System and method for virtual microphone in active noise cancellation

Abstract

Methods and systems for a vehicle audio system are disclosed, in one example, including a method for noise cancellation in a vehicle having: a reference sensor configured to acquire a reference signal; a plurality of speakers configured to emit a noise cancellation signal; and a plurality of error microphones configured to acquire a residual signal. The method: processes the reference signal with a time domain adaptive weight filter to generate the noise cancellation signal; a frequency domain filtered virtual microphone signal is estimated from the noise cancellation signal and the residual signal, and a frequency domain filtered reference signal is estimated from the reference signal. The method comprises: decomposing the frequency domain filtered virtual microphone signal and the frequency domain filtered reference signal into a plurality of frequency domain sub-band signals; and updating the time domain adaptive weight filter based on a weight transform of a plurality of frequency domain sub-band adaptive filter weights.

Description

Technical Field

[0001] This disclosure relates to systems and methods for active noise cancellation, and more particularly, to systems and methods for calculating an estimated virtual microphone signal and adjusting an adaptive weighted filter based on the estimated virtual microphone signal. Background Technology

[0002] Active noise cancellation (ANC) is a method for generating sound waves that cancel out unwanted sound waves. These canceling sound waves can be generated by transducers, such as those used in loudspeakers, to combine with the unwanted sound waves. This technology is widely used in various applications, including headphones, residential and commercial buildings, and automotive environments, to create a quieter and more comfortable acoustic experience. In automotive applications, ANC systems are particularly beneficial for reducing road noise, engine noise, and other external sounds that can penetrate the vehicle's cabin, thereby enhancing passenger comfort.

[0003] Virtual microphone technology (VMT) is used in current automotive applications. In automotive applications, it is desirable to eliminate noise near the driver's or passenger's ears, but placing microphones in those locations may be impractical. The location of one or more virtual microphones can be the area where noise cancellation is attempted. A VMT system in an automotive application may include one or more physical microphones placed within the vehicle cabin and a virtual microphone (VM) algorithm that calculates sound waves that can be generated by transducers such as speakers within the vehicle to create a quiet area at the location of one or more virtual microphones. Conventional VM algorithms are based on either the time-domain filtered x least mean square (FXLMS) algorithm or the time-frequency domain FXLMS algorithm.

[0004] However, the inventors have recognized potential problems with such systems in this paper. The FXLMS algorithm has inherent structural limitations when applied to VMT systems. These structural limitations constrain VM performance in terms of broadband noise cancellation, and specifically in terms of reducing high-frequency noise. Furthermore, conventional VM algorithms process signals on a sample-by-sample basis, which places a significant burden on the system's computational resources. Specifically, the FXLMS algorithm requires significant computational power to execute effectively, which can increase the amount of space and power required for computing devices within the vehicle. Compared to conventional methods, adapting the FXLMS algorithm using a subband adaptive filtering (SAF) algorithm has shown a certain reduction in computational requirements and an increase in broadband noise cancellation performance; however, further gains are desired. Summary of the Invention

[0005] This application provides systems and methods for virtual microphone technology (VMT) in active noise cancellation (ANC) systems, specifically for eliminating high-frequency noise incident on the listener's ears. This application discloses a method including a time-frequency subband virtual microphone (TFSVM) algorithm that addresses the computational limitations of conventional VMT systems and increases noise reduction over a wide frequency range.

[0006] In a first aspect, a method for noise cancellation in a vehicle is provided, the vehicle comprising: a reference sensor configured to acquire a reference signal relating to noise within a vehicle cabin; a plurality of loudspeakers positioned within the vehicle cabin, the plurality of loudspeakers configured to emit a noise cancellation signal to cancel noise around the ears of one or more vehicle occupants; and a plurality of physical error microphones positioned within the vehicle cabin, the plurality of physical error microphones configured to acquire a residual signal. The method includes: processing the reference signal using a time-domain adaptive weighted filter to generate the noise cancellation signal; and estimating a frequency-domain filtered virtual microphone signal based on the noise cancellation signal and the residual signal, and estimating a frequency-domain filtered reference signal based on the reference signal. The method further includes: decomposing the frequency-domain filtered virtual microphone signal and the frequency-domain filtered reference signal into a plurality of frequency-domain sub-band signals. A frequency-domain sub-band adaptive filter is calculated for each sub-band of the plurality of frequency-domain sub-band signals to generate a plurality of frequency-domain sub-band adaptive filters. The method further includes: updating the time-domain adaptive weighted filter based on a weighted transformation of the plurality of frequency-domain sub-band adaptive filters.

[0007] In a second aspect, a noise cancellation system for a vehicle is provided, the noise cancellation system comprising: a reference sensor configured to acquire a reference signal related to noise within a vehicle cabin; a time-domain adaptive weighted filter electronically communicating with the reference sensor, the time-domain adaptive weighted filter being configured to apply an adaptive filtering process to the reference signal to generate a noise cancellation signal, thereby eliminating noise around the ears of one or more vehicle occupants; a plurality of loudspeakers positioned within the vehicle cabin and electronically communicating with the time-domain adaptive weighted filter, the plurality of loudspeakers being configured to transmit the noise cancellation signal into the vehicle cabin; and a plurality of physical error microphones positioned within the vehicle cabin and configured to acquire a residual signal generated by the interaction between the noise cancellation signal and the noise within the vehicle cabin. The system includes: a signal processing unit electronically communicating with the reference sensor, the plurality of loudspeakers, and the plurality of physical error microphones. The signal processing unit includes: a processor; and a non-transitory memory storing a set of frequency domain subband filters and instructions. When executing the instructions, the processor is configured to: estimate a frequency-domain filtered virtual microphone signal based on the noise cancellation signal and the residual physical error microphone signal, and estimate a frequency-domain filtered reference signal based on the reference signal. The processor is configured to: decompose the frequency-domain filtered virtual microphone signal and the frequency-domain filtered reference signal into multiple frequency-domain sub-band signals; and calculate a frequency-domain sub-band adaptive filter for each sub-band of the multiple frequency-domain sub-band signals to generate multiple frequency-domain sub-band adaptive filters. The processor is configured to: update the time-domain adaptive weighted filter based on a weighted transformation of the multiple frequency-domain sub-band adaptive filters.

[0008] In a third aspect, a method for noise cancellation in a vehicle is provided. The method includes: acquiring a reference signal using a reference sensor, wherein the reference signal is correlated with noise in the vehicle compartment; processing the reference signal using a time-domain adaptive weighted filter to generate a noise-cancelled signal, the noise-cancelled signal being emitted using a plurality of loudspeakers positioned in the vehicle compartment to cancel noise around the ears of one or more vehicle occupants. The method includes: acquiring residual signals from a plurality of physical error microphones positioned in the vehicle compartment; and using a Fast Fourier Transform (FFT) to transform the noise-cancelled signal, the residual signal, and the reference signal into a frequency-domain noise-cancelled signal, a frequency-domain residual signal, and a frequency-domain reference signal, respectively. The method includes: applying a plurality of secondary path filters to the frequency-domain noise-cancelled signal and the frequency-domain residual signal to generate a frequency-domain filtered estimated virtual microphone signal, and applying the plurality of secondary path filters to the frequency-domain reference signal to generate a frequency-domain filtered reference signal. The method employs a set of frequency-domain sub-band filters to decompose the frequency-domain filtered reference signal and the frequency-domain filtered estimated virtual microphone signal into multiple frequency-domain sub-band filtered reference signals and multiple frequency-domain filtered estimated virtual microphone signals, respectively, and calculates a frequency-domain sub-band gradient for each sub-band based on the frequency-domain sub-band reference signals and frequency-domain sub-band error signals. The method includes determining a frequency-domain normalization step size for each sub-band based on the power contributions of the frequency-domain sub-band reference signals and the frequency-domain sub-band error signals. The method includes updating a set of frequency-domain sub-band adaptive filter weights based on the frequency-domain sub-band gradients and the frequency-domain normalization step size to generate an updated set of frequency-domain sub-band adaptive filter weights. The method includes transforming the updated set of frequency-domain sub-band adaptive filter weights from the frequency domain to the time domain using an inverse fast Fourier transform to generate an updated time-domain adaptive weight filter. The method includes transmitting the noise cancellation signal based on the updated time-domain adaptive weight filter to reduce noise around the ears of one or more vehicle occupants.

[0009] In this way, compared with conventional VM methods, the disclosed system and method reduce more broadband road noise near the ears of one or more vehicle occupants while requiring less computational resources. Attached Figure Description

[0010] This disclosure can be better understood by referring to the following description of non-limiting embodiments, in which:

[0011] Figure 1 Schematic diagrams illustrating active noise cancellation systems according to one or more embodiments of the present disclosure;

[0012] Figure 2A schematic diagram illustrating a virtual microphone noise cancellation system according to one or more embodiments of the present disclosure;

[0013] Figure 3 This is a graph comparing the performance of a conventional virtual microphone noise cancellation algorithm with the proposed Time-Frequency Subband Virtual Microphone (TFSVM) algorithm according to one or more embodiments of this disclosure.

