Adaptive active noise reduction method and system

Abstract

Adaptive active noise reduction method and system. The embodiments of the present application provide an ear cavity frequency response EFCR adaptive noise reduction ANC method and system that performs path compensation on the entire main path MPED leading to the user's eardrum. Many ANC filter models are pre-trained to include corresponding anti-noise path ANP filter models and corresponding MPED filter models representing ANC filter configurations. When a user wears an earphone earpiece, the wearer's characteristics and wearing position / orientation show the wearer / wearing state. The technology described in the present application can continuously or periodically and effectively determine which pre-trained ANC filter model most accurately describes the current MPED of the current wearer / wearing state, and can continuously or periodically update the ANC filter configuration based on the pre-trained model to maintain high-performance ANC including EFCR path compensation.

Description

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS

[0002] This application claims priority to the following application, the contents of which are incorporated by reference in their entirety into this application: U.S. patent application entitled “ACTIVE NOISE CANCELLATION FILTER ADAPTATION WITH EARCAVITY FREQUENCY RESPONSE COMPENSATION” filed on March 15, 2022 and patent application number 17 / 695,816. Technical Field

[0003] Embodiments described herein relate generally to audio processing and noise reduction, and more particularly to an adaptive active noise reduction filter with ear cavity frequency response compensation in a headphone system. Background Art

[0004] Many high-performance headphone systems include active noise cancellation (ANC). The purpose of ANC is to improve the wearer's listening experience by dynamically canceling ambient and / or other noise. This cancellation typically involves generating synthetic noise through an audio transducer (e.g., a speaker located in the area or location) so that the generated signal, the "anti-noise" signal, ideally has the same amplitude and phase as the noise, but opposite polarity. Playing the anti-noise signal along with the desired audio can effectively cancel most of the noise.

[0005] For various reasons, such as the dynamically changing nature of noise, the anti-noise signal often cannot eliminate all noise. Typically, an error sensor (e.g., an error microphone) is placed in the audio output path to receive a mixture of ambient noise and the generated synthetic anti-noise signal. The mixture received by the error sensor is an error signal, which changes over time. Typically, ANC algorithms are used in ANC filter design to dynamically adapt and constantly seek to minimize the noise experienced by the wearer. In some conventional ANC implementations, the ANC filter is set in a laboratory and / or manufacturing setting to generate a fixed ANC, and the fixed ANC is incorporated into the headphone environment. In this case, the headphones may not include an integrated error sensor, and the ANC filter settings may not be dynamically updated when the headphones are used. In other conventional implementations, the headphones include an integrated error sensor, and the ANC filter settings can be dynamically updated in real time based on error signal feedback during use.

[0006] Many different types of approaches have been used with conventional ANC implementations to improve noise reduction. For example, various techniques are used to shorten feedback paths, reduce processing delays, adapt to broadband and narrowband types of noise, etc. However, even very high-performance ANC implementations often do not take into account errors (e.g., noise) caused by the wearer's physiology and / or the specific position of the headphones relative to the wearer's physiology. For example, the shape of the user's inner ear, the distance to the user's eardrum, the relative position of the headphone output speakers to the user's ear structure, the effectiveness of the acoustic seal provided by the headphones, etc., all affect the way the user hears the sound generated by the headphones, including the user's experience of the desired audio, the mixing of the desired audio with ambient noise, and the anti-noise signal generated. Conventional ANC methods often cannot compensate for this physiological effect at all, and such rare conventional attempts are usually unsuccessful. Summary of the invention

