demand for comfort, high-end vehicle models have adopted ANC active noise reduction technology to enhance the luxury and comfort of driving. Traditional passive noise reduction technology mainly achieves noise reduction by using measures such as silent tires, optimizing the suspension structure and elastic elements, and using acoustic wrapping materials. However, these methods have led to problems such as increased costs, performance requirements, and large energy of low-frequency noise.

[0003] The layout of acquisition sensors in the in-vehicle noise reduction system mainly includes the layout of acceleration vibration sensors, acoustic sensors, and speakers. The acceleration vibration sensors are mainly used to collect the body vibration signals caused by road excitation, and are usually arranged at the attachment points of the body and the chassis, such as the body tower package position and the connection position of the subframe and the body. The acoustic sensors are used to detect the noise level after noise reduction in the vehicle, and are usually arranged at a position on the ceiling far from the human ear. The signal-to-noise ratio and sensitivity of the acoustic sensors are relatively high to ensure accurate detection of the noise level. The speakers are used to emit secondary sound sources, covering the full frequency bands of low, medium, and high frequencies. The low-frequency speakers are generally arranged at the rear of the vehicle, the medium-frequency speakers are arranged in the middle channel or IP, and the high-frequency speakers are arranged at the door panel or ceiling position.

[0004] Since the active noise reduction technology needs to comprehensively detect and collect wheel vibrations, tire noises, and engine operation noises, and effectively cancel the noises by emitting reverse sound waves, it requires high-precision sensors and powerful computing and processing capabilities to ensure accurate acquisition of noise signals. Moreover, the complexity of the algorithm increases with the increase of the noise reduction space, and the performance requirements for the processor are relatively high. In addition, vehicle noises not only come from the engine and transmission system, but also the wind noise and road noise during high-speed driving are also objects that need to be reduced. The dynamic changes of the engine, wind noise, and road noise lead to a lag in the response of the noise reduction algorithm and insufficient stability.

[0005] Among them, the acoustic sensors play a crucial role in the in-vehicle noise reduction system. They help the system analyze parameters such as the frequency, amplitude, and phase of the sound by capturing the sound signals inside and outside the vehicle, so as to evaluate the stability of the sound line. In applications, errors in the acoustic sensors will lead to inaccurate capture of the sound signals, which will in turn affect the results of the spectrum analysis. This may cause the system to fail to correctly identify the source and nature of abnormal noises and cannot effectively suppress the noises. Secondly, spectrum analysis is an important means for evaluating the stability of the sound line. If the data of the acoustic sensors is inaccurate, the results of the spectrum analysis will also be affected, which may lead to incorrect judgments of the noise distribution and affect the noise reduction effect.

[0006] In addition, incorrect noise recognition and suppression can lead to ineffective reduction of the in-vehicle noise level, affecting driving and riding comfort. Customers will experience abnormal noise or unstable sound, which will affect the overall driving experience. Summary of the Invention

[0007] In view of this, the present invention aims to provide an in-vehicle noise reduction system that can improve the stability of noise reduction under complex working conditions and ensure the noise reduction effect.

[0008] To achieve the above object, the technical solution of the present invention is realized as follows:

[0009] An in-vehicle noise reduction system, a multi-modal sensor array, including at least one vibration sensor, several high-precision acoustic sensors, an IMU unit, and a temperature and humidity sensor;

[0010] A signal processing module for performing noise feature extraction, multi-sensor data fusion, and dynamic weight allocation;

[0011] An adaptive control module, including a filtering module and a local silent area control module, for connecting to the in-vehicle CAN bus and adjusting the noise reduction parameters according to the vehicle state;

[0012] A self-calibration unit periodically detects the health status of the sensors and performs temperature compensation and redundant switching, and filters the monitored sound waves, analyzes based on the filtered sound wave information, and compares the sound wave information with the noise reduction parameters to calculate the sound wave information difference, and transmits the sound wave information difference to the adaptive control module.

