Method of generating imperceptible stimulus signal for automotive active noise control system

Abstract

A system and method for generating a stimulation signal as a solution for an adaptive secondary path component of an active noise cancellation (ANC) system, the system having: at least one audio signal source for providing an audio signal in a cabin of the vehicle; a shaped noise generator that provides a shaped noise signal in the cabin of the vehicle; at least one controller programmed to activate the shaped noise generator and add the shaped noise signal to the audio signal to generate a stimulation signal; and at least one speaker for projecting the stimulation signal within the cabin of the vehicle to cause the ANC system to generate an anti-noise signal.

Description

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS

[0002] The entire disclosure of U.S. non-provisional application serial number 17 / 975,782, filed on October 28, 2022, entitled “SYSTEM AND METHOD FOR ESTIMATING SECONDARY PATH IMPULSE RESPONSE FOR ACTIVE NOISE CANCELLATION,” and U.S. non-provisional application serial number 17 / 976,048, filed on October 28, 2022, entitled “SYSTEM AND METHOD FOR SECONDARY PATH SWITCHING FOR ACTIVE NOISE CANCELLATION,” are hereby incorporated by reference into the present disclosure. Technical Field

[0003] Various aspects disclosed herein generally relate to a system and method for generating an imperceptible stimulation signal for online adaptive secondary path switching of an active noise control (ANC) system. These and other aspects will be discussed in greater detail herein. Background Art

[0004] Active noise cancellation (ANC) systems use feed-forward and feedback structures to attenuate unwanted noise in order to adaptively remove unwanted noise within a listening environment, such as within a vehicle cabin. ANC systems cancel or reduce unwanted noise by generating cancellation sound waves to destructively interfere with the unwanted audible noise. ANC systems implemented on a vehicle that minimize noise within a vehicle cabin include road noise cancellation (RNC) systems that minimize unwanted road noise and engine order noise cancellation (EOC) systems that minimize unwanted engine noise within a vehicle cabin.

[0005] ANC systems typically use digital signal processing and digital filtering techniques. For example, a noise sensor such as a microphone, accelerometer, or revolutions per minute (RPM) sensor outputs an electrical reference signal that represents the interfering noise signal generated by the noise source. This reference signal is fed to an adaptive filter. The filtered reference signal is then supplied to an acoustic actuator (e.g., a speaker) that generates a compensating sound field that ideally has an opposite phase and nearly the same amplitude as the noise signal. This compensating sound field eliminates or reduces the noise signal within the vehicle cabin.

[0006] An RNC system is a specific ANC system implemented on a vehicle to minimize unwanted road noise inside the vehicle cabin. An RNC system uses vibration sensors to sense road-induced vibrations generated from the tire-road interface, which can cause unwanted audible road noise. Eliminating this road noise can provide a more pleasant ride for vehicle passengers and enable automakers to use lightweight materials, thereby reducing energy consumption and emissions. An EOC system is a specific ANC system implemented on a vehicle to minimize unwanted engine noise inside the vehicle cabin. An EOC system uses non-acoustic sensors such as an engine speed sensor to generate a signal representing the rotational speed of the engine crankshaft in revolutions per minute (RPM) as a reference. RNC systems are generally designed to eliminate broadband signals, while EOC systems are designed and optimized to eliminate narrowband signals, such as individual engine order noises. An ANC system within a vehicle can provide both RNC and EOC technologies.

[0007] The residual noise signal can be measured using, for example, a microphone to provide an error signal to an adaptive unit of an adaptive filter, wherein filter coefficients (also referred to as parameters) of the adaptive filter are modified to produce a norm of the error signal. The adaptive unit of the adaptive filter can use digital signal processing methods, such as least mean square (LMS), filtered x least mean square (FxLMS), improved filtered x least mean square (MFxLMS) or other techniques to reduce the error signal.

[0008] When applying many variants of the LMS algorithm, such as the FxLMS and MFxLMS algorithms, an estimated model representing the acoustic transmission path from the loudspeaker to the microphone is used. This acoustic transmission path is often referred to as the secondary path of the ANC system. In contrast, the acoustic transmission path from the noise source to the microphone is often referred to as the primary path of the ANC system. The secondary path transfer function represented in the time domain is often referred to as an impulse response or IR.

[0009] The way the estimated secondary path transfer function matches the actual secondary path transfer function affects the stability of the ANC system. A changing secondary path transfer function can negatively impact the ANC system because the actual secondary path transfer function no longer matches the "a priori" estimated secondary path transfer function used in the FxLMS or MFxLMS algorithm when it changes. The estimated model of the secondary path is typically measured once and approximated during the production tuning process, where the secondary path transfer function is estimated for a "nominal" acoustic scenario (i.e., one occupant, windows closed, seats in default positions). However, the secondary path can change for a number of different reasons, such as changes in the number of occupants, seat position, items in the listening environment, etc. These differences between the stored, estimated secondary path and the actual secondary path can result in insufficient noise cancellation system performance and can even cause the adaptive filter to diverge, which results in the generation of undesirable noise in the listening environment, commonly referred to as noise enhancement.