[0007] Embodiments include systems and methods for active noise reduction or control in a manner that compensates for the entire main noise path to the user's eardrum. For example, an ear cavity frequency response (EFCR) adaptive noise cancelation (ANC) system is integrated into an earphone receiver and performs path compensation on the entire main path to the eardrum (MPED). The system includes an ANC filter system for outputting an anti-noise signal at a target location (on the front of the earphone receiver) based on a current ANC filter configuration, such that the anti-noise signal output at the target location is an inverse estimate of an in-ear noise signal. Due to the acoustic path effect of the MPED, when the user wears the earphone receiver in the current wearer / wearing state relative to the user's outer ear, the in-ear noise signal corresponds to the ambient noise and is displayed at the target location. An error microphone records a residual noise signal representing a combination of the anti-noise signal and the in-ear noise signal in the current wearer / wearing state. A model data storage device may store K (e.g., 50 or more) pre-trained ANC models, each K-th pre-trained ANC model having a K-th anti-noise path (ANP) filter model and a k-th MPED filter model previously trained under the K-th trained wearer / wearing state.

[0008] The user adaptive engine can be coupled to the ANC filter system, the error microphone, and the model data storage. The user adaptive engine can be used for each K-th pre-trained ANC model to: calculate the k-th estimated actual noise signal using the k-th ANP filter model to convert the anti-noise signal into a path-compensated anti-noise signal, and remove the path-compensated anti-noise signal from the residual noise signal; calculate the k-th estimated eardrum noise signal using the k-th MPED filter model to convert the reference noise signal; and calculate the k-th candidate residual noise between the k-th estimated actual noise signal and the k-th estimated eardrum noise signal. The user adaptive engine can then select one of the pre-trained ANC models as the model that generates the lowest k-th candidate residual noise, and can direct the ANC filter system to replace the current ANC filter configuration with a new ANC filter configuration based on one of the selected pre-trained ANC models. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] The benefits and advantages of the inventive concept will be apparent from the detailed description of the embodiments of the present disclosure and the accompanying drawings, in which like reference numerals and numbers refer to like parts throughout the drawings. The drawings are not drawn to scale; emphasis is placed on illustrating the principles of the inventive concept.

[0010] Figure 1 A simplified diagram of a headphone environment including a conventional active noise cancellation (ANC) system is shown.

[0011] Figure 2 Another simplified diagram of a conventional headphone environment is shown, which includes a conventional hybrid active noise reduction system more explicitly represented as separate wideband (WB) ANC subsystem and narrowband (NB) ANC subsystem.

[0012] Figure 3 A simplified block diagram of an exemplary ANC architecture for hybrid ambient noise reduction with ear cavity frequency response compensation is shown in accordance with various embodiments described herein.

[0013] Figure 4 A simplified block diagram of an exemplary ECFR-trained ANC architecture is shown in accordance with various embodiments described herein.

[0014] Figure 5 An exemplary implementation of a user adaptation engine according to various embodiments is shown.

[0015] Figure 6 An example operational block diagram is shown of switching from one ANC filter configuration to another ANC filter configuration based on updating a selected index in response to filter selection by a user adaptation engine.

[0016] Figure 7 A flowchart of an exemplary method for pre-training an ANC model based on ear cavity frequency response according to various embodiments described in this application is shown.

[0017] Figure 8 A flow chart of an exemplary method for ear cavity frequency response adaptive noise reduction with path compensation over the entire main path to the eardrum according to various embodiments described herein is shown.

[0018] Specific implementation method

[0019] The embodiments described in the present application provide active noise reduction filter adaptation with dynamic ear cavity frequency response compensation in a headphone system. As used in the present application, "headphones", "headphone systems", etc. may generally refer to any type of wearable speaker system in which a speaker is positioned directly above and / or in one or both ears of a user. For example, the embodiments described in the present application may be used for closed headphones, open headphones, on-ear headphones, ear-hook headphones, in-ear headphones, earplugs, and / or any other suitable type of wired or wireless headphones. Typically, users want to hear "desired audio", such as music, recordings, etc. At the same time, various factors can reduce the user's experience of the desired audio. Therefore, many high-performance headphone systems include active noise reduction ANC.