[0014] Figure 4 A schematic diagram of a TFSVM system according to one or more embodiments of the present disclosure is shown;

[0015] Figure 5A A flowchart illustrating an example method for active noise cancellation using the TFSVM algorithm according to one or more embodiments of the present disclosure is shown.

[0016] Figure 5B The following describes one or more embodiments of the present disclosure. Figure 5A The flowchart continues with the example method; and

[0017] Figure 5C The following describes one or more embodiments of the present disclosure. Figure 5A The flowchart continues with the example method. Detailed Implementation

[0018] In one of many exemplary embodiments, an active noise cancellation (ANC) system, as described herein, reduces unwanted sounds present in the environment. Unwanted sounds are sounds that annoy a listener, such as vehicle engine sounds, road noise, etc., but when, for example, a listener wants to make a phone call, unwanted sounds can also be music or other people's voices. The disclosed systems and methods include virtual microphone (VM) algorithms for active noise cancellation (ANC) systems. VM algorithms can reduce noise around the ears of one or more vehicle occupants, regardless of the placement of the error microphone. The disclosed methods address many challenges that plague conventional VM algorithms, including structural limitations that limit the broadband frequency noise reduction performance of conventional methods, and signal processing requirements that impose a significant computational burden on conventional VM systems.

[0019] To address the aforementioned challenges, a novel Time-Frequency Subband Virtual Microphone (TFSVM) algorithm based on a time-frequency domain subband adaptive filter structure is proposed. Compared to conventional VM algorithms, the TFSVM algorithm computes the estimated virtual microphone signal in each subband and updates the adaptive filter individually for each frequency range. This method significantly reduces the computational load. Furthermore, to enhance the performance and convergence speed of the disclosed TFSVM algorithm, a method of subband flexible adaptive step size normalization is applied. In this way, compared to conventional methods, the method reduces more broadband noise near the ears of vehicle occupants while requiring less computational power.

[0020] The accompanying figures below illustrate various aspects of the systems and methods claimed herein. Figure 1 This is a schematic diagram of a noise cancellation system. The noise cancellation system may include multiple error microphones, a processor, and a loudspeaker. These error microphones, processor, and loudspeaker are capable of detecting ambient noise inside the vehicle cabin, processing the ambient noise, and determining a signal to be output to the loudspeaker to cancel the ambient noise in the vehicle cabin. Figure 2 This is a schematic diagram of a virtual microphone noise cancellation system. One implementation of a noise cancellation system in a vehicle involves using a virtual microphone, where noise cancellation can be focused on a location where a physical microphone is absent, and this location can be referred to as the virtual microphone's position. In this case, noise detected by a physical microphone within the cabin can be processed to predict the sound at the virtual microphone's position. The predicted sound at the virtual microphone's position can be used to determine a signal generated by a loudspeaker to cancel the noise at the virtual microphone's position, and can be used to determine residual noise at the virtual microphone's position after noise cancellation has been performed. The residual noise can be analyzed to further refine the noise cancellation signal. Figure 3 The graph compares the broadband noise cancellation performance implemented by the conventional VM algorithm and the proposed TFSVM algorithm, which is based on a time-frequency domain subband adaptive filter structure. Figure 4 This is a schematic representation of the TFSVM algorithm, which involves estimating the virtual microphone signal block by block in the frequency domain, reducing computational complexity. The graph also includes the application of a time-frequency subband adaptive filter structure that updates all adaptive filters in each subband across the entire frequency range. Figures 5A to 5C It describes the implementation. Figure 4 A flowchart illustrating the method of the TFSVM algorithm.

[0021] Go to Figure 1This diagram illustrates a block diagram of a vehicle noise cancellation system 100. The vehicle noise cancellation system 100 is configured to enhance the acoustic environment within a vehicle cabin 130 by actively reducing unwanted noise. The vehicle noise cancellation system 100 is equipped with a reference sensor 102, the task of which is to acquire a reference signal associated with noise present within the vehicle cabin 130. This reference signal serves as the basis for generating a noise cancellation signal to cancel the detected noise. In some examples, the reference sensor 102 may include at least one of: an accelerometer configured to detect vibrations associated with the vehicle; a microphone configured to detect ambient noise outside the vehicle cabin; and a non-acoustic sensor configured to detect operating parameters indicating noise generation of the vehicle.

[0022] The vehicle noise cancellation system 100 includes an adaptive weighted filter 104 that processes a reference signal obtained from a reference sensor 102 and applies an adaptive filtering algorithm to generate a noise-cancelled signal. The adaptive weighted filter 104 is capable of dynamically adjusting its filtering characteristics to reduce residual signals acquired through multiple error microphones 108.

[0023] The vehicle noise cancellation system 100 includes a plurality of speakers 106 strategically positioned within the vehicle compartment 130. The speakers 106 are configured to transmit noise cancellation signals into the compartment space, thereby creating an anti-noise sound field that interferes with unwanted noise to reduce or eliminate said unwanted noise.

[0024] To monitor the effectiveness of the noise cancellation process, multiple error microphones 108 are also positioned within the vehicle compartment 130. In one example, the multiple error microphones 108 may include a reference... Figure 2 One or more physical error microphones and one or more virtual error microphones are described in more detail. These error microphones 108 capture residual signals, which are the resulting sound after the emitted noise-cancelling signal interacts with the original noise. The residual signals provide feedback on the performance of the vehicle noise cancellation system 100. For example, reducing the residual signals captured by multiple error microphones may indicate greater effectiveness of the noise cancellation process.

[0025] The signal processing unit 110 serves as the computational center of the vehicle noise cancellation system 100. The signal processing unit 110 electronically communicates with both the reference sensor 102 and the error microphone 108. The signal processing unit 110 houses a processor 112 and a non-transitory memory 114, which together execute machine-readable instructions for canceling noise within the vehicle compartment 130. The processor 112 in the signal processing unit 110 is a hardware component designed to execute machine-executable instructions stored in the non-transitory memory 114, including instructions for computational tasks related to real-time signal processing, such as adaptive filtering algorithms, subband decomposition, and gradient calculation of filter weights. The processor 112 processes data to generate a real-time noise cancellation signal that cancels out unwanted noise in the vehicle compartment 130.

[0026] Non-transitory memory 114 stores machine-readable instructions and data of the vehicle noise cancellation system 100, thereby maintaining this information even when the system is off. The non-transitory memory includes firmware, software, and data structures or databases used by the processor 112, and may include ROM, flash memory, or other non-volatile storage technologies. Non-transitory memory 114 also stores historical data and adaptive filter coefficients used for system learning and performance enhancement.

[0027] The signal processing unit 110 of the vehicle noise cancellation system 100 uses a set of subband filters 116 to decompose the audio signal into multiple frequency subbands. The subband filters 116 are designed to divide a wide frequency range of the reference and residual signals into narrower bands, thereby allowing noise cancellation strategies to be tailored to the acoustic properties of each subband. The subband filters 116 begin with a prototype low-pass filter, which is then modulated to create a series of bandpass filters covering the entire frequency range of interest.

[0028] The impulse response of each subband filter (denoted as h) m The impulse response of the m-th subband filter is obtained from the filter through a modulation process that shifts the passband of the prototype filter to the frequency range of the target subband. A mathematical transformation incorporating modulation and windowing effects is used to determine the impulse response of the m-th subband filter. The number of subbands M and the length l of each subband filter are also considered. swThese are parameters that affect the resolution and computational requirements of the subband filtering process. Through filter bank analysis, subband filters 116 are used for both the reference signal and the residual signal. This involves convolving the input signal with the impulse response of each subband filter to isolate the subband components. The output is a set of subband reference signals and subband error signals, reflecting the frequency content of the original signal within each subband. By operating in the subband domain, the vehicle noise cancellation system 100 can more effectively perform noise cancellation tasks by focusing on and eliminating specific frequencies of sound.

[0029] Secondary path filters 118 are applied to the sub-band reference signal to generate a filtered sub-band reference signal. These secondary path filters 118 model the acoustic transfer function from loudspeaker 106 to error microphone 108 within each sub-band. These filters reflect the characteristics of the acoustic environment of the vehicle cabin, including cabin geometry, trim materials, and the variable presence of passengers or cargo. Each secondary path filter in the secondary path filters 118 corresponds to a specific sub-band and processes the associated sub-band reference signal while taking into account the frequency-dependent behavior of sound transmission, including reflection, absorption, and diffraction. The secondary path filters can be learned during secondary path calibration, as previously disclosed. To maintain accuracy, the system may include a calibration mechanism that adjusts the filter coefficients in response to environmental changes.

[0030] An adaptive step size determination module 120 is included in the signal processing unit 110. The adaptive step size determination module 120 adjusts the step size in the adaptive filtering algorithm of the vehicle noise cancellation system 100 on a per-sub-band basis. This adjustment affects the convergence rate and stability of the adaptive filter weights within the adaptive weighted filter 104. The adaptive step size determination module dynamically modifies the step size in each sub-band based on the power contribution of the error microphone signal and the reference signal within each corresponding sub-band, the step size affecting the convergence speed and stability of the adaptive filter weights. The adaptive step size determination module 120 calculates a normalized step size for each sub-band by evaluating factors such as the sum of the power of the filtered reference signal and the error signal, the individual power contribution of these signals, and a smoothness parameter related to their power. The module may also consider a power contribution parameter reflecting the maximum power within each sub-band. The obtained normalized step size is then applied to update the sub-band adaptive filter weights, thereby seeking a balance between convergence speed and stability.