[0010] A known solution is that the ANC system selects the closest impulse response from a set of reference impulse responses based on online (real-time) measurements using decorrelated audio signals. This component of the ANC is called an Adaptive Secondary Path (ASP) system. ASP uses an audio signal from the audio, whose content covers the frequency range of interest to typical noise cancellation algorithms, such as 20Hz to 1000Hz. In addition, the audio signal received at the microphone must have sufficient sound pressure level for the adaptive filter to converge. ASP requires each speaker to emit a decorrelated signal at the same time, and usually uses audio (such as music) as the source of the decorrelated signal so that the decorrelated signal has a sufficient level while not being perceived by the occupants.

[0011] In many cases, the ASP system is active, but the audio level is below the level required to receive the audio signal at the microphone at a sufficient sound pressure level that will cause the adaptive filter to converge. Moreover, when this occurs, the ASP will not function properly. Examples of audio levels below the necessary level may occur, for example, when the occupant turns off the audio, listens to low-volume audio, listens to a listener hotline, communicates with a voice assistant, participates in a voice conversation in other media forms, and participates in a telephone conversation.

[0012] Therefore, a method for generating multi-channel sound for ASP is needed. The stimulus signal is less perceptible to the occupants in the vehicle. Summary of the invention

[0013] The inventive subject matter is a system and method for generating multi-channel sounds used by an ASP when learning a secondary path. The multi-channel sounds used by the ASP are multiple stimulus signals, where a stimulus signal is generated for each speaker, such as in a vehicle audio system. When the audio level is insufficient to learn the secondary path, the stimulus signals are less perceptible to the occupants. This is accomplished by generating shaped noise to play a unique signal on the vehicle's audio system. The shaped noise is added to an alternate audio source signal so that each unique stimulus signal is generated for the ASP system with content sufficient to cause the adaptive filter in the ASP to converge.

[0014] The inventive subject matter includes a feedback system for adjusting noise volume based on vehicle operation and a method for generating shaped noise to be used in an ASP system.

[0015] A system and method for generating a stimulus signal for an adaptive secondary path of an active noise cancellation (ANC) system in a vehicle is described below. The system and method have: at least one audio signal source for providing an audio signal in a cabin of the vehicle; a shaped noise generator for providing a shaped noise signal in the cabin of the vehicle; at least one controller programmed to activate the shaped noise generator and add the shaped noise signal to the audio signal to generate a stimulus signal; and at least one speaker for projecting the stimulus signal within the cabin of the vehicle to cause the ANC system to generate an anti-noise signal.

[0016] In one or more embodiments, the microphone provides a signal indicative of the noise level in the cabin, and a gain scheduling component adjusts the gain of the shaped noise signal to psychoacoustically mask the shaped noise signal and maintain a constant signal-to-noise ratio (SNR). The gain-adjusted shaped noise signal is added to the audio signal to generate the stimulus signal.

[0017] In one or more embodiments, the shaped noise generator is a pre-recorded signal.

[0018] In one or more embodiments, the pre-recorded signal is a cyclic masking signal.

[0019] In one or more embodiments, the pre-recorded signal is a horn of the vehicle.

[0020] In one or more embodiments, the shaped noise generator has an active sound design (ASD) system.

[0021] In one or more embodiments, the linear predictive coding component applies a set of predetermined LPC synthesis filters to generate the stimulation signal.

[0022] In one or more embodiments, the linear predictive coding component uses input from an error microphone and updates the LPC synthesis filter in real time. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 is a block diagram of an active noise control (ANC) system having an adaptive secondary path (ASP) system;

[0024] Figure 2 is a block diagram of one or more embodiments of a shaped noise generator;

[0025] Figure 3 is a block diagram of one or more embodiments of a shaped noise generator;

[0026] Figure 4 is a block diagram of one or more embodiments of a shaped noise generator;

[0027] Figure 5 is based on Figure 4 Block diagram of LPC analysis using a shaped noise generator as shown in;

[0028] Figure 6 is a block diagram of one or more embodiments of a shaped noise generator; and

[0029] Figure 7 is a flow chart of a method for performing secondary path switching using IR fingerprint analysis according to one embodiment. DETAILED DESCRIPTION

[0030] Figure 1 An overall block diagram 100 of active noise control (ANC) with an adaptive secondary path (ASP) for use in a vehicle cabin 102 is generally depicted. The system 100 includes at least one controller and a memory. In accordance with the inventive subject matter, a shaped noise generator 104 generates a shaped noise signal 104 that is supplied to an adaptive secondary path (ASP) system 106. The ASP 106 and its operation are not described herein, but are described in detail in U.S. Non-Provisional Application Serial No. 17 / 975,782, entitled “SYSTEM AND METHOD FOR ESTIMATING SECONDARY PATH IMPULSE RESPONSE FOR ACTIVE NOISE CANCELLATION,” filed on October 28, 2022, and U.S. Non-Provisional Application Serial No. 17 / 976,048, entitled “SYSTEM AND METHOD FOR SECONDARY PATH SWITCHING FOR ACTIVE NOISE CANCELLATION,” both of which are incorporated by reference into this disclosure. The inventive subject matter provides a solution for using an ASP when there is a lack of audio with sufficient content to cause the adaptive filter in the ASP system to converge.