[0020] The presence of ambient noise is a group of factors that prevent the headset from providing the desired audio to the user in an optimal manner. The term ambient noise used in this application generally includes any ambient noise and / or any other audio that the user can hear but is not part of the desired audio. For example, if the user attempts to listen to audio in a noisy environment, the user may not be able to hear the audio well, and / or the ambient sound may also interfere with the listening experience. The specific relationship between the headset and the user's ear physiology may be another group of factors that prevent the headset from providing the desired audio to the user in an optimal manner. For example, whether the earpiece of the headset system is in the user's ear, above the ear, or around the ear, the combined environment of the earpiece, the ear cavity, the eardrum, and other related structures all show a specific acoustic environment with a specific frequency response. For each user's ear and the different positions of the earpiece relative to the ear, the acoustic environment and the corresponding frequency response may be different.

[0021] As used in this application, the term "wearer" is used to refer to an individual wearing headphones, whether on one or both ears, above one or both ears, in one or both ears, etc. This application assumes that when audio is played through the headphones (e.g., music, speech, etc.), the wearer hears a combination of the played audio and ambient noise (i.e., assuming no ANC). The anatomical structure of the wearer's ear can be divided into the outer ear, the middle ear, and the inner ear. The outer ear (pinna) is a structure formed primarily of cartilage that can effectively collect sound waves and guide them to the eardrum (tympanic membrane) through the ear cavity (sound canal). The shape, size, and other features of the outer ear anatomical structure (e.g., helix, anti-helix, tragus, anti-tragus, ear canal, earwax, etc.) can define the frequency response of the outer ear up to the eardrum. For simplicity, this application assumes that the wearer "hears" mainly based on the way the desired and undesired audio information (sound waves) entering the outer ear eventually reach the eardrum and interact with the eardrum. For example, this includes the effects of outer ear anatomy and physiology on the audio information, but does not include the anatomical or physiological effects of middle or inner ear structures.

[0022] When an individual wears the headphones, each earpiece of the headphones (or in some cases a single earpiece) is placed in a wearing position relative to the wearer's ear anatomy. It can generally be assumed that the wearing position defines an outward-facing environment and an inward-facing environment, and that there may be an intermediate environment between the outward-facing environment and the inward-facing environment (e.g., in the space occupied by the headphones). The headphones are intended to project the desired audio (e.g., an anti-noise signal in the case of ANC) into the inward-facing environment using the primary earpiece speaker, and to block and / or eliminate ambient noise from the outward-facing environment. In a typical ANC implementation, this may involve recording ambient noise from the outward-facing environment, generating an anti-noise signal based on the recorded ambient noise and an intermediate environment model, and projecting the anti-noise signal together with the desired audio into the inward-facing environment (e.g., using a secondary anti-noise speaker).

[0023] Figure 1 A simplified diagram of a headphone environment 100 including a conventional active noise reduction ANC system 110 is shown. For example, the ANC system 110 can be integrated into a headphone system, such as into a headphone earpiece. As described above, placing each earpiece in the wearer's ear can effectively define an outward environment, an inward environment, and an intermediate environment. For example, a reference microphone records ambient noise at an outward boundary of the intermediate environment, the ANC system 110 generates an anti-noise signal 115 (synthetic noise signal) based at least on the ambient noise, and the earphone outputs the anti-noise signal 115 at the inward boundary of the intermediate environment (target location). The target location is typically very close to the location of the main earpiece speaker (and / or error microphone), from which the earpiece outputs the desired audio.

[0024] Typically, the purpose of outputting the anti-noise signal 115 at the target location is to at least partially cancel any version of the ambient noise (illustrated as the in-ear noise signal 125) present at the same target location. For example, by passing through the intermediate environment on the "main noise path", a portion of the ambient noise from outside the ear arrives at the target location as the in-ear noise signal 125. A small speaker located substantially at the target location outputs the anti-noise signal 115 so that the target location now includes a combination (e.g., superposition) of the anti-noise signal 115 and the in-ear noise signal 125 (and any desired audio in addition). Generally, it is believed that maximized noise reduction occurs when the anti-noise signal 115 is the same as the inverse phase of the in-ear noise signal 125. For example, the combination of the anti-noise signal 115 and the in-ear noise signal 125 generates a residual noise signal 135, which is essentially an error signal for noise reduction. Some conventional implementations include an error microphone (also located at the target location) that records the residual noise signal 135 and feeds the residual noise signal 135 back to the ANC system 110 for dynamic adaptation.