[0031] The gradient determination module 122 calculates the sub-band gradient for each sub-band based on the filtered sub-band reference signal and the corresponding sub-band error signal. This continuous real-time adjustment allows the vehicle noise cancellation system 100 to effectively adapt to different noise conditions, thereby improving the acoustic experience inside the vehicle cabin.

[0032] The sub-band adaptive weight update module 124 updates the adaptive filter weights in each sub-band based on the calculated gradient and the determined adaptive step size. The sub-band adaptive weight update module 124 ensures that the vehicle noise cancellation system 100 adapts to the noise conditions within the vehicle compartment 130 in real time.

[0033] The weight transformation module 126 integrates the updated adaptive filter weights from each sub-band to produce the final weights of the adaptive weight filter 104 in the time domain. These updated weights are then applied to the adaptive weight filter 104 to adjust the noise cancellation signal for optimal noise reduction within the vehicle compartment 130.

[0034] Figure 2 A block diagram of a VM noise cancellation system 200 is shown. The VM noise cancellation system 200 is configured to reduce interior noise around the ears of a driver or passenger, regardless of the placement of the error microphone. In one example, the VM noise cancellation system 200 may be included in a multiple-input multiple-output (MIMO) active noise cancellation system as part of an overall method for reducing unwanted noise, such as road noise. In one example, the VM noise cancellation system 200 may be included in a vehicle noise cancellation system 100 within a vehicle compartment 130.

[0035] VM noise cancellation system 200 includes a reference sensor configured to acquire a reference signal related to noise within the vehicle compartment. In one example, the vehicle compartment may be vehicle compartment 130, and the reference sensor may be... Figure 1 The reference sensor 102 is located within the vehicle compartment 130. The VM noise cancellation system 200 includes: a plurality of speakers 206 positioned within the vehicle compartment 130, the plurality of speakers being configured to emit noise cancellation signals to cancel noise around the ears of one or more vehicle occupants; and a plurality of error microphones 208 positioned within the vehicle compartment 130, the plurality of error microphones being configured to acquire residual signals. In one example, the plurality of speakers 206 and the plurality of error microphones 208 may each include... Figure 1 The vehicle contains some or all of a plurality of loudspeakers 106 and a plurality of error microphones 108. Residual signals may include filtered noise signals present within the vehicle compartment 130. For example, physical microphone signals may include residual signals that are the product of filtering road noise using noise immunity signals generated by transducers such as loudspeakers 206.

[0036] The VM noise cancellation system 200 includes an adaptive weighted filter 204 that processes a reference signal obtained from a reference sensor 102 and applies an adaptive filtering algorithm to generate a noise-cancelled signal. The adaptive weighted filter 204 may be a time-domain adaptive weighted filter that dynamically adjusts based on a disclosed TFSVM algorithm to reduce residual signals.

[0037] Multiple error microphones 208 positioned within the vehicle compartment 130 can be configured to estimate the acoustic environment near a passenger's ears, regardless of their physical location or placement within the vehicle compartment 130. For example, multiple virtual microphones 208a can be modeled based on residual signals detected by the multiple physical error microphones 208b. The multiple physical error microphones 208b receive the residual signals, which are the resulting sound after filtering the interaction between the emitted noise-cancelled signal and the original noise using an adaptive weighted filter 204. The residual signals provide feedback on the performance of the VM noise cancellation system 200.

[0038] The VM noise cancellation system 200 includes a signal processing unit 110, a processor 112, and a non-transitory memory 114, as described in the aforementioned vehicle noise cancellation system 100, which together execute machine-readable instructions for using the TFSVM algorithm to cancel noise near the ears of one or more vehicle occupants. The signal processing unit 110 communicates electronically with a reference sensor 102, multiple loudspeakers 206, and multiple error microphones 208. In addition to executing the reference... Figure 1 In addition to the machine-executable instructions, processor 112 can also execute the TFSVM algorithm. Processor 112 processes data to adjust the adaptive weighted filter 204 corresponding to the noise cancellation signal to achieve targeted noise reduction near the ears of passengers in vehicle compartment 130.

[0039] The frequency-domain filtered signal module 230 is included in the signal processing unit 110. The frequency-domain filtered signal module 230 includes a frequency-domain transformation module 234, an overlap-preservation module 232, and multiple secondary path filters 236. The frequency-domain filtered signal module 230 uses a Fast Fourier Transform (FFT) process to transform the reference signal, noise-cancelled signal, and residual signal to the frequency domain, thereby applying an overlap-preservation method to mitigate the surround effect caused by cyclic correlation in the frequency domain. The multiple secondary path filters 236 may include frequency-domain physical secondary paths, frequency-domain virtual secondary paths, and frequency-domain virtual paths calculated or measured offline or online. For example, the multiple secondary path filters 236 may be calculated based on a sub-band adaptive filtering process or by other methods. The output of the frequency-domain filtered signal module 230 includes a frequency-domain filtered virtual microphone signal and a frequency-domain filtered reference signal.

[0040] A set of frequency-domain subband filters 216 in the signal processing unit 110 of the VM noise cancellation system 200 is used to decompose the frequency-domain filtered virtual microphone signal and the frequency-domain filtered reference signal into multiple subbands. Similar to the subband filter 116 described above in the reference vehicle noise cancellation system 100, the frequency-domain subband filters 216 are designed to divide the wide frequency range of the frequency-domain filtered virtual microphone signal and the frequency-domain filtered reference into narrower frequency bands. Each subband filter within the set corresponds to a different frequency range within the vehicle cabin noise spectrum. The design of the frequency-domain subband filters 216 includes multiple subband filters derived from a prototype filter (typically a low-pass filter with a specific window function). The window function in the prototype filter design can be adjusted based on application conditions. The choice of a window function such as Hamming or Kaiser can depend on the predetermined frequency response characteristics of each subband. The prototype filter is then modulated to create a series of bandpass filters spanning the entire frequency range of interest. The frequency range of each subband filter is equal to , It is the system sampling rate, and It refers to the number of sub-bands.

[0041] As referenced above Figure 1 As described above, the impulse response of each subband filter is obtained from the prototype filter, and the resolution and computational requirements of the subband filtering process can be adjusted by changing the length and number of subband filters. Through filter bank analysis, the frequency domain subband filter 216 is used for the frequency domain filtered virtual microphone signal and the frequency domain filtered reference signal, as described above and referenced. Figure 4 To describe in more detail: The output is multiple frequency-domain sub-band signals, each comprising the frequency content of the original signal within its respective sub-band. In this way, the time-domain adaptive filter is updated across the frequency range in each sub-band. Compared to full-band processing, this method reduces computational requirements and achieves greater noise reduction over a wide frequency range.

[0042] An adaptive step size normalization module 220 is included in the signal processing unit 110. The adaptive step size normalization module 220 adjusts the step size in the frequency domain on a per-subband basis. As the adaptive weighted filter 204 is updated and adapted to the vehicle noise environment, the adjustment affects the convergence rate and stability of the filter weights. The step size normalization module 220 is configured to... Figure 1 The adaptive step size determination module 120 is similar and is referenced below. Figure 4The method described in more detail dynamically modifies the step size in each subband based on the power contributions of the estimated virtual microphone signal and the reference signal within each corresponding subband. The resulting normalized step size is implemented for the weight transformation of the adaptive filter in each frequency domain subband to update the full-length adaptive weighted filter.

[0043] The gradient determination module 222 calculates the frequency domain sub-band gradient for each sub-band based on the filtered reference signal and the corresponding filtered error signal of the frequency domain sub-band. The output of the gradient determination module 222 is used to adjust the frequency domain sub-band adaptive filter in each sub-band to minimize residual signal.

[0044] The subband adaptive weight update module 224 updates the frequency domain subband adaptive filter in each subband based on the calculated gradient and the normalized step size. The subband adaptive weight update module 224 adapts the virtual microphone noise cancellation system 200 to the real-time noise conditions within the vehicle compartment 130.

[0045] The weighting transformation module 226 integrates the updated frequency-domain subband adaptive filter from each subband. The weighting transformation module uses inverse fast Fourier transform to pass the frequency-domain subband adaptive filter to the full-length adaptive weighted filter in the time domain.