[0031] It should be recognized that the system includes a controller disclosed herein, and the method is performed by a controller disclosed herein. The controller may include various microprocessors, integrated circuits, memory devices (e.g., FLASH, random access memory (RAM), read-only memory (ROM), electrically programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM) or other suitable variations thereof) and software that cooperate with each other to perform the (multiple) steps disclosed herein. In addition, such controllers as disclosed utilize one or more microprocessors to execute computer programs embodied in non-transitory computer-readable media that are programmed to perform any number of functions as disclosed. In addition, the (multiple) controllers as provided herein include a housing and various numbers of microprocessors, integrated circuits, and memory devices (e.g., FLASH, random access memory (RAM), read-only memory (ROM), electrically programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM)) positioned within the housing. The disclosed (multiple) controllers also include hardware-based inputs and outputs, which are used to receive data from and transmit data to other hardware-based devices as discussed herein, respectively.

[0032] The backup audio source 108 provides an audio signal 110 that may include typical entertainment or speech audio. The audio signal 110 should include content that spans the frequency range of interest to typical noise cancellation algorithms and have sufficient sound pressure levels to allow the adaptive filter of the ASP system 106 to converge. The noise level within the vehicle cabin 102 is monitored by a microphone 112 in the vehicle cabin 102, and the ANC system 100 will determine when the ASP system 106 needs to be operated, and the ANC system 100 will enable or disable the shaped noise generator 104 as needed. The microphone 112 measures the error signal used by the ANC system, the gain scheduling component 120, and the shaped noise generator 104. It is worth noting that although one microphone 112 is shown, there are typically multiple microphones 112 throughout the vehicle cabin 102. For example, at each seat position within the vehicle cabin 102, but for simplicity, only one microphone is shown. Similarly, the method described below generates a unique shaped noise signal 114 for each seat position within the vehicle cabin. However, for simplicity, a single step or a single element is described.

[0033] For situations where ASP is active but audio signal 110 lacks audio whose content meets ASP requirements, ANC system 100 enables shaped noise generator 104 so that shaped noise generator 104 provides shaped noise signal 114 which is gain adjusted 116 and then added 118 to audio signal 110 .

[0034] Based on the noise level detected by the microphone 112 in the vehicle cabin 102, the ASP system 106 will instruct the gain scheduling component 120 to adjust 116 the gain of the shaped noise signal 114 output by the shaped noise generator 104 to generate a gain-adjusted shaped noise signal 126. The gain-adjusted shaped noise signal 126, when combined with the audio signal 110, generates a stimulus signal 122 sufficient to enable the ASP system 106 to operate efficiently. As described in detail in U.S. non-provisional application serial number 17 / 975,782, entitled "System and Method for Estimating Secondary Path Impulse Response for Active Noise Cancellation," filed on October 28, 2022, the stimulus signal 122 is processed to ensure that the signal is decorrelated before being added to the corresponding ANC anti-noise signal 128 output to one or more speakers 124 in the vehicle cabin 102.

[0035] The shaped noise generator 104 can be implemented in one of several embodiments described in detail later herein. The gain scheduling component 120 increases or decreases the volume of the shaped noise signal 114 with the goal of maintaining a constant signal-to-noise ratio (SNR). For example, where a constant 10 dB SNR is desired and the vehicle cabin 102 has a 40 dBA sound pressure level (SPL), the output of the shaped noise generator 104 can be gained up to 50 dBA SPL.

[0036] The gain scheduling component 120 also seeks to achieve one or more psychoacoustic masking requirements to ensure that the stimulus signal 122 is not perceived by the vehicle occupants. For example, if the primary engine order noise is 50 Hz, the gain of the output of the shaped noise generator 104 can be increased within the critical bandwidth of the 50 Hz tone masking. The calculation of the gain can be accomplished using any one or more known methods. For example, for a single-input single-output (SISO) system, the gain can be calculated as follows:

[0037] in,

[0038] P e [n] = λP e [n-1]+(1-λ)e 2 [n],

[0039]

[0040] Extension to multiple-input multiple-output (MIMO) systems means that for microphone m, the gain can be calculated as follows:

[0041]

[0042] G final [n] = max(G m[n])

[0043] where exponential smoothing is applied to the final gains.

[0044] For any of the methods for generating shaped noise described later herein, the gain adjusted shaped noise signal 126 and the audio signal 110 are combined to generate a stimulus signal 122. The stimulus signal 122 is played through the vehicle's audio system and has content sufficient to converge the adaptive filter in the ASP 106. The stimulus signal 122 is constructed or masked so that it is not perceived by vehicle occupants when it is played through the vehicle's audio system during normal vehicle operation.