[0025] As described above, ambient noise is considered to propagate from the outward environment to the target location through the main noise path, otherwise referred to as the main path. The main path is actually a collection of physical paths that ambient noise can reach the target location, so that the main path also has a specific main path frequency response affected by the physical structure of the headset and some anatomical structures of the outer ear. As shown in the figure, the conventional ANC system 110 can consider the anti-noise path 120, also known as the secondary noise path, or more simply the secondary path. The anti-noise path 120 can be a path (or a model representation of the path) between the location where the anti-noise signal 115 is generated and the location where the anti-noise signal 115 is output. For example, it can be mathematically assumed that the reference noise signal 105 and the feedback residual noise signal 135 are both received at a spatial point at a certain distance from the spatial point coinciding with the target location. This distance can add delays and / or other artificial signals to the anti-noise signal 115. Because headphones with ANC features usually receive reference noise at the outward component of each earpiece and output the generated anti-noise signal 115 at the inward component of each earpiece, conventional ANC designs usually assume that the main path is substantially equivalent to the intermediate environment, as shown in the figure. Therefore, conventional ANC systems 110 are intended to use the anti-noise path 120 as a representation of the primary path, so that compensation for the anti-noise path 120 is considered compensation for the primary path effects.

[0026] In general, the ANC system 110 seeks to generate the anti-noise signal 115 such that the anti-noise signal 115 as an output at the target location after passing through the anti-noise path 120 combines with any ambient noise at the target location to minimize the residual noise at the target location. For example, the ANC system 110 includes one or more filters having filter parameters that are adjustable based on feedback of the reference noise signal 105 and / or the residual noise signal 135 to generate the anti-noise signal 115. Dynamic control of the filter parameters seeks to generate the anti-noise signal 115 such that the corresponding anti-noise signal 115 passing through the anti-noise path 120 will combine with (i.e., destructively interfere with) the in-ear noise signal 125 to minimize the residual noise signal 135. Because ambient noise generally varies over time, any delay in generating the anti-noise signal and transmitting the anti-noise signal on the anti-noise path 120 will result in a mismatch between the generated anti-noise signal 115 and the noise it is trying to cancel (represented by the in-ear noise signal 125). Therefore, the anti-noise path 120 is expected to have as little delay as possible.

[0027] Typically, there are at least three methods for an ANC system to synthesize an anti-noise signal 115. One method synthesizes noise by capturing reference noise using a reference sensor. The reference noise can be an earlier version of the noise with additional reflections of the noise through multiple paths. The synthesis of the noise can be accomplished by applying an adaptive filter or controller to the reference noise, thereby minimizing the error (difference) between the noise and the synthesized noise played. The noise synthesis must be done quickly so that it adds little delay and the synthesized noise can arrive at the target location almost synchronously with the noise being eliminated. This ANC method is called "feedforward" ANC. Because there is a reference sensor to sense the earlier version of the noise, feedforward ANC can eliminate relatively wideband noise. Therefore, this feedforward-based ANC method is referred to as wideband ANC in this application.

[0028] A second approach to synthesizing an anti-noise signal by an ANC system seeks to cancel narrowband noise (typically from 20 Hz to 1000 Hz for ANC purposes), or noise components that include multiple tonal signals, by predicting the synthesized noise. For example, an implementation of an ANC system may estimate the noise using an error signal from an error sensor (e.g., an error microphone), and the noise after passing through a secondary path may be predicted from the estimate. This ANC approach is referred to as "feedback" ANC. Using a pure feedback ANC approach, a reference sensor may not be required. If the secondary path and processing have low latency, the prediction performance can be higher and the waterbed effect (unwanted noise with a relatively narrow frequency band) is lower. Therefore, for certain bandwidths, feedback ANC has better performance and lower latency. If the signal is not narrowband, the narrowband requirement can be achieved by emphasizing some frequency bands that require noise reduction. This feedback-based ANC approach is referred to as narrowband ANC in this application.