[0046] Figure 3Plots 300, 310, 320, 330, 340, 350, 360, and 370 show simulation results comparing the performance of a conventional virtual microphone algorithm and the disclosed TFSVM algorithm in reducing noise at multiple virtual microphone locations. In each plot, frequencies in Hz are plotted on the x-axis, and sound pressure levels (SPL) in dB(A) are plotted on the y-axis. Plot 300 shows a first unfiltered noise signal 302 on a first virtual microphone, a first filtered signal 304 filtered by the conventional algorithm, and a first TFSVM-filtered signal 306 filtered by the disclosed TFSVM algorithm. Plot 310 shows a second unfiltered noise signal 312 on a second virtual microphone, a second filtered signal 314 filtered by the conventional algorithm, and a second TFSVM-filtered signal 316 filtered by the disclosed TFSVM algorithm. Figure 320 shows a third unfiltered noise signal 322, a third filtered signal 324 filtered using a conventional algorithm, and a third TFSVM-filtered signal 326 filtered using the disclosed TFSVM algorithm on a third virtual microphone. Figure 330 shows a fourth unfiltered noise signal 332, a fourth filtered signal 334 filtered using a conventional algorithm, and a fourth TFSVM-filtered signal 336 filtered using the disclosed TFSVM algorithm on a fourth virtual microphone. Figure 340 shows a fifth unfiltered noise signal 342, a fifth filtered signal 344 filtered using a conventional algorithm, and a fifth TFSVM-filtered signal 346 filtered using the disclosed TFSVM algorithm on a fifth virtual microphone. Figure 350 shows a sixth unfiltered noise signal 352, a sixth filtered signal 354 filtered using a conventional algorithm, and a sixth TFSVM-filtered signal 356 filtered using the disclosed TFSVM algorithm on a sixth virtual microphone. Figure 360 ​​shows the seventh unfiltered noise signal 362 on the seventh virtual microphone, the seventh filtered signal 364 filtered by a conventional algorithm, and the seventh TFSVM-filtered signal 366 filtered by the disclosed TFSVM algorithm. Figure 370 shows the eighth unfiltered noise signal 372 on the eighth virtual microphone, the eighth filtered signal 374 filtered by a conventional algorithm, and the eighth TFSVM-filtered signal 376 filtered by the disclosed TFSVM algorithm.

[0047] To enhance the overall performance and reliability of noise cancellation systems and virtual microphone technology, the greater the reduction from unfiltered to filtered noise signals, the better the system performance. As shown in Figures 300, 310, 320, 330, 340, 350, 360, and 370, the overall reduction in road noise produced by TFSVM filtering of unfiltered noise signals (e.g., TFSVM-filtered signals 306, 316, 326, 336, etc.) is greater than the reduction produced by conventional VM filtering methods. This is particularly evident in the high-frequency range, where the overall reduction at the virtual microphone location increases by 1.4 dB (above 290 Hz). In other words, the TFSVM algorithm reduces more road noise near the ears of vehicle occupants compared to conventional methods. The TFSVM algorithm achieves this improved performance while also reducing computational costs compared to the traditional time-frequency filtering X-Least Mean Square (FXLMS) algorithm.

[0048] Figure 4 This is a block diagram illustrating a Time-Frequency Subband Virtual Microphone (TFSVM) system 400. The TFSVM system 400 is based on a time-frequency domain subband adaptive filter structure that calculates the estimated virtual microphone signal in each subband and updates the adaptive filter individually for each frequency range. The TFSVM system 400 can be compared with a reference... Figure 2 The VM noise cancellation system 200 described herein is the same as or similar to that described above. In one example, the TFSVM system 400 may be used in a vehicle's noise cancellation system (such as reference 400). Figure 1 This is implemented in the vehicle noise cancellation system 100. In one example, the TFSVM system 400 can be used in a multiple-input multiple-output (MIMO) system for noise cancellation in a vehicle. The signal path is represented in the diagram by a line with an arrow indicating the direction of signal transmission.

[0049] The TFSVM system 400 includes: a reference sensor 402 configured to acquire a reference signal related to noise within a vehicle compartment; a plurality of loudspeakers 406 positioned within a vehicle compartment 404, configured to emit noise cancellation signals to cancel noise around the ears of one or more vehicle occupants; and a plurality of physical error microphones 408 positioned within the vehicle compartment 404, configured to acquire residual signals generated by filtering the reference signal using a time-domain adaptive weighting filter 401. The time-domain adaptive weighting filter 401 includes a full-length adaptive filter in the time domain, which is dynamically adjusted to reduce residual signals at the locations of the plurality of estimated virtual microphones within the vehicle compartment. The noise cancellation signal and the residual signal are also referred to herein as the loudspeaker output signal and the physical microphone signal, respectively.

[0050] In order to obtain the residual physical error microphone signal at block 409 The physical error microphone signal is expressed as follows:

[0051]

[0052]

[0053] in yes The residual physical microphone signal on the physical error microphone yes The primary signal on the error physical microphone, From Speaker to The actual acoustic impulse response of the secondary path of the physical error microphone. yes Reference signal, It is a full-length adaptive filter. It is the length of the adaptive filter, and It is a linear convolution operator. In other words, the residual signal is obtained by linearly convolving the primary signals on multiple physical error microphones, the actual impulse responses from multiple loudspeakers to multiple physical error microphones, and a full-length adaptive filter.

[0054] Conventionally, the time-domain VM method processes the estimation error microphone signal in the time domain. This process requires significant computational resources. Instead, the TFSVM system 400 uses a Fast Fourier Transform (FFT) process to calculate and estimate the virtual microphone signal. To avoid the loop-dependent surround effect in the FFT process, an overlap preservation method is applied. Therefore, the physical microphone signal block... and speaker output block The expression is as follows:

[0055]

[0056]

[0057] in yes The physical microphone signal on the physical microphone channel block vector, yes The speaker output signal on the speaker channel comes first. Add at each point Zero block vector, It is a block index. It is equal to the full length of the adaptive filter. The block size is ). In other words, the TFSVM system 400 can form a noise cancellation signal 2N block vector and a residual signal 2N block vector at blocks 410a and 410b by adding N zero blocks to the speaker output signal and the physical microphone signal, respectively.

[0058] Then, FFT is used to vectorize the physical microphone signal blocks. and speaker output block vector Transforming to the frequency domain, the calculation is as follows.

[0059]

[0060]

[0061] in yes The frequency domain physical microphone signal of the physical microphone channel, and yes The frequency domain speaker output signal of the speaker channel. In other words, the TFSVM system 400 applies FFT to the noise cancellation signal 2N block vector and the residual signal 2N block vector to generate the frequency domain noise cancellation signal and the frequency domain residual signal at blocks 412a and 412b, respectively, where N is the block size equal to the full length of the time-domain adaptive weighted filter 401.

[0062] To further reduce the computational load, frequency domain estimation of the physical secondary path is applied at blocks 414a, 414b, and 414c, respectively. Frequency domain estimation virtual secondary path And frequency domain estimation virtual path In one example, the secondary path filter can be computed or measured online or offline. For instance, the secondary path filter can be computed based on a subband adaptive filtering process, while in other examples, it can be computed based on a head-related transfer function process. At block 415, the secondary path filter is used to compute the frequency-domain filtered estimate of the virtual error microphone signal. The corresponding secondary path filtering equation is as follows:

[0063]

[0064]

[0065] in yes Frequency domain estimation error signal of physical microphone channel, yes Frequency domain estimation of virtual error signal for virtual microphone channel. yes The frequency domain speaker output signal of the speaker channel. From Physical error microphone to The frequency domain secondary path of the loudspeaker From Virtual error microphone to The speaker's frequency domain secondary path, and From Physical error microphone to The frequency domain secondary path of the virtual error microphone.

[0066] Reference signal block at block 403 The relevant frequency domain reference signal at block 413 is expressed as follows:

[0067]

[0068]

[0069] in yes Reference signal on reference channel block vector, yes The frequency domain reference signal of the reference channel. It is a block index. It is equal to the full length of the adaptive filter. The block size is N. In other words, the TFSVM system 400 can form a reference signal 2N block vector from the reference signal at block 403 at block 410c, and apply an FFT to the reference signal 2N block vector at block 412c to generate a frequency domain reference signal at block 413, where N is the block size equal to the full length of the time-domain adaptive weighted filter 401.

[0070] The frequency domain reference signal at block 413 is input into the following calculation to generate a frequency domain filtered reference signal.

[0071]

[0072] in yes Reference Channel Virtual microphone channels and The frequency domain of the speaker channel is filtered by the reference signal. From Virtual error microphone to The speaker's frequency domain estimation virtual secondary path. In other words, the TFSVM system 400 can apply a frequency domain virtual secondary path filter at block 414d to the frequency domain reference signal to generate a frequency domain filtered reference signal at block 417.

[0073] To reduce the complexity of the VM adaptive filtering process, a novel frequency domain subband processing is applied, which directly processes the signal from the full-band frequency domain to the subband frequency domain. To break down or segment the input signal into individual frequency domain subbands, a set of frequency domain subband analysis filters is used, where each subband represents a different frequency range. In one example, the set of frequency domain subband analysis filters includes a frequency domain filter bank 416. The frequency domain subband reference signal is calculated as follows...