[0045] One feature of the stimulus signal 122 is that it has a limited bandwidth, which makes it almost imperceptible. Human hearing is less sensitive to lower frequencies. Therefore, the bandwidth of the stimulus signal 122 is limited to frequencies below what the human ear can hear.

[0046] In one or more embodiments, the stimulus signal is barely perceptible because the gain scheduling component 120 adjusts 116 the gain of the shaped noise signal 114 based on the noise level in the vehicle cabin 102. In one or more embodiments, through spectral shaping, the stimulus signal 122 can be kept at a frequency near the tonal noise, thereby being masked by the tonal noise and thus not perceptible to the vehicle occupants. In one or more embodiments, the sounds before and after the stimulus signal are masked through temporal gain scheduling so that the stimulus signal is not perceptible to the human ear.

[0047] refer to Figure 2 , one or more embodiments of the shaped noise generator 104 are shown in block diagram 200. The shaped noise signal 114 may be a pre-recorded signal 202, which may be in a compressed format or an uncompressed format depending on available memory. For example, but not limited to, a wave file (.WAV or .WAVE), which is a digital audio file encoded in a standard audio format. The pre-recorded signal 202 may be stored in memory, and when the ASP system 106 is active and the audio signal 110 from the backup audio source 108 is deemed insufficient by the ANC, the ASP activates the shaped noise generator to output the pre-recorded signal 202 as the shaped noise signal 114. If desired, the gain scheduling component 120 may modify the shaped noise signal 114 to generate a gain shaped noise signal 126, which is added 118 to the audio signal 110 to generate the stimulus signal 122.

[0048] The pre-recorded signal 202 may be a predetermined looping file, i.e., a masked signal. Additionally or alternatively, the pre-recorded signal 202 may be a vehicle chime that is played from time to time during normal vehicle operation, such as a welcome sound. Regardless, the pre-recorded signal 202 is a predetermined sound that is designed to meet all of the characteristics required of the stimulus signal 122 in order for the ASP system 106 to operate effectively while remaining unnoticed by vehicle occupants.

[0049] refer to Figure 3 , one or more embodiments of the shaped noise generator 104 are shown in block diagram 300. The shaped noise signal 114 may be generated by an active sound design (ASD) system 302. ASD is a method of synthesizing combustion or electric vehicle powertrain sounds. It may include one of several additive synthesis techniques and / or a looping waveform player. Since the ASD system 302 is an existing in-cabin sound generator, there are no additional requirements or costs associated with using an ASD as a shaped noise generator for generating the shaped noise signal 114. The ASD filters inputs from engine and vehicle speed, pedal input, exhaust noise, and vehicle vibration to produce the desired output. Since the sound generated by the ASD system 302 produces an audio that is already integrated into the normal sound of the vehicle cabin, the SNR of the shaped noise signal 114 produced by the ASD system 302 can advantageously have a higher SNR. The ASD output is designed to meet all the characteristics required of the stimulus signal 122 to enable the ASP system 106 to operate effectively while remaining unnoticed by the vehicle occupants.

[0050] refer to Figure 4 , one or more embodiments of the shaped noise generator 104 are shown in block diagram 400 . Figure 4 The system shown can be used with Figure 3 The ASP system 302 works in conjunction with the ASD system to generate the shaped noise signal 114. The ASP system activates the white noise generator 402, whose output signal 404 is fed to the ASD system 302. A residual noise signal 406 (the current noise in the vehicle measured by the error microphone 112 of the ANC system) is also fed to the ASD system 302. The linear predictive coding (LPC) component 410 of the ASD system 302 uses the residual noise signal 406 to fit the frequency domain response to the residual noise. Using the white noise generator output signal 404 and the output of the LPC component 410, the ASD system 302 estimates the residual noise spectrum 412.

[0051] The psychoacoustic masking component 414 applies a psychoacoustic masking curve to the estimated residual noise spectrum 412 , and the output is passed through an inverse secondary path filter 416 to generate the shaped noise signal 114 .

[0052] refer to Figure 5, a block diagram 500 of the LPC 410 is shown in more detail. The residual noise signal 406 from the error microphone 112 is downsampled 502 or decimated to a frequency of 1.5 kHz. An overlap analysis 504 is performed to eliminate duplicate data. Smoothing 506 is performed via a Hanning window. An autocorrelation 508 is performed, and a Levinson-Dubin algorithm 510 with a predetermined LPC order 512 is applied. The processed signal is synthesized 514 with the output 404 of the white noise generator 402, and the synthesized signal 516 is upsampled 518 to 48 kHz.

[0053] refer to Figure 5 One or more embodiments described provide an online update for the LPC synthesis filter. In one or more embodiments, Figure 6 As shown, the shaped noise generator 104 is shown in block diagram 600, wherein the LPC synthesis filter (such as Figure 5 described) is tuned off-line. Figure 6 , an LPC filter 602 is applied to the output 404 of the white noise generator 402. The LPC filter 602 is predetermined, stored in memory, and selected by the ASP system 106 based on an ASP 106 input representing a vehicle characteristic, such as engine RPM or vehicle speed.