[0029] A third approach to synthesizing an anti-noise signal by an ANC system seeks to cancel both wideband and narrowband noise.Such an approach effectively combines a WB ANC implementation and a NB ANC implementation into a hybrid ANC, which is referred to in this application as a "hybrid" ANC. Figure 1 The conventional ANC system 110 of can be viewed as a simplified illustration of a hybrid ANC method because it includes both a reference noise signal 105 of a feedforward type and a residual noise signal 135 of a feedback type. In the WB ANC path of the hybrid method, a reference microphone can be used to detect the reference noise signal 105, corresponding to the ambient noise in the external environment. The reference microphone can pass the detected reference noise signal 105 to the ANC system 110, which generates an anti-noise signal 115. The anti-noise signal 115 is output to the target location via the anti-noise path 120 (e.g., while simultaneously outputting the desired audio at substantially the same target location). In an ideal case, the anti-noise signal 115 completely cancels any ambient noise at the target location, as represented by the in-ear noise signal 125. However, in reality, there will be some cancellation errors, so the sum of the anti-noise signal 115 at the target location and the in-ear noise signal 125 at the target location (represented by the adder 130) generates a residual noise, represented as the residual noise signal 135. In the NB ANC path of the hybrid approach, the residual noise signal 135 is fed back to the ANC system 110 for dynamic feedback control of the ANC filter parameters. Although not explicitly shown, feedback of the residual noise signal 135 can be achieved by placing an error sensor at or near a target location to detect the residual noise signal 135. For example, one or more error microphones can be integrated with the headphone system (e.g., at an inward-facing position on one or both earpieces) to record noise at or near a target location.

[0030] Figure 2 Another simplified diagram of a conventional headphone environment 200 is shown, which includes a conventional hybrid active noise reduction system 110 more specifically represented as a separate wideband ANC subsystem 110w and a narrowband ANC subsystem 110n. Figure 1 Components and signals similar to those in the drawings are labeled with the same reference designations. As shown, each ANC subsystem receives its own input signal from which it generates its own anti-noise component 215 to counter the noise signal 115. Specifically, as described above, the WB-ANC subsystem 110w generates its anti-noise component 215w based on receiving a reference noise signal 205w from the reference microphone 105, and the NB-ANC subsystem 110n generates its anti-noise component 215n based on receiving a feedback signal 205n related to the residual noise signal 135.

[0031] The WB anti-noise component 215w and the NB anti-noise component 215n are combined at the summing node 210 to generate the anti-noise signal 115. Figure 1 As shown, through Figure 1 The anti-noise path 120 corresponding to the anti-noise path 120 of the anti-noise path 120 sends the anti-noise signal 115 to the target location. At the target location, the anti-noise signal 115 destructively interferes with the local version of the ambient noise (the in-ear noise signal 125) at the target location to generate a residual noise signal 135. For example, the adder 130 indicates that the interference at the target location between the in-ear noise signal 125 and the anti-noise signal 115 causes the residual noise signal 135 (i.e., representing the noise reduction error). As shown, an embodiment may include an error microphone 235 to record the residual noise signal 135.

[0032] In the illustrated implementation, rather than feeding the residual noise signal 135 directly back to the NB-ANC 110n, the feedback is adjusted to compensate for the impulse response of the anti-noise path 120. As shown, the ANPi filter 220 is used to match the frequency response (impulse response) of the anti-noise path 120. The NB anti-noise component 215n may be added to the residual noise signal 135 by the ANPi filter 220 and at the feedback compensation node 230, and the output of the feedback compensation node 230 may be fed back as the feedback signal 205n is input to the NB-ANC subsystem 110n. In some other cases, the residual noise may be used directly as a NB ANC reference.