[0013] Further, the signal processing module further includes a noise source separation module, and the noise source separation module includes vibration signal processing and acoustic signal processing;

[0014] The vibration signal processing is used to perform wavelet packet decomposition on the vibration signal and extract the frequency band energy characteristics;

[0015] The acoustic signal processing is used to perform STFT transformation on the microphone signal and classify the noise type through a convolutional neural network.

[0016] Further, the dynamic weight allocation method is as follows:

[0017] Calculate the confidence of each sensor based on the D-S evidence theory;

[0018] Weight calculation formula:

[0019] Where C i is the confidence, and E i is the proportion of noise energy.

[0020] Further, the temperature compensation algorithm of the self-calibration unit is as follows:

[0021] Scal(T) = S0 × [1 + k1(T - T0) + k2(T - T0) 2

[0022] where k1 = 0.005, k2 = -0.0002, and T0 = 25°C.

[0023] Further, the filtering module adopts a dynamic weight adjustment mechanism:

[0024] Based on the error signal e(n) = d(n) - y(n)

[0025] where D(n) is the desired signal and y(n) is the filter output signal.

[0026] The weights are updated through the LMS algorithm:

[0027] w(n + 1) = w(n) + μ * x(n) * e(n)

[0028] where μ is the step factor, which controls the convergence speed and stability.

[0029] Further, the filter output signal

[0030] where is the i-th frame of the original speech signal, n(t) is the noise, and y(n) is the speech signal contaminated by the noise;

[0031] S(n) = y(n) * H(n);

[0032] where H(n) is a linear filter.

[0033] Further, the vibration sensor is set at the position of the engine compartment, the top of the tower, and the connection between the subframe and the vehicle body; the acoustic sensor is set at the ceiling, the center console, and the headrest; the temperature and humidity sensor is set at the air outlet of the air conditioning system, the air duct, the air filter housing of the engine compartment, and the intake pipe.

[0034] Further, the acoustic sensor includes an outer cover arranged in an arc shape and a base arranged at the bottom of the outer cover;

[0035] Two adjacent microphones are arranged on the base, and a wind noise filtering part is arranged above each microphone. A pressing seat is arranged between the wind noise filtering part and the outer cover;

[0036] A main board is arranged between the microphone and the base, and the main board is electrically connected to the microphone.

[0037] ​Further, the self - calibration unit further includes an LSTM neural network prediction module. When the self - calibration unit detects the failure of the multimodal sensor, the LSTM neural network is enabled to predict the noise spectrum. The inputs of the LSTM network include historical noise data, IMU signals, and CAN bus vehicle speed information.

[0038] Compared with the prior art, the present invention has the following advantages:

[0039] The vehicle - mounted noise reduction system of the present invention collects vehicle noise and environmental parameters in real - time through an array of multimodal sensors such as vibration sensors, acoustic sensors, IMUs, and temperature - humidity sensors. By adopting the dynamic weight theory, the extraction and data fusion of the characteristic information collected by the multimodal sensors are realized. The noise reduction effect is optimized through the self - calibration unit, and it can judge whether the acoustic sensor fails according to the calibration information. It also filters and analyzes the acoustic wave information, and calculates the difference value of the acoustic wave information by comparing with the parameters to be noise - reduced, increasing the further verification of the data collected by the acoustic sensor, and avoiding the situation of large data errors caused by road noise and wind noise. At the same time, through the self - calibration unit, the IMU data and vibration sensor signals are used in time, and the prediction mode is switched, so as to avoid the failure of the noise reduction effect caused by data acquisition errors and effectively ensure the in - vehicle quiet performance. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] The drawings constituting a part of the present invention are used to provide a further understanding of the present invention. The schematic embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation to the present invention. In the drawings:

[0041] Figure 1 is a schematic flow chart of the vehicle - mounted noise reduction system according to an embodiment of the present invention;

[0042] Figure 2 is a cross - sectional view of the acoustic sensor according to an embodiment of the present invention.