[0054] refer to Figure 7 , a method 700 for performing secondary path switching using IR fingerprint analysis is shown, as described in U.S. Non-Provisional Application Serial No. 17 / 976,048, filed on October 28, 2022, entitled “SYSTEM AND METHOD FOR SECONDARY PATH SWITCHING FOR ACTIVE NOISE CANCELLATION,” which is hereby incorporated by reference into this disclosure. In step 702, the controller sets up an algorithm for performing secondary path switching, IR fingerprint analysis, and matching conditions to initial conditions. In step 704, the controller remains idle and runs checks. For example, the controller may maintain a secondary acoustic transfer function The estimate of the estimation algorithm (or secondary path IR re-estimation) is inactive until a predetermined condition has been met. For example, in step 704, the controller can continuously monitor the levels of each group of signals (as will be discussed in more detail below) to determine when to enable or activate the secondary path IR re-estimation algorithm. In addition to monitoring the levels of each group of signals, the controller can also determine whether there is sufficient spectral content and signal-to-noise ratio (SNR) in the audio signal. Once these conditions are met, method 700 proceeds to 712. In one embodiment, one or more of these quality checks can be omitted. In one embodiment, different quality checks, such as confidence scores, can be obtained as described in step 728, thereby reducing the need for one or more conditions or quality checks. Step 728 will be described in more detail below.

[0055] In step 706, the controller determines whether the adaptive secondary path (ASP) is initialized. ASP corresponds to a process that performs online secondary path estimation in a multiple-input / multiple-output (MIMO) environment. If the ASP process can identify a closer match to the secondary path, this aspect may improve cancellation performance and the secondary path will be utilized.

[0056] If this condition is true, the method 700 proceeds to step 708. If not, the method 700 returns to step 704. In step 708, the controller determines whether an increasing true audio (TA) error has occurred. The TA error corresponds to the difference between the predicted audio signal at the error microphone and the actual audio signal at the error microphone 112. If this difference is increasing, the TA error indicates that the model of the secondary path is inaccurate.

[0057] If this condition is true, the method 700 proceeds to step 710. If not, the method 700 returns to step 704.

[0058] In step 710, the controller determines whether there is sufficient audio content and whether the SNR of the audio signal is above a predetermined level. The characterization is non-intrusive, and it is desirable that the level of any test signal be as low as possible so that the test signal is as inaudible as possible to the passengers. However, the level of the test signal needs to be high enough so that the test signal can be detected at the error microphone. This aspect provides an optimal amplitude range for the test signal. If this condition is true, the method 700 proceeds to step 712. If not, the method 700 proceeds to step 742 because the poor signal-to-noise ratio has caused the test signal to be too noisy to be used to reliably estimate

[0059] In step 742, the controller activates the shaped noise generator to generate a gain shaped noise signal that, when combined with the audio signal, has sufficient spectral content and signal-to-noise ratio (SNR) in the audio signal. In step 744, the controller executes a wait state during which the controller waits for the level to stabilize, which is indicated when the shaped noise generator generates an output. Once the level stabilizes, the controller proceeds to step 712.

[0060] In step 712, the controller prepares for processing. In other words, the controller activates the secondary path IR re-estimation algorithm. In step 714, the controller is committed to identifying the current audio configuration in the vehicle (or the cabin in the vehicle). In step 716, the controller continues to determine whether the audio content is sufficient in terms of spectral density and SNR. Generally speaking, the adaptive filter can only be adjusted if the audio content is sufficiently flat over the required bandwidth. To achieve this aspect, method 700 can use spectral descriptors (such as spectral flatness) to determine whether the audio content allows the adaptive filter to converge appropriately.

[0061] In terms of the SNR level of the error microphone in the vehicle, if the background noise in the vehicle is much louder than the audio played in the car, such noise may dominate the adaptive filter estimate (or secondary path estimate Therefore, in this case, if the audio is lower than the background noise of the vehicle, this situation may indicate a secondary path estimate In addition, the overall level in a microphone (e.g., an ANC microphone) relative to the error level at the microphone can also be used to determine the estimated secondary path For example, if the microphone exhibits a consistently low amplitude, this could indicate that the estimated secondary path If the controller determines that the audio content is adequate in terms of spectral density and SNR, the method 700 proceeds to step 718. If not, the method 700 proceeds to step 720.

[0062] In step 720, the controller determines whether an insufficient amount of audio is being received for more than a predetermined amount of time. If this condition is true, the method 700 proceeds to step 724 and terminates. If not, the method 700 returns to step 714. Additionally, in step 714, the controller monitors one or more signals from the secondary path IR re-estimation algorithm to determine whether convergence has been achieved. In general, the error microphone provides an output indicating the has converged to an acceptable error of S(z). Additionally or alternatively, a signal derived from the error (such as a gradient value from an adaptive filter using a gradient descent method) may be used to perform Online estimation of .