[0033] Mathematically, the residual noise signal 135 (“e”) can be described as: e=n–SP*y, where “n” is the target position noise signal 125, “y” is the anti-noise signal 115, and “SP” is the corresponding representation of the anti-noise path 120. The in-ear noise signal 125 is unknown, but it can be estimated as: n=e+SPi*y, where “SPi” is the pre-computed impulse response of SP, as implemented by the ANPi filter 220. In some implementations, the reference input 205w of the WB-ANC subsystem 110w is the reference microphone signal 105. In other implementations, the reference input 205w of the WB-ANC subsystem 110w can correspond to an earlier version of the reference noise, such as with additional reflections of the noise via one or more paths. In such an implementation, the reference input 205w (“x”) can be described as: x=SPi*xw, where “xw” is the reference microphone signal 105. The output of the WB-ANC subsystem 110w may be described as: yw = WB*x, where "yw" is the WB anti-noise component 215w generated by applying the WB-ANC subsystem 110w filter, as indicated by "WB". Therefore, the residual noise after applying the filtering from the WB-ANC subsystem 110w may be described as: ew = n - yw; and the filter coefficients of the WB-ANC subsystem 110w may be updated according to the following formula: WB = WB + mu*ew*x, where "mu" is an adaptive coefficient, which may be a variable depending on which adaptive filtering algorithm is used.

[0034] The filter coefficients of the NB-ANC subsystem 110n may also be adaptively updated according to the same or any other suitable adaptive algorithm. It can be seen that the feedback reference input 205n of the NB-ANC subsystem 110n corresponds to the residual noise filtering of the WB-ANC subsystem 110w after passing through the anti-noise path 120 (the response is modeled as SPi). Therefore, the feedback reference signal 205n can be described as: xn=SPi*ew. The output of the NB-ANC subsystem 110n can be described as: yn=NB*xn, where "yn" is the NB anti-noise component 215n generated by applying the NB-ANC subsystem 110n filter, as shown by "NB". Therefore, the residual noise after applying the filtering from the WB-ANC subsystem 110w and the NB-ANC subsystem 110n can be described as: ee=ew-yn; and the filter coefficients of the NB-ANC subsystem 110n can be updated according to the following formula: NB=NB+mu*ee*xn.

[0035] In general, the WB-ANC subsystem 110w implements a filter for which the reference microphone 105 provides an adaptive filter reference (i.e., input signal 205w), and the filter parameters are dynamically adapted to minimize residual noise at a target location (e.g., detected by the error microphone 235) without being influenced by the NB-ANC subsystem 110n. The NB-ANC subsystem 110n implements a filter for which the path compensation feedback from the target location provides an adaptive filter reference (i.e., input signal 205n), and after being influenced by the WB-ANC subsystem 110w, the filter parameters are dynamically adapted to minimize residual noise at the target location (e.g., detected by the error microphone 235). Each of the WB-ANC subsystem 110w and the NB-ANC subsystem 110n can be optimized in different ways. For example, some methods optimize each of the WB-ANC subsystem 110w and the NB-ANC subsystem 110n independently, in series, or in parallel. Other approaches jointly optimize the NB-ANC subsystem 110n and the WB-ANC subsystem 110w.

[0036] Various ANC designs can be implemented using analog and / or digital circuits. In some designs, such as for lightweight devices, active resistor and capacitor (RC) circuits may be effective for analog ANC designs. However, it is difficult to change RC circuit parameters in real time to adapt to changing environments. In addition, due to differences between RC components, they may vary from device to device even if the device acoustics are of the same type. This may involve using different component values ​​in different RC circuits for each device, which may involve a lot of design work and may be impractical or even impossible to overcome in mass production. Digital designs can be more flexible than analog designs because the processing of modern algorithms can be easily implemented by digital components, such as digital signal processors (DSPs), etc. Therefore, many high-performance ANC designs are designed to be implemented primarily or exclusively using digital circuits. The filtered least mean squares (LMS) algorithm with a finite impulse response (FIR) filter is widely used.