[0043] DESCRIPTION OF THE REFERENCE NUMERALS:

[0044] 1. Outer cover; 2. Base; 3. Microphone; 4. Wind noise filtering part; 5. Pressing seat; 6. Main board. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0045] It should be noted that, without conflict, the embodiments in the present invention and the features in the embodiments can be combined with each other.

[0046] In the description of the present invention, it should be noted that the orientation or positional relationship indicated by the terms "upper", "lower", "inner", "back", etc. is based on the orientation or positional relationship shown in the drawings. It is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation. Therefore, it should not be construed as a limitation to the present invention. In addition, the terms "first" and "second" are only used for descriptive purposes and cannot be construed as indicating or implying relative importance.

[0047] In addition, in the description of the present invention, unless otherwise clearly defined, the terms "installation", "connection", "connection", "connector" should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be directly connected or indirectly connected through an intermediate medium, and it can be the communication inside two elements. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood in combination with specific situations.

[0048] The present invention will be described in detail below with reference to the drawings and in combination with embodiments.

[0049] This embodiment relates to a vehicle noise reduction system, which includes a multi-modal sensor array, including at least one vibration sensor, several high-precision acoustic sensors, an IMU unit, and a temperature and humidity sensor. A signal processing module for performing noise feature extraction, multi-sensor data fusion, and dynamic weight allocation. An adaptive control module, including a filtering module and a local silent area control module, for connecting to the vehicle CAN bus and adjusting the noise reduction parameters according to the vehicle state.

[0050] Among them, a self-calibration unit periodically detects the health status of the sensors and performs temperature compensation and redundancy switching, and filters the monitored sound waves, analyzes based on the filtered sound wave information, compares the sound wave information with the parameters to be noise-reduced to calculate the sound wave information difference, and transmits the sound wave information difference to the adaptive control module.

[0051] The vehicle noise reduction system described in this embodiment collects vehicle noise and environmental parameters in real time through an array of multimodal sensors, such as vibration sensors, acoustic sensors, IMUs, and temperature and humidity sensors. By adopting the dynamic weight theory, it realizes the extraction and data fusion of the characteristic information collected by the multimodal sensors, optimizes the noise reduction effect through the self-calibration unit, and can judge whether the acoustic sensor fails according to the calibration information. It also filters and analyzes the acoustic wave information, and compares it with the parameter to be noise-reduced to calculate the difference value of the acoustic wave information, which further verifies the data collected by the acoustic sensor and avoids the large data error of the noise caused by road noise and wind noise. At the same time, through the self-calibration unit, it timely switches to using IMU data and vibration sensor signals and switches to the prediction mode, thereby avoiding the failure of the noise reduction effect caused by incorrect data collection and effectively ensuring the in-vehicle quiet performance.

[0052] As a preferred implementation manner, the signal processing module further includes a noise source separation module, and the noise source separation module includes vibration signal processing and acoustic signal processing. The vibration signal processing is used to perform wavelet packet decomposition on the vibration signal to extract the frequency band energy characteristics, and the acoustic signal processing is used to perform STFT transformation on the microphone signal and classify the noise type through a convolutional neural network.

[0053] Regarding the vibration signal processing: perform wavelet packet decomposition on the acceleration signal. For example, the mother wavelet is db8, the decomposition level is 5, extract the frequency band energy characteristic E_band, the extraction efficiency is increased, and the calculation delay time is significantly reduced.

[0054] Regarding the acoustic signal processing: perform short-time Fourier transform on the microphone signal, with a frame length of 512 points and an overlap rate of 50%. Generate the time-frequency matrix and input it into the pre-trained ResNet-18 model to output the noise type label. The classification accuracy of the ResNet-18 model for wind noise, tire noise, and motor noise reaches 92.5%.