[0063] In step 718, the controller determines whether the audio content exhibits a small stable gradient. For example, the controller can compare the gradient on the audio content to a predetermined gradient value. A high gradient in the system may indicate that the filter of the estimated secondary path (e.g., IR that has not yet converged to the actual secondary path S(z). Since the gradient is usually a vector quantity, the L2 norm can be used to determine the estimated secondary path (e.g., Convergence of the filter (or convergence of the adaptive filter). When the L2 norm is consistently low (or below a predetermined gradient value), this may indicate that the estimated secondary path Therefore, if the condition is true (e.g., the gradient of the audio content is less than the predetermined gradient value), the method 700 proceeds to step 722. If not (e.g., the gradient of the audio content is greater than the predetermined gradient value), the method 700 returns to step 714.

[0064] In general, you may want to ensure that the secondary path estimates is reliable and can be used to estimate the secondary path using Update the secondary acoustic transfer function S(z) so that the adaptive unit can use the secondary path estimate To calculate the filter coefficients of the transfer function W(z) of the adaptive filter 108. In general, although the system can reliably generate But this may not mean The secondary path estimate of the ANC system will produce a robust or reliable fingerprint analysis. When the ANC system is started, the ANC system is configured for the default Or you can use the final determination in the last step of the vehicle The possible candidates are determined based on the selection that minimizes the performance degradation for the largest number of possible cabin configurations. The controller also includes instructions for executing an adaptive filter (or a deep learning or system identification algorithm) that iteratively solves for the secondary path based on a reference component of the microphone output signal and the reception of that signal by all error microphones through the actual S(z).

[0065] There are several reasons why it may be unreliable. For example, if the estimated secondary path estimate is obtained by the controller's adaptive filter, the adaptive filter may be incorrectly adjusted for various reasons, or include coefficients with values ​​approaching infinity (or provide a divergent filter). If the system is obtained through deep learning or system identification, the estimate may not be accurate if the system has not been trained and tested on as many real-world acoustic combinations as possible. The coefficients are analyzed and updated for fingerprint analysis Before calculating coefficients, it is usually necessary to assess the reliability of these coefficients.

[0066] High errors in ANC systems may indicate that the filter of the estimated secondary path There is a deviation from the actual IR of the secondary path S(z). Large errors may indicate that convergence has not occurred, leading to Therefore, to achieve this, the state machine (or method 700) can compare the current error to a tunable threshold, and if the error is below the threshold, the method 700 can conclude that for this metric, Generally speaking, the controller monitors the error in the microphone output.

[0067] In general, since the secondary path estimate can be estimated using one or more of these methods Therefore, it is used to verify the secondary path estimation The specific process may vary. Method 700 provides a method for using a number of possible methods that can be used to validate secondary path estimates based on adaptive system identification. Any combination of the following techniques may be used. An example of this can be seen between steps 704 (e.g., idle) and 714 (e.g., run IR estimation) of method 700, and also includes steps 706, 708, 710, 716, 718, and 720.

[0068] In step 722, the controller disables or deactivates the secondary path IR re-estimation algorithm and initializes the IR fingerprint analysis and matching process.

[0069] In step 724, the controller activates or triggers the matching algorithm to perform IR fingerprinting. In this step, a fingerprint of one or more IRs is derived. In general, the method 700 and ANC system provide, for example, fingerprinting IRs to dynamically switch secondary path parameters in an online single-input single-output (SISO) or MIMO ANC system. Fingerprinting involves applying signal processing techniques to estimate impulse responses from at least one or each of the ANC system. Each IR in a multi-channel ANC system displays unique properties about the electro-acoustic path. These properties take into account signal processing components such as analog-to-digital (A / D) and digital-to-analog (D / A) converters, amplifiers, speakers, microphones, as well as the acoustic path itself. The response of the electro-acoustic path, i.e., the frequency response of the electro-acoustic path, is greatly affected by the acoustic characteristics of the vehicle cabin. For a typical vehicle cabin, the frequency response exhibits peaks and valleys at different frequencies. Furthermore, the amplitude and frequency of the peaks and valleys will shift based on a variety of factors such as temperature, seating configuration, number of vehicle occupants, etc. It is these variations in the frequency response that provide the opportunity to extract a unique fingerprint from the IR.

[0070] This can be estimated from production vehicle IR measurements or different secondary paths The estimation technique obtains a library of pre-measured IRs from various configurations of the vehicle cabin. Another approach may involve using a secondary path estimate The estimation algorithm itself is derived from a library of pre-measured IRs from various acoustic cabin configurations. In addition, the controller can assign each cabin acoustic configuration in the library to a label (library 1, library 2, etc.) or use a look-up table (LUT). The library of pre-measured IRs can be stored as pre-stored IRs in a memory associated with the controller (or LUT 320) for comparison with the estimated impulse response. One or more of the .

[0071] IR measured in a production vehicle can provide a more accurate fingerprint for multiple cabin configurations because it uses a test signal to excite all frequencies. The estimation algorithm itself derives the IR from various acoustic combinations. Finally, each cabin acoustic configuration in the library is assigned a label (library 1, library 2, etc.).