[0037] refer to Figure 1 and Figure 2Conventional hybrid ANC designs are discussed, which represent only a portion of many different types of hybrid ANC design architectures. As described above, hybrid ANC methods generally use feedback filters to predict noise to eliminate low-frequency and / or tone-like noise, and use feedforward filters to synthesize anti-noise from reference noise to eliminate broadband or wideband noise. Both analog and digital circuits can be used. Over the years, digital circuits with advanced algorithms have been successfully used in many ANC designs, but noise reduction performance is often limited due to at least the high latency in the playback synthetic noise path (i.e., the anti-noise path, or secondary path). Recently, fast digital processing circuits have been used to adjust filter coefficients in near real time. While this improves performance, ANC designs are still limited in their ability to fully utilize complex algorithms due to the high computational costs and power consumption associated with running such algorithms in real time. Many improved hybrid ANC designs are described in U.S. Patent No. 11,189,261, entitled "Hybrid Active Noise Control System," filed on May 31, 2020, which is incorporated herein by reference in its entirety.

[0038] While many different ANC designs are known, conventional ANC designs are not intended to account for acoustic effects in the outer ear region between a conventional target location and the wearer's eardrum. As described above, headphones that include ANC features may typically include inward-facing components, such as a speaker for outputting at least a synthesized anti-noise signal 115 at the target location, and a microphone for detecting a residual noise signal 135 at the target location. When an individual wears such headphones, each individual earphone (i.e., a right earphone and a left earphone) is placed or partially placed on a corresponding ear, thereby positioning the inward-facing components at a specific location on the wearer's outer ear. Because the location of the inward-facing components in each ear is removed from the eardrum of that ear (the earphones are not designed to contact the eardrum), an additional acoustic environment is formed between the target location and the eardrum of each ear. Each such acoustic environment has its own frequency response, which depends on many factors, including "wearer" factors and "wearing" factors. Wearer factors may include, for example, the anatomical structure of the corresponding area of ​​the outer ear and the physiological properties of those outer ear structures. Wearing factors may include, for example, the position and orientation of the earphone in the outer ear, the acoustic seal between the outward-facing and inward-facing environments due to insert earphones, the acoustic properties of the earphone itself, etc. As used in this application, when an earphone is placed in the ear by a wearer, it is said to be in a particular "wearer / wearing state," which corresponds to a set of wearers and wearing factors. Therefore, any wearer / wearing state will exhibit a characteristic frequency response, referred to in this application as the "ear cavity frequency response," or ECFR. While conventional ANC implementations seek to minimize the effects of ambient noise at a target location (where the inward-facing components of the earphone are located in the wearer's outer ear), conventional implementations do not aim to consider ECFR in noise reduction.

[0039] Embodiments described in the present application provide new methods for dynamically adjusting ANC features to optimize audio delivery using ECFR-compensated ambient noise reduction. For example, physical in-ear sensor hardware can be used in a test environment to pre-train hybrid ANC filter parameters for a large number of ECFR models. Each pre-trained ECFR model can be associated with an identifier. During use, a user adaptation engine can use a hybrid ANC filter to continuously and adaptively eliminate ambient noise while also continuously iterating the pre-trained ECFR model to dynamically determine which combination of ANC filter parameters can minimize residual noise in the ear cavity. Some embodiments may further adjust the filter parameters away from the pre-trained ECFR model settings for additional optimization. In some such embodiments, those further optimized parameters may be stored as additional ECFR models.

[0040] Figure 3 A simplified block diagram of an exemplary ANC architecture 300 for hybrid ambient noise reduction with ear cavity frequency response (ECFR) compensation according to various embodiments described herein is shown. The exemplary ANC architecture 300 includes the ANC 110. For greater clarity, the details of the ANC 110 are shown substantially as Figure 2 A hybrid ANC implementation including a WB-ANC subsystem 110w and a NB-ANC subsystem 110n. It should be understood that although some novel features are described herein in the context of a particular illustrated hybrid ANC implementation, those same features may be applied to any suitable ANC design, including an ANC design having only a WB-ANC subsystem 110w, only a NB-ANC subsystem 110n, and the like.