[0055] Preferably, the vibration sensor is set at the engine compartment, the top position, and the connection between the subframe and the vehicle body; specifically, the vibration sensor is located in the middle of the chassis, close to the drive system and the exhaust pipe, and is easily affected by the vibration of the power assembly and exhaust resonance, so key vibration isolation treatment is required. Close to the tire and suspension system, directly receiving the vibration energy of road noise and tire noise, it is a key coverage area for the sound insulation and noise reduction project. The vibration sensor adopts a MEMS accelerometer with a frequency response range of 5Hz to 5kHz, which is used to capture the contact noise between the tire and the road surface.

[0056] Among them, the acoustic sensor is set on the ceiling, the center console, and the headrest. Four digital microphones (SNR≥70dB) are distributed on the roof lining to form a tetrahedral structure, and the direction of the noise source is located through beamforming technology.

[0057] The temperature and humidity sensors are set at the air outlet of the air conditioning system, in the air duct, on the air filter housing of the engine compartment and the intake pipe, as well as inside the door seal strip, to monitor environmental parameters to compensate for the temperature drift of the sensors, such as the non-linear error of the microphone sensitivity varying with temperature. The IMU inertial measurement unit integrates a three-axis gyroscope and an accelerometer to monitor the vehicle's yaw angle, pitch angle and acceleration in real time, and is used to correct the offset of the noise transmission path caused by the change of the vehicle attitude.

[0058] Further, the dynamic weight allocation method is to calculate the confidence of each sensor based on the D-S evidence theory;

[0059] Weight calculation formula:

[0060] where C i is the confidence, and E i is the proportion of noise energy.

[0061] With such settings, the credibility of each sensor signal can be improved and the evaluation error can be reduced.

[0062] Preferably, the temperature compensation algorithm of the self-calibration unit is:

[0063] Scal(T)=S0×[1+k1(T-T0)+k2(T-T0) 2

[0064] where k1 = 0.005, k2 = -0.0002, and T0 = 25 °C.

[0065] In addition, the filtering module adopts a dynamic weight adjustment mechanism:

[0066] Based on the error signal e(n) = d(n) - y(n)

[0067] D(n) is the desired signal, and y(n) is the filter output signal:

[0068] Update the weight through the LMS algorithm:

[0069] w(n+1)=w(n)+μ*x(n)*e(n)

[0070] where μ is the step factor, which controls the convergence speed and stability.

[0071] Further, the filter output signal

[0072] where is the i-th frame of the original speech signal, n(t) is the noise, and y(n) is the speech signal contaminated by the noise;

[0073] S(n)=y(n)*H(n); ​

[0074] Among them, H(n) is a linear filter.

[0075] By calculating a linear filter H(n), the estimated value of the noisy speech y(n) after passing through the linear filter is S(n), so as to separate the pure original speech signal from the noisy speech interfered by noise, and improve the accuracy of the acquisition data of the acoustic sensor.

[0076] As a preferred implementation manner, as Figure 2 shown, the acoustic sensor includes an outer cover 1 arranged in an arc shape, and a base 2 provided at the bottom of the outer cover 1. Two adjacent microphones 3 are provided on the base 2. A wind noise filtering part 4 is provided above each microphone 3. A pressing seat 5 is provided between the wind noise filtering part 4 and the outer cover 1. A main board 6 is provided between the microphone 3 and the base 2, and the main board 6 is electrically connected to the microphone 3. Through the setting of the above structure, an acoustic sensor with two microphones 3 is set at the same detection point, and as Figure 2 shown. The outlets of the two microphones 3 are both arranged at the lower air inlet of the outer cover 1, so that the detection error caused by wind noise can be reduced.

[0077] In addition, the setting of the outer cover 1 and the wind noise filtering part 4 forms a simultaneous noise prevention treatment for the outer layer and the inner layer, further improving the data monitoring accuracy of the acoustic sensor, and more effectively helping to analyze the coherence function of the vehicle transmission path and the in-vehicle microphone 3 to determine the best noise reduction scheme.