[0074]

[0075]

[0076]

[0077]

[0078]

[0079] in, yes Reference Channel Virtual microphone channels and frequency domain of speaker channel Sub-band reference signal. yes Frequency domain of virtual error microphone channel Sub-band error signal It includes A zero-block vector, It is the length of the adaptive filter. It is the length of the subband adaptive filter. This refers to the number of sub-bands. In other words, the TFSVM system 400 can decompose the frequency-domain filtered virtual microphone signal at block 415 and the frequency-domain filtered reference signal at block 417 into multiple frequency-domain sub-band signals, wherein the multiple frequency-domain sub-band signals include multiple frequency-domain sub-band filtered reference signals and multiple frequency-domain filtered estimated virtual microphone signals.

[0080] In addition, based on the frequency domain subband filtered reference signal and frequency domain subband error signal Frequency domain subband gradient The expression is as follows:

[0081]

[0082] in yes Reference channels and Frequency domain on speaker channel Subband gradient, yes Reference Channel Virtual error microphone channel and Frequency domain on speaker channel The complex conjugate of the subband filtered reference signal. In other words, the frequency domain subband gradient calculation indicated by blocks 418a, 418b, and 418c involves performing a complex conjugate multiplication of the frequency domain subband filtered reference signal and the frequency domain subband error signal. The frequency domain subband adaptive filter in each subband is adjusted based on the frequency domain subband gradient calculation, which is part of the subband LMS process.

[0083] To adjust the convergence speed of the frequency domain subband adaptive filter and enhance the performance of the disclosed TFSVM process, a frequency domain subband flexible adaptive step size normalization is proposed. This normalization adjusts the step size based on the power contributions of the error microphone signal and the reference signal. The frequency domain subband flexible adaptive normalization step size... Calculate using the following equation.

[0084]

[0085]

[0086] in yes The sum of the filtered reference and error signals in the sub-band. yes In subband filters Reference channels and Speaker channel normalization step size yes Reference Channel Virtual microphone channels and frequency domain of speaker channel Sub-band filtered reference signal, It is the sum of the power contributions of the error signals. It is a smoothing parameter for the power of the filtered reference signal and error signal. This is the power contribution parameter representing the maximum power of the filtered reference signal and the error signal in each sub-band. The frequency domain sub-band flexible adaptive step size normalization process ensures that the signal levels of different sub-bands converge at a uniform rate.

[0087] Frequency domain subband adaptive filter weights The following equation is used for calculation and updating, based on the frequency domain subband gradient. and frequency domain normalization step size ,

[0088]

[0089] in yes Reference channels and frequency domain of speaker channel Subband adaptive filter yes Reference channels and frequency domain of speaker channel Subband normalization step size, and yes Subband leakage. In other words, the TFSVM system 400 can update the set of frequency domain subband adaptive filter weights indicated by blocks 420a, 420b, and 420c based on the frequency domain subband gradient and the frequency domain normalization step size to produce an updated set of frequency domain subband adaptive filter weights.

[0090] To obtain a full-length adaptive filter in the time domain The subband weight transformation process is implemented at block 422, which transforms the frequency domain subband adaptive filter. Passed to the full-length adaptive filter in the time domain .

[0091]

[0092]

[0093]

[0094]

[0095] in yes Reference channels and Full-band adaptive filter in the frequency domain for the speaker output channel. It is a frequency domain flip. The complex conjugate, yes Reference channels and Time-domain FIR filter for speaker channel, It is the length of the subband filter. It is the length of the fully adaptive filter, and It is a complex conjugate. In other words, the TFSVM system 400 uses the inverse fast Fourier transform (IFFT) at block 424 to transform the updated set of frequency-domain subband adaptive filter weights from the frequency domain to the time domain to produce an updated time-domain adaptive weighted filter. As part of the overlap-preservation method described above, in order to update the time-domain adaptive weighted filter based on the weighted transform of the multiple frequency-domain subband adaptive filters, the TFSVM system 400 is further configured to discard the last N-zero block from the output of the weighted transform of the multiple frequency-domain subband adaptive filters as indicated by block 426. As part of the overlap-preservation method, discarding the last N-zero block restores the signal to the original adaptive filter length. The VM noise cancellation system then transmits a noise cancellation signal based on the updated time-domain adaptive weighted filter to reduce noise around the ears of one or more vehicle occupants.

[0096] Figure 5A , Figure 5B and Figure 5C This is a flowchart depicting a method 500 for active noise cancellation incorporating the disclosed TFSVM algorithm. Method 500 may be similar to the method described in reference VM noise cancellation system 200, which may include... Figure 2 and Figure 1 In the vehicle noise cancellation system 100 described above, instructions for performing method 500 can be executed by a controller based on computer-readable instructions stored in the controller's memory and in conjunction with signals received from sensors of the vehicle system, such as signal processing unit 110, processor 112, non-transitory memory 114, reference sensor 102, multiple error microphones 208, and the aforementioned reference... Figure 1 and Figure 2 The aforementioned collection of frequency domain subband filters 216. According to the method described below, the controller can employ actuators of the vehicle system (such as multiple speakers 206) to adjust the operation of the vehicle system.

[0097] At 502, method 500 includes: acquiring a reference signal from a reference sensor. In one example, the reference signal corresponds to noise within the vehicle cabin. One or more reference sensors capable of detecting noise within the vehicle cabin may be present and placed within the vehicle cabin.

[0098] At 504, method 500 includes: processing a reference signal using a time-domain adaptive weighting filter to generate a noise-cancelled signal. The noise-cancelled signal (also referred to as a speaker output signal) may be emitted by multiple speakers or transducers within the vehicle cabin to attempt to cancel (e.g., reduce) sound near the ears of one or more vehicle occupants within the vehicle cabin. The characteristics of the time-domain adaptive weighting filter are iteratively updated to minimize residual signals detected at multiple physical microphones used to estimate noise at multiple virtual microphone locations. The residual signals may include ambient noise remaining after active noise cancellation of the reference signal has been performed using the time-domain adaptive weighting filter.

[0099] At 506, method 500 includes: estimating the frequency-domain filtered virtual microphone (VM) signal, as in method 500. Figure 5B Further details are provided below. Therefore, see [link to relevant documentation]. Figure 5B Step 506 of method 500 includes receiving a noise cancellation signal and a residual physical error microphone signal or a residual signal at 550.

[0100] At 552, method 500 includes applying an overlap-preservation method to the noise-cancelled signal and the residual signal. An example of applying the overlap-preservation method may include forming a 2N block vector of the noise-cancelled signal and a 2N block vector of the residual signal by adding N zero blocks to the noise-cancelled signal and the residual signal, respectively, where N is the block size equal to the full length of the time-domain adaptive weighted filter.

[0101] At 554, method 500 includes: applying a Fast Fourier Transform to the noise-cancelled signal 2N block vector and the residual signal 2N block vector. The noise-cancelled signal 2N block vector and the residual signal 2N block vector are transformed respectively to generate a frequency-domain (FD) noise-cancelled signal and a frequency-domain residual signal.

[0102] At 556, method 500 includes applying a plurality of secondary path filters to the frequency-domain noise cancellation signal and the frequency-domain residual signal to calculate a frequency-domain estimated virtual error microphone signal. The plurality of secondary path filters may include a frequency-domain estimated physical secondary path representing a frequency-domain secondary path from a plurality of physical error microphones to a loudspeaker. The frequency-domain estimated physical secondary path can be used to calculate the frequency-domain estimated error signal based on the frequency-domain noise cancellation signal and the frequency-domain physical microphone signal. The plurality of secondary path filters also include a frequency-domain estimated virtual secondary path representing a frequency-domain secondary path from a plurality of virtual error microphones to a plurality of loudspeakers, and a frequency-domain estimated virtual path representing a frequency-domain secondary path from a plurality of physical error microphones to a plurality of virtual error microphones. The frequency-domain estimated virtual secondary path filter and the frequency-domain estimated virtual path filter are used to calculate the frequency-domain estimated virtual error microphone signal based on the frequency-domain estimated error signal and the frequency-domain noise cancellation signal.

[0103] After step 556, Figure 5B The flowchart at step 508 returns Figure 5A .

[0104] At 508, method 500 includes: determining a frequency-domain filtered reference signal, as in method 500. Figure 5C Further details are provided below. Therefore, see [link to relevant documentation]. Figure 5C Step 508 of method 500 includes receiving a reference sensor signal at 560.

[0105] At 562, method 500 includes applying an overlap preservation method to the reference sensor signal. The overlap preservation method may include forming a 2N block vector of the reference signal, where N is the block size equal to the full length of the time-domain adaptive weighted filter.

[0106] At 564, method 500 includes: applying an FFT process to a 2N block vector of the reference signal. The 2N block vectors of the reference signal are then transformed to generate a frequency domain reference signal.

[0107] At position 566, method 500 includes: applying a frequency-domain estimated virtual secondary path filter to the frequency-domain reference signal. (See above reference...) Figure 5B The frequency domain estimation virtual secondary path represents the frequency domain secondary path from multiple virtual error microphones to multiple loudspeakers. The frequency domain estimation virtual secondary path filter is used to calculate the frequency domain filtered reference signal based on the frequency domain reference signal.

[0108] After step 556, Figure 5C The flowchart at step 510 can be returned to Figure 5A .