[0072] Once the estimated After all fingerprint features are extracted, the ANC system and method 700 can estimate the secondary path A match is performed with an existing database of fingerprints (or predetermined (or pre-stored) secondary path estimates) from various cabin configurations to find the closest match. Finally, the adaptive unit loads the IR coefficients from the closest matching configuration into the adaptive filter of the ANC system. The ANC system performs the above method of estimating the impulse response and determining whether each speaker and microphone pair in the vehicle matches the pre-stored impulse response.

[0073] The matching process performed by the ANC system may involve the controller calculating a distance metric, determining a winner (e.g., the closest matching IR value to a pre-stored IR), estimating secondary paths, and determining a winner. The matching confidence between the closest matching IR value and each pre-stored IR value, and the winning margin (MOV) between the closest matching IR value and multiple of the pre-stored IRs are determined.

[0074] In step 724, the controller calculates the distance metric to determine the secondary path estimate IR, The closest match to a previously stored IR library. Examples of distance metrics may include the Euclidean norm, the Lp norm, or their variants such as Hausdorff distance and normalized misalignment.

[0075] In step 726, the controller determines whether the match is complete. For example, the controller builds a voting matrix of the closest matching IR values ​​relative to the closest previously stored IRs stored in memory. If this condition is true (e.g., the voting matrix has been built), the method 700 proceeds to step 728. If not, the method 700 returns to step 724.

[0076] In step 728, the controller builds a confidence metric by considering various aspects of the voting matrix to determine the final winning IR configuration. This will also be combined with Figure 6 The controller determines whether the IR candidate exhibits a confidence greater than a predetermined value (eg, 50%). If this condition is met, the method 700 proceeds to step 732. If not, the method 700 returns to step 712.

[0077] In step 730, the controller determines the secondary path estimate The controller determines whether the MOV is above a predetermined MOV threshold (e.g., 20%). If the MOV is above the predetermined MOV threshold, the method 700 proceeds to step 732. If not, the method 700 returns to step 712.

[0078] In step 732, the controller determines whether the selected IR time slot is different from the current time slot. If the selected IR time slot is exactly the same as the current IR time slot, the ANC system is currently using the most suitable IR and does not need to change to another IR. If this condition is true, the method 700 proceeds to step 734 to perform an IR switch. If not, the method 700 returns to step 712. In step 734, the controller performs or makes an IR switch. For example, the controller performs a triggering IR switch mechanism. In this case, the controller performs an IR switch mechanism to start from the old IR. Parameter transition to The new parameters are matched to the measurements. Abrupt switching may temporarily degrade ANC performance, but is possible. To avoid this, the IR switching mechanism includes a smooth turnaround to minimize any audible artifacts. The time constant of the turnaround may be tunable or variable, but may be done in the range of, for example, 100 milliseconds to several seconds. Another embodiment may include using an adaptive filter The coefficients are instantaneously updated for the IR of the secondary estimation path. It is generally believed that one aspect may involve utilizing Measurements match new parameters, rather than impulse response measurements (or measurement) corresponding to the parameters.

[0079] Generally speaking, the controller may perform one of the following options when executing the IR switching mechanism. In the first option, the controller updates all coefficients of the adaptive filter through the adaptive unit. For example, the LMS parameters of the step size and leakage may be temporarily modified so that the adaptive filter of the ANC system may not respond when an unexpected transient occurs. In the second option, the controller performs a slewing (average) update of all coefficients of the adaptive filter through the adaptive unit over a short period of time. In this case, the coefficients are updated stepwise using, for example, an averaging filter to incorporate the coefficients from the secondary path estimate. The current value of the coefficient is mixed into the secondary path estimate The averaging can be linear or exponential, and the total update time can be tunable (for example, the total update time can range from 40 milliseconds to 5 seconds).

[0080] In step 736, the controller determines whether the coefficients provided by the adaptation unit to the adaptive filter have been correctly loaded. For example, the controller monitors the state of the ANC after the switch has occurred in step 734. If the appropriate coefficients have been correctly loaded, the method 700 proceeds to step 738. If not, the method 700 returns to step 434.

[0081] In step 738, the controller executes a wait state. For example, the controller determines whether the selected IR (or new IR) is stable. Generally, after the update is complete, the controller monitors the effect of the newly updated parameters. If the controller detects divergence, or if these new IRs are inconsistent with the newly measured The error signal between is too high, or higher than the error signal immediately before the coefficient swap. The controller can quickly revert to previous parameters because they are stored rather than deleted. However, if no problem is detected, the controller may The estimated coefficients of the adaptive filter are kept frozen for a longer time (e.g., 5 to 30 minutes), and then the IR estimation algorithm is resumed and the same logic is repeated to decide whether, when, and how to update again. This process may continue to run for the duration of the vehicle ignition cycle. If the selected IR (or estimated secondary path ) is stable, method 700 proceeds to step 740. If not, method 700 returns to step 734. In step 740, the controller determines whether the timer for the wait state has expired. The controller uses a timer and waits for the timer to expire to determine whether the state of the selected IR (or new IR) is stable. If the state of the selected IR is stable after the time expires, method 700 proceeds to step 704. If not, method 700 returns to step 738.