[0041] As described above, conventional ANC structures seek to minimize ambient noise at a target location corresponding to the location of the error microphone 235 of the earpiece. The conventional target location is also typically the location of the main earpiece speaker that outputs the desired audio or a location very close to the main earpiece speaker. For example, conventional ANC is intended to use one or more reference signals (e.g., reference noise signal 205w and / or feedback signal 205n) to generate an anti-noise signal 115, which is output at approximately the same location as the error microphone 235 via an anti-noise path 120 corresponding to the ANC secondary path. The ambient noise arrives at approximately the same location via the conventional main path, which can be considered to be the approximate main path to the error microphone (main path to the error microphone, MPEM). The MPEM has a specific frequency response, which causes the ambient noise to appear as an in-ear noise signal 125 at the conventional target location. The desired interference between the anti-noise signal 115 and the in-ear noise signal 125 results in a residual noise signal 135 corresponding to the noise reduction error at the conventional target location.

[0042] As shown, the listening environment can be considered to include an additional "third" path between the conventional target position at the end of the MPEM and the wearer's eardrum. For example, the third path includes at least the ear cavity. In a simple representation, the individual "hears" the sound signal when interacting with the eardrum, rather than the other parts of the outer ear (of course, a more complete representation includes audio transmission through bone conduction and / or other mechanisms, which is basically ignored here). Therefore, any ambient noise at the conventional target position (as represented by the in-ear noise signal 125) is different from the ambient noise experienced at the eardrum (at the end of the third path). The performance of ambient noise at the eardrum is represented as an eardrum noise signal 325. The eardrum noise signal 325 is essentially the result of the entire main path (MPED) of the ambient noise passing through the eardrum, which includes the MPEM and the third path. In fact, any ambient noise detected at the conventional target position is also affected by the third path because the ambient noise is reflected back to the conventional target position from the surrounding of the ear cavity and other third path structures.

[0043] Conventional ANC implementations seek to minimize the residual noise signal 135 without considering the entire MPED. For example, a conventional ANC filter can be initially trained (e.g., in a laboratory environment, etc.) based on detecting the residual noise signal 135 with the error microphone 235 and feeding back the residual noise signal 135 for updating the ANC filter components, thereby optimizing the synthesis of the anti-noise signal 115. However, such conventional training is based only on optimizing the synthesis of the anti-noise signal 115 into the anti-noise path 120 and the MPEM, which ignores the third path effect of any wearer / wearing state. As described above, when an individual wears the earpiece of the headset, the specific characteristics of the wearer (ear anatomy and physiology) and the wearing position combine to form a specific wearer / wearing state. For example, each person may have a different ear shape, a different ear cavity geometry, etc. In addition, even a slight change in the position and / or orientation of the earpiece in the wearer's ear will be intended to change multiple characteristics of the MPED. For example, movement of the earpiece to a looser fit in the ear may result in changes in the seal and / or earpiece orientation, changes in the position of the error microphone 235 relative to the eardrum, slight changes in the shape of the ear anatomy, and / or other factors that may affect the frequency response of the MPEM and third path. Thus, each wearer / wearing state may have different associated frequency response characteristics.

[0044] When a wearer wears a headphone device equipped with ANC 110, ANC 110 needs to adaptively select ANC parameters and filters to achieve optimal noise reduction performance for a specific wearer / wearing state. It is further desired to ensure that the adaptation time is very short (e.g., within a few seconds) so that the user feels that the noise reduction performance is smooth and comfortable. ANC 110 needs to optimally adjust its filter parameters within a very short time, such as within a few seconds. This may involve designing the parameters of the ANC filter to quickly adapt to the entire main path, including the entire MPED in the novel embodiment of the present application. In addition, during the wearing and use of the headphones (i.e., over time), the wearing position changes, which changes the corresponding frequency response of the entire acoustic path and MPED. For example, the earpiece can become