[0109] At 510, method 500 includes: applying a set of frequency-domain subband filters to decompose the frequency-domain filtered reference signal (full-bandwidth signal) into multiple frequency-domain subband filtered reference signals. As described above, the frequency-domain subband analysis filters separate the full-band frequency-domain filtered reference signal into individual subbands based on frequency. In some examples, low-pass and high-pass filters may be used to decompose the full-band frequency-domain filtered reference signal into multiple frequency-domain subband signals. An allowable error signal is associated with each subband to separate the full-band frequency-domain filtered reference signal into multiple frequency-domain subband signals.

[0110] At 512, method 500 includes: applying a set of frequency-domain subband filters to decompose the frequency-domain filtered estimated virtual microphone signal (full bandwidth signal) into multiple frequency-domain filtered estimated virtual microphone signals. The frequency-domain subband filters separate the frequency-domain filtered estimated virtual microphone signals into frequency subbands identical to those into which the frequency-domain filtered reference signal is separated, and this separation can be similarly accomplished using high-pass and low-pass filters. Using the same frequency subbands at 510 ensures that each subband reference signal has a corresponding subband error signal.

[0111] At 514, the process for determining the frequency domain subband adaptive filter for each individual subband is applied to each subband separately from the other subbands. Before the method continues through 514, the sub-method within 514 is applied to each subband created by the frequency domain subband adaptive filter at 510 and 512.

[0112] Within 514, at 516, method 500 includes: determining a frequency domain subband gradient based on a filtered reference signal of the frequency domain subband sharing the same subband and a frequency domain subband error signal. The subband gradient provides information about how the weights of the frequency domain adaptive weighted filter can be adjusted to reduce subband signal error.

[0113] At 518, method 500 includes: determining a frequency domain normalization step size. The normalization step size can be adjusted based on the power contributions of the frequency domain sub-band filtered reference signal and the frequency domain error signal according to a previously determined step size. The frequency domain normalization step size can affect the convergence rate and stability of the frequency domain adaptive filter weights, and balancing the normalization step size so that the adaptive filter weights converge quickly but do not change drastically between updates can be advantageous.

[0114] At point 520, method 500 includes determining the frequency-domain adaptive filter weights based on the normalized step size and gradient. The frequency-domain adaptive filter weights can be determined from a reference... Figure 4 The method described above updates the frequency-domain adaptive filter weights and may depend on previous iterations of the frequency-domain adaptive filter weights, the frequency-domain subband gradient, the frequency-domain normalization step size, and the leakage of the frequency-domain subband adaptive filter. Subband leakage can indicate the extent to which previous iterations of the frequency-domain subband adaptive filter affect the updated frequency-domain subband adaptive filter to reduce road noise effects.

[0115] At 522, method 500 includes: applying a subband weight transformation to each frequency domain subband adaptive filter. The subband weight transformation process integrates the updated adaptive filter weights from each subband to produce the final weights for the adaptive weighted filter in the time domain. The subband weight transformation process can be based on... Figure 4The method may be applied and may include using an inverse fast Fourier transform to transform each frequency domain subband adaptive filter to the time domain, and discarding the last N-zero block from the transformed filter as part of an overlap-preservation method.

[0116] At 524, the method further includes updating the weights of an adaptive weighted filter based on a sub-band adaptive filter with weight transformation. The weights can be updated to minimize residual signals detected at multiple physical error microphones and estimated at multiple virtual microphone locations.

[0117] In this way, compared with traditional filtering X-mean square algorithms or time-frequency FXLMS algorithms, the system and method for implementing the TFSVM algorithm disclosed in this paper achieve greater noise reduction over a wide frequency range by calculating the estimated virtual microphone signal in each sub-band and updating the adaptive filter individually for each frequency range. Furthermore, as disclosed, the method of flexible adaptive step size normalization for sub-bands increases the performance and convergence speed of the proposed TFSVM algorithm. The technical advantage of this application is that it enhances the performance of virtual microphone technology with less computational resource requirements.

[0118] This disclosure also provides support for a method for noise cancellation in a vehicle, the vehicle having: a reference sensor configured to acquire a reference signal related to noise within the vehicle cabin; a plurality of loudspeakers positioned within the vehicle cabin, the plurality of loudspeakers configured to emit a noise cancellation signal to cancel noise around the ears of one or more vehicle occupants; and a plurality of physical error microphones positioned within the vehicle cabin, the plurality of physical error microphones configured to acquire a residual signal, the method comprising: processing the reference signal using a time-domain adaptive weighted filter to generate a noise cancellation signal; estimating a frequency-domain filtered virtual microphone signal based on the noise cancellation signal and the residual signal, and estimating a frequency-domain filtered reference signal based on the reference signal; decomposing the frequency-domain filtered virtual microphone signal and the frequency-domain filtered reference signal into a plurality of frequency-domain sub-band signals; calculating a frequency-domain sub-band adaptive filter for each sub-band of the plurality of frequency-domain sub-band signals to generate a plurality of frequency-domain sub-band adaptive filters; and updating the time-domain adaptive weighted filter based on a weight transformation of the plurality of frequency-domain sub-band adaptive filters to generate an updated time-domain adaptive weighted filter. In a first example of the method, the method further includes applying an updated time-domain adaptive weighted filter to eliminate noise in the vehicle. In a second example of the method, optionally including the first example, estimating the frequency-domain filtered virtual microphone signal based on the noise-cancelled signal and the residual signal includes using a Fast Fourier Transform (FFT) to transform the noise-cancelled signal and the residual signal into a frequency-domain noise-cancelled signal and a frequency-domain residual signal, respectively. In a third example of the method, optionally including one or both of the first and second examples, transforming the noise-cancelled signal and the residual signal into a frequency-domain noise-cancelled signal and a frequency-domain residual signal further includes forming a 2N-block vector of the noise-cancelled signal and a 2N-block vector of the residual signal by adding N zero blocks to the noise-cancelled signal and the residual signal, respectively; and applying an FFT to the 2N-block vector of the noise-cancelled signal and the 2N-block vector of the residual signal, where N is the block size equal to the full length of the time-domain adaptive weighted filter. In a fourth example of the method, which optionally includes one or more or each of the first to third examples, estimating the frequency-domain filtered virtual microphone signal based on the noise-cancelled signal and the residual signal further includes applying a plurality of secondary path filters to the frequency-domain noise-cancelled signal and the frequency-domain residual signal, wherein the plurality of secondary path filters includes a frequency-domain physical secondary path, a frequency-domain virtual secondary path, and a frequency-domain virtual path. In a fifth example of the method, which optionally includes one or more or each of the first to fourth examples, estimating the frequency-domain filtered reference signal based on a reference signal includes: forming a 2N block vector of reference signal from the reference signal; applying a Fast Fourier Transform to the 2N block vector of reference signal to generate a frequency-domain reference signal, where N is the block size equal to the full length of the time-domain adaptive weighted filter; and applying a frequency-domain virtual secondary path filter to the frequency-domain reference signal.In a sixth example, which optionally includes one or more, or each of the methods described in the first to fifth examples, decomposing the frequency-domain filtered virtual microphone signal and the frequency-domain filtered reference signal into multiple frequency-domain sub-band signals includes: filtering the frequency-domain filtered virtual microphone signal and the frequency-domain filtered reference signal using a frequency-domain filter bank comprising a set of frequency-domain sub-band filters, each sub-band filter corresponding to a different frequency range within the vehicle cabin noise spectrum. In a seventh example, which optionally includes one or more, or each of the methods described in the first to sixth examples, calculating a frequency-domain sub-band adaptive filter for each of the multiple frequency-domain sub-band signals includes: determining a frequency-domain sub-band gradient for each sub-band based on the frequency-domain sub-band filtered reference signal and a frequency-domain sub-band error signal; and determining a frequency-domain normalization step size for each sub-band based on the power contribution of the frequency-domain sub-band filtered reference signal and the frequency-domain sub-band error signal. In an eighth example, which optionally includes one or more, or each of, the methods described in the first to seventh examples, determining the frequency domain subband gradient for each subband includes performing a complex conjugate multiplication of the frequency domain subband filtered reference signal and the frequency domain subband error signal. In a ninth example, which optionally includes one or more, or each of, the methods described in the first to eighth examples, the weight transformation includes updating the set of frequency domain subband adaptive filter weights based on the frequency domain subband gradient and the frequency domain normalization step size to produce an updated set of frequency domain subband adaptive filter weights; and using an inverse fast Fourier transform to transform the updated set of frequency domain subband adaptive filter weights from the frequency domain to the time domain to produce an updated time-domain adaptive weighted filter.