[0082] In general, the various aspects disclosed herein provide distance measurements (or misalignments) that are derived from an estimated With actual IR S p The L2 normalized squared error between (z) is defined. This can be defined by the following mathematical formula:

[0083] This can then be converted to a logarithmic scale. It should be appreciated that the disclosed system can also perform the above equations in the time domain. This is generally referred to as a relative modeling error. An example of such a relative modeling error is generally described in "A New Online Secondary Path Modeling Method with An Auxiliary Noise Power Scheduling Strategy for Narrowband Active Noise Control Systems" by Sun et al., published on November 29, 2017, which is hereby incorporated by reference in its entirety. The controller provides a distance vector between each IR and the stored IRs in the distance matrix.

Claims

1. A system for generating a stimulus signal for an adaptive secondary path of an active noise cancellation (ANC) system in a vehicle, the system comprising: at least one audio signal source for providing an audio signal in a cabin of the vehicle; a shaped noise generator for providing a shaped noise signal in said compartment of said vehicle; at least one controller programmed to activate the shaped noise generator and add the shaped noise signal to the audio signal to generate a stimulation signal; as well as At least one speaker for projecting the stimulation signal within the cabin of the vehicle to cause the ANC system to generate an anti-noise signal.

2. The system of claim 1, further comprising: at least one microphone for providing a signal indicative of a noise level in said cabin of said vehicle; as well as A gain scheduling component is configured to adjust a gain of the shaped noise signal to psychoacoustically mask the shaped noise signal and maintain a constant signal-to-noise ratio (SNR), wherein the gain adjusted shaped noise signal is added to the audio signal to generate the stimulation signal.

3. The system of claim 1, wherein: The shaped noise generator also includes a pre-recorded signal.

4. The system of claim 3, wherein: The pre-recorded signal is a cyclic masking signal.

5. The system of claim 3, wherein: The pre-recorded signal is a horn of the vehicle.

6. The system of claim 1, wherein: The shaped noise generator also includes an active sound design (ASD) system.

7. The system of claim 6, wherein: The shaped noise generator works together with the ASD and also includes: a white noise generator; and Linear predictive coding component.

8. The system of claim 1, wherein: The linear predictive coding component also includes: a white noise generator; and Linear predictive coding component.

9. The system of claim 8, wherein: The linear predictive coding component also includes a set of predetermined LPC synthesis filters applied to the output of the white noise generator to generate the stimulation signal.

10. The system of claim 9, wherein: The controller selects the predetermined LPC synthesis filter to be applied to the output of the white noise generator based on vehicle parameters.

11. The system of claim 8, wherein: The linear predictive coding component also includes an input from an error microphone, and the controller updates the LPC filter in real time based on the input from the error microphone.

12. The system of claim 11, further comprising: at least one microphone for providing a signal indicative of a noise level in said cabin of said vehicle; as well as A gain scheduling component is configured to adjust a gain of the shaped noise signal, wherein the gain-adjusted shaped noise signal is added to the audio signal to generate the stimulation signal.

13. A method for generating a stimulus signal for an adaptive secondary path of an active noise cancellation (ANC) system in a vehicle, the system comprising: generating an audio signal in a cabin of the vehicle using at least one audio source; generating a shaped noise signal in the cabin of the vehicle using a shaped noise generator; masking the shaped noise signal; adding a masked shaped noise signal to the audio signal to generate a stimulation signal; as well as The stimulation signal is projected within the cabin of the vehicle for the ANC system to generate an anti-noise signal.

14. The method of claim 13, further comprising the steps of: detecting noise levels using a microphone in a cabin of the vehicle; as well as The gain of the masked shaped noise signal is adjusted to psychoacoustically mask the shaped noise signal and maintain a constant signal-to-noise ratio (SNR).

15. The method of claim 13, further comprising the steps of: Upon determining that the audio signal does not have sufficient content for the ANC system to generate an anti-noise signal, the shaped noise generator is activated to generate the shaped noise signal.

16. The method of claim 13, wherein: The step of masking the shaped noise signal also includes: generating a white noise signal; and A bank of linear predictive coding (LPC) filters is applied to the white noise signal.

17. The method of claim 16, wherein: The set of LPC filters is a predetermined set of LPC filters, and the step of applying the set of LPC filters also includes selecting an LPC filter to be applied to the white noise signal from the predetermined set of LPC filters, wherein the LPC filter to be applied is selected based on vehicle speed or engine revolutions per minute (RPM).

18. The method of claim 16, wherein: The LPC filters in the set of LPC filters are tuned online based on input from an error microphone, and the method further comprises the step of updating the LPC filters in real time based on the input from the error microphone.

19. The method of claim 18, wherein: The input from the error microphone is downsampled before applying the LPC filter, and the output of the LPC filter is upsampled before being added to the audio signal.

Images