[0078] The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A vehicle noise reduction system, characterized in that, Comprising: A multi-modal sensor array, including at least one vibration sensor, several high-precision acoustic sensors, an IMU unit, and a temperature and humidity sensor; A signal processing module for performing noise feature extraction, multi-sensor data fusion, and dynamic weight allocation; An adaptive control module, including a filtering module and a local mute zone control module, for connecting to the vehicle-mounted CAN bus and adjusting the noise reduction parameters according to the vehicle state; A self-calibration unit that periodically detects the health status of the sensors and performs temperature compensation and redundancy switching, and filters the monitored sound waves, analyzes based on the filtered sound wave information, and calculates the sound wave information difference by comparing the sound wave information with the noise reduction parameters to be adjusted, and transmits the sound wave information difference to the adaptive control module.

2. The vehicle-mounted noise reduction system according to claim 1, wherein: The signal processing module further includes a noise source separation module, and the noise source separation module includes vibration signal processing and acoustic signal processing; The vibration signal processing is used for performing wavelet packet decomposition on the vibration signal to extract the frequency band energy characteristics; The acoustic signal processing is used for performing STFT transformation on the microphone signal and classifying the noise type through a convolutional neural network.

3. The vehicle-mounted noise reduction system according to claim 2, wherein: The dynamic weight allocation method is: Calculating the confidence of each sensor based on the D-S evidence theory; Weight calculation formula: Among them, C i is the confidence level, and E i is the proportion of noise energy.

4. The vehicle-mounted noise reduction system according to claim 3, wherein: The temperature compensation algorithm of the self-calibration unit is: Scal(T) = S0 × [1 + k1(T - T0) + k2(T - T0) 2 ​ k1 = 0.005, k2 = -0.0002, T0 = 25°C.

5. The vehicle-mounted noise reduction system according to claim 4, wherein: The filtering module adopts a dynamic weight adjustment mechanism: Based on the error signal e(n) = d(n) - y(n) D(n) is the desired signal, y(n) is the filter output signal: Updating the weight through the LMS algorithm: w(n + 1) = w(n) + μ * x(n) * e(n) Wherein, μ is the step factor, controlling the convergence speed and stability.

6. The vehicle-mounted noise reduction system according to claim 5, wherein: The output signal of the filter Among them, is the original speech signal of the i-th frame, n(t) is the noise, and y(n) is the speech signal contaminated by the noise; S(n) = y(n) * H(n); Wherein, H(n) is a linear filter.

7. The vehicle-mounted noise reduction system according to claim 1, wherein: The vibration sensor is arranged at the engine compartment, the top position, and the connection between the subframe and the vehicle body; The acoustic sensors are arranged at the ceiling, the center console, and the headrests; the temperature and humidity sensors are arranged at the air outlet of the air conditioning system, the air duct, the air filter housing of the engine compartment, and the intake pipe.

8. The vehicle-mounted noise reduction system according to claim 2, wherein: The acoustic sensor includes an outer cover (1) arranged in an arc shape, and a base (2) arranged at the bottom of the outer cover (1); Two adjacent microphones (3) are arranged on the base (2), and a wind noise filtering part (4) is arranged above each microphone (3), and a pressing seat (5) is arranged between the wind noise filtering part (4) and the outer cover (1); A main board (6) is provided between the microphone (3) and the base (2), and the main board (6) is electrically connected to the microphone (3).

9. The vehicle-mounted noise reduction system according to claim 1, wherein: The self-calibration unit further includes an LSTM neural network prediction module. When the self-calibration unit detects the failure of the multimodal sensor, the LSTM neural network is enabled to predict the noise spectrum. The inputs of the LSTM network include historical noise data, IMU signals, and CAN bus vehicle speed information.

Images