[0119] This disclosure also provides support for a noise cancellation system for a vehicle, the noise cancellation system comprising: a reference sensor configured to acquire a reference signal related to noise within the vehicle cabin; a time-domain adaptive weighted filter electronically communicating with the reference sensor, the time-domain adaptive weighted filter being configured to apply an adaptive filtering process to the reference signal to generate a noise cancellation signal, thereby eliminating noise around the ears of one or more vehicle occupants; a plurality of loudspeakers positioned within the vehicle cabin and electronically communicating with the time-domain adaptive weighted filter, the plurality of loudspeakers being configured to transmit the noise cancellation signal into the vehicle cabin; and a plurality of physical error microphones positioned within the vehicle cabin and configured to acquire residual noise resulting from the interaction between the noise cancellation signal and noise within the vehicle cabin. The signal; and a signal processing unit for electronic communication with a reference sensor, multiple loudspeakers, and multiple physical error microphones, wherein the signal processing unit includes: a non-transitory memory storing a set of frequency domain subband filters and instructions; and a processor, wherein when executing the instructions, the processor is configured to: estimate a frequency domain filtered virtual microphone signal based on a noise cancellation signal and a residual physical error microphone signal, and estimate a frequency domain filtered reference signal based on a reference signal; decompose the frequency domain filtered virtual microphone signal and the frequency domain filtered reference signal into multiple frequency domain subband signals; calculate a frequency domain subband adaptive filter for each subband of the multiple frequency domain subband signals to generate multiple frequency domain subband adaptive filters; and update a time-domain adaptive weighted filter based on a weighted transformation of the multiple frequency domain subband adaptive filters. In a first example of the system, to estimate a frequency-domain filtered virtual microphone signal based on a noise-cancelled signal and a residual signal, the processor is further configured to: form a 2N block vector of noise-cancelled signal and a 2N block vector of residual signal by adding N zeros to the noise-cancelled signal and the residual signal, respectively; apply a Fast Fourier Transform (FFT) to the 2N block vectors of noise-cancelled signal and the 2N block vectors of residual signal to generate a frequency-domain noise-cancelled signal and a frequency-domain residual signal, where N is the block size equal to the full length of the time-domain adaptive weighted filter; and apply a plurality of secondary path filters to the frequency-domain noise-cancelled signal and the frequency-domain residual signal. In a second example of the system, optionally including the first example, to estimate a frequency-domain filtered reference signal based on a reference signal, the processor is further configured to: form a 2N block vector of reference signal from the reference signal; apply an FFT to the 2N block vectors of reference signal to generate a frequency-domain reference signal, where N is the block size equal to the full length of the time-domain adaptive weighted filter; and apply a frequency-domain estimation virtual secondary path filter to the frequency-domain reference signal.In a third example of the system, optionally including one or both of the first and second examples, in order to update the time-domain adaptive weighted filter based on a weighted transform of multiple frequency-domain subband adaptive filters, the processor is further configured to discard the last N-zero block from the output of the weighted transform of the multiple frequency-domain subband adaptive filters. In a fourth example of the system, optionally including one or more of the first to third examples, the set of frequency-domain subband filters includes multiple subband filters derived from a prototype filter using high-pass and low-pass filters, each subband filter corresponding to a different frequency range within the vehicle cabin noise spectrum. In a fifth example of the system, optionally including one or more of the first to fourth examples, the noise cancellation system is a multi-input multi-output active noise cancellation system.

[0120] This disclosure also provides support for a method for noise cancellation in a vehicle, the method comprising: acquiring a reference signal using a reference sensor, wherein the reference signal is correlated with noise in a vehicle compartment; processing the reference signal using a time-domain adaptive weighted filter to generate a noise-cancelled signal; emitting the noise-cancelled signal using a plurality of loudspeakers positioned in the vehicle compartment to cancel noise around the ears of one or more vehicle occupants; acquiring a residual signal from a plurality of physical error microphones positioned in the vehicle compartment; transforming the noise-cancelled signal, the residual signal, and the reference signal into a frequency-domain noise-cancelled signal, a frequency-domain residual signal, and a frequency-domain reference signal, respectively, using a Fast Fourier Transform (FFT); applying a plurality of secondary path filters to the frequency-domain noise-cancelled signal and the frequency-domain residual signal to generate a frequency-domain filtered estimated virtual microphone signal, and applying a plurality of secondary path filters to the frequency-domain reference signal to generate a frequency-domain filtered reference signal; applying the frequency domain... A set of subband filters is used to decompose the frequency-domain filtered reference signal and the frequency-domain filtered estimated virtual microphone signal into multiple frequency-domain subband filtered reference signals and multiple frequency-domain filtered estimated virtual microphone signals, respectively; a frequency-domain subband gradient is determined for each subband based on the frequency-domain subband reference signal and the frequency-domain subband error signal; a frequency-domain normalization step size is determined for each subband based on the power contribution of the frequency-domain subband reference signal and the frequency-domain subband error signal; a set of frequency-domain subband adaptive filter weights is updated based on the frequency-domain subband gradient and the frequency-domain normalization step size to produce an updated set of frequency-domain subband adaptive filter weights; an inverse fast Fourier transform is used to transform the updated set of frequency-domain subband adaptive filter weights from the frequency domain to the time domain to produce an updated time-domain adaptive weighted filter; and a noise cancellation signal is emitted based on the updated time-domain adaptive weighted filter to reduce noise around the ears of one or more vehicle occupants. In a first example of the method, using FFT to transform the noise-cancelled signal, the residual signal, and the reference signal into a frequency-domain noise-cancelled signal, a frequency-domain residual signal, and a frequency-domain reference signal, respectively, further includes applying an overlap-preservation method to mitigate the wrap-around effect caused by cyclic correlation in the frequency domain. In a second example of the method, optionally including the first example, applying the overlap-preservation method includes: forming a 2N-block vector of the noise-cancelled signal and a 2N-block vector of the residual signal by adding N zero blocks to the noise-cancelled signal and the residual signal, respectively; forming a 2N-block vector of the reference signal; transforming the 2N-block vector of the noise-cancelled signal, the 2N-block vector of the residual signal, and the 2N-block vector of the reference signal to the frequency domain; and discarding the last N zero blocks from the updated time-domain adaptive weighting filter, where N is the block size equal to the full length of the time-domain adaptive weighting filter. In a third example of the method, optionally including one or both of the first and second examples, the plurality of secondary path filters include a frequency-domain estimated physical secondary path, a frequency-domain estimated virtual secondary path, and a frequency-domain estimated virtual path.

[0121] The description of the implementation scheme has been presented for illustrative and descriptive purposes. Suitable modifications and variations to the implementation scheme can be obtained by practicing the methods described above, either in light of the foregoing description or by means of practice. For example, unless otherwise stated, one or more of the described methods can be made by suitable equipment and / or combinations of equipment (such as those referenced in the document). Figure 1 and Figure 2 The method is performed by the vehicle noise cancellation system 100 and the VM noise cancellation system 200. The method can be performed by executing stored instructions using one or more logic devices (e.g., processors) in conjunction with one or more additional hardware elements, such as storage devices, memories, hardware network interfaces / antennas, switches, actuators, clock circuits, etc. The described method and associated actions can also be performed in various orders other than those described herein, in parallel, and / or simultaneously. Individual steps may be omitted in certain embodiments. The described systems are exemplary in nature and may include additional elements and / or omitted elements. The subject matter of this disclosure includes all novel and non-obvious combinations and sub-combinations of various systems and configurations, as well as all other features, functions, and / or properties disclosed. As used herein, unless otherwise specified, "about" is interpreted as meaning ±5% of that range.

[0122] As used in this application, an element or step described in the singular and followed by the words "an" or "a" should be understood as not excluding multiple said elements or steps, unless such exclusion is stated. Furthermore, references to "an embodiment" or "an example" in this disclosure are not intended to be construed as excluding the existence of additional embodiments also incorporated into the described features. The terms "first," "second," and "third," etc., are used merely as designations and are not intended to impose numerical requirements or a particular order of placement on their objects. The following claims specifically point to subject matter deemed novel and non-obvious from the foregoing disclosure.

Claims

1. A method for noise cancellation in a vehicle, the vehicle having: a reference sensor configured to acquire a reference signal related to noise within the vehicle cabin; Multiple speakers positioned within the vehicle cabin are configured to emit noise cancellation signals to eliminate noise around the ears of one or more vehicle occupants. The method includes: and multiple physical error microphones positioned within the vehicle compartment, the multiple physical error microphones being configured to acquire residual signals; The reference signal is processed using a time-domain adaptive weighted filter to generate the noise cancellation signal; The frequency-domain filtered virtual microphone signal is estimated based on the noise cancellation signal and the residual signal, and the frequency-domain filtered reference signal is estimated based on the reference signal; The frequency-domain filtered virtual microphone signal and the frequency-domain filtered reference signal are decomposed into multiple frequency-domain sub-band signals; For each subband of the plurality of frequency domain subband signals, a frequency domain subband adaptive filter is calculated to generate a plurality of frequency domain subband adaptive filters; and The time-domain adaptive weighted filter is updated based on the weight transformation of the plurality of frequency-domain sub-band adaptive filters to generate an updated time-domain adaptive weighted filter.

2. The method of claim 1, further comprising: The updated time-domain adaptive weighted filter is applied to eliminate noise in the vehicle.

3. The method of claim 1, wherein estimating the frequency-domain filtered virtual microphone signal based on the noise cancellation signal and the residual signal comprises: The noise-cancelled signal and the residual signal are transformed into frequency-domain noise-cancelled signal and frequency-domain residual signal, respectively, using Fast Fourier Transform (FFT).

4. The method of claim 3, wherein transforming the noise-cancelled signal and the residual signal into the frequency-domain noise-cancelled signal and the frequency-domain residual signal further comprises: A 2N-block vector of noise cancellation signal and a 2N-block vector of residual signal are formed by adding N zeros to the noise cancellation signal and the residual signal, respectively. as well as Apply the FFT to the 2N block vectors of the noise-cancelled signal and the 2N block vectors of the residual signal. Where N is the block size equal to the full length of the time-domain adaptive weighted filter.

5. The method of claim 3, wherein estimating the frequency-domain filtered virtual microphone signal based on the noise cancellation signal and the residual signal further comprises: Multiple secondary path filters are applied to the frequency domain noise-cancelled signal and the frequency domain residual signal, wherein the multiple secondary path filters include frequency domain physical secondary paths, frequency domain virtual secondary paths, and frequency domain virtual paths.

6. The method of claim 1, wherein estimating the frequency-domain filtered reference signal based on the reference signal comprises: The reference signal forms a 2N block vector of reference signals; Apply a Fast Fourier Transform to the 2N block vectors of the reference signal to generate a frequency domain reference signal, where N is the block size equal to the full length of the time-domain adaptive weighted filter; as well as A frequency-domain virtual secondary path filter is applied to the frequency-domain reference signal.

7. The method of claim 1, wherein decomposing the frequency-domain filtered virtual microphone signal and the frequency-domain filtered reference signal into the plurality of frequency-domain sub-band signals comprises: The frequency-domain filtered virtual microphone signal and the frequency-domain filtered reference signal are filtered by a frequency-domain filter bank comprising a set of frequency-domain sub-band filters, each sub-band filter corresponding to a different frequency range within the vehicle cabin noise spectrum.

8. The method of claim 1, wherein calculating the frequency domain subband adaptive filter for each subband of the plurality of frequency domain subband signals comprises: The frequency domain subband gradient is determined for each subband based on the filtered reference signal and the frequency domain subband error signal; and the frequency domain normalization step size is determined for each subband based on the power contribution of the filtered reference signal and the frequency domain subband error signal. Determining the frequency domain sub-band gradient for each sub-band includes performing a complex conjugate multiplication of the filtered reference signal and the error signal of the frequency domain sub-band. The weight transformation includes: updating the set of frequency domain subband adaptive filter weights based on the frequency domain subband gradient and the frequency domain normalization step size to generate an updated set of frequency domain subband adaptive filter weights; and using inverse fast Fourier transform to transform the updated set of frequency domain subband adaptive filter weights from the frequency domain to the time domain to generate the updated time domain adaptive weighted filter.

9. A noise cancellation system for a vehicle, the noise cancellation system comprising: A reference sensor, configured to acquire a reference signal related to noise within the vehicle cabin; A time-domain adaptive weighted filter that communicates electronically with the reference sensor is configured to apply an adaptive filtering process to the reference signal to generate a noise cancellation signal, thereby eliminating noise around the ears of one or more vehicle occupants. Multiple loudspeakers located within the vehicle compartment and in electronic communication with the time-domain adaptive weighted filter are configured to transmit the noise cancellation signal into the vehicle compartment. Multiple physical error microphones are positioned within the vehicle compartment and configured to acquire residual signals resulting from the interaction between the noise cancellation signal and the noise within the vehicle compartment. as well as A signal processing unit that electronically communicates with the reference sensor, the plurality of loudspeakers, and the plurality of physical error microphones, wherein the signal processing unit includes: Non-transitory memory, which stores a set of frequency domain subband filters and instructions; and A processor, wherein when the instructions are executed, the processor is configured to: The frequency domain filtered virtual microphone signal is estimated based on the noise cancellation signal and the residual physical error microphone signal, and the frequency domain filtered reference signal is estimated based on the reference signal. The frequency-domain filtered virtual microphone signal and the frequency-domain filtered reference signal are decomposed into multiple frequency-domain sub-band signals; For each subband of the plurality of frequency domain subband signals, a frequency domain subband adaptive filter is calculated to generate a plurality of frequency domain subband adaptive filters; and The time-domain adaptive weighted filter is updated based on the weighted transformation of the multiple frequency-domain sub-band adaptive filters.

10. The noise cancellation system of claim 9, wherein, in order to estimate the frequency-domain filtered virtual microphone signal based on the noise cancellation signal and the residual signal, the processor is further configured to: A 2N-block vector of noise cancellation signal and a 2N-block vector of residual signal are formed by adding N zeros to the noise cancellation signal and the residual signal, respectively. Apply Fast Fourier Transform (FFT) to the 2N block vectors of the noise-cancelled signal and the 2N block vectors of the residual signal to generate a frequency-domain noise-cancelled signal and a frequency-domain residual signal, where N is the block size equal to the full length of the time-domain adaptive weighted filter; as well as Multiple secondary path filters are applied to the frequency domain noise-cancelled signal and the frequency domain residual signal. In order to estimate the frequency-domain filtered reference signal based on the reference signal, the processor is further configured to: The reference signal forms a 2N block vector of reference signals; The FFT is applied to the 2N block vector of the reference signal to generate a frequency domain reference signal, where N is the block size equal to the full length of the time-domain adaptive weighted filter; as well as A frequency-domain estimation virtual secondary path filter is applied to the frequency-domain reference signal. In order to update the time-domain adaptive weighted filter based on the weighted transform of the plurality of frequency-domain subband adaptive filters, the processor is further configured to discard the last N-zero block from the output of the weighted transform of the plurality of frequency-domain subband adaptive filters.

11. The noise cancellation system of claim 9, wherein the set of frequency domain subband filters comprises a plurality of subband filters derived from a prototype filter using high-pass and low-pass filters, each subband filter corresponding to a different frequency range within the vehicle cabin noise spectrum.

12. The noise cancellation system of claim 9, wherein the noise cancellation system is a multiple-input multiple-output active noise cancellation system.

13. A method for noise cancellation in a vehicle, the method comprising: A reference sensor is used to acquire a reference signal, wherein the reference signal is correlated with the noise in the vehicle cabin; The reference signal is processed using a time-domain adaptive weighted filter to generate a noise-cancelled signal; Multiple speakers positioned in the vehicle compartment are used to emit the noise cancellation signal to eliminate noise around the ears of one or more vehicle occupants; Residual signals are acquired from multiple physical error microphones located in the vehicle compartment; The noise-cancelled signal, the residual signal, and the reference signal are transformed into a frequency-domain noise-cancelled signal, a frequency-domain residual signal, and a frequency-domain reference signal, respectively, using a Fast Fourier Transform (FFT). Multiple secondary path filters are applied to the frequency domain noise-cancelled signal and the frequency domain residual signal to generate a frequency domain filtered estimated virtual microphone signal, and the multiple secondary path filters are applied to the frequency domain reference signal to generate a frequency domain filtered reference signal. A set of frequency domain sub-band filters is used to decompose the frequency domain filtered reference signal and the frequency domain filtered estimated virtual microphone signal into multiple frequency domain sub-band filtered reference signals and multiple frequency domain filtered estimated virtual microphone signals, respectively. The frequency domain subband gradient is determined for each subband based on the frequency domain subband reference signal and the frequency domain subband error signal. The frequency domain normalization step size is determined for each sub-band based on the power contribution of the frequency domain sub-band reference signal and the frequency domain sub-band error signal. The set of frequency domain subband adaptive filter weights is updated based on the frequency domain subband gradient and the frequency domain normalization step size to generate the updated set of frequency domain subband adaptive filter weights. The updated set of frequency domain subband adaptive filter weights is transformed from the frequency domain to the time domain using the inverse fast Fourier transform to produce an updated time-domain adaptive weighted filter. as well as The noise cancellation signal is emitted based on the updated time-domain adaptive weighted filter to reduce noise around the ears of the one or more vehicle occupants.

14. The method of claim 13, wherein using FFT to transform the noise-cancelled signal, the residual signal, and the reference signal into the frequency-domain noise-cancelled signal, the frequency-domain residual signal, and the frequency-domain reference signal, respectively, further comprises: An overlap-based preservation method is applied to mitigate the wraparound effect caused by cyclic correlation in the frequency domain. The overlapping preservation method includes: forming a 2N block vector of noise-cancelled signal and a 2N block vector of residual signal by adding N zeros to the noise-cancelled signal and the residual signal respectively; forming a 2N block vector of reference signal; transforming the 2N block vector of noise-cancelled signal, the 2N block vector of residual signal and the 2N block vector of reference signal to the frequency domain; and discarding the last N zeros from the updated time-domain adaptive weighting filter, where N is the block size equal to the full length of the time-domain adaptive weighting filter.

15. The method of claim 13, wherein the plurality of secondary path filters includes frequency domain estimated physical secondary paths, frequency domain estimated virtual secondary paths, and frequency domain estimated virtual paths.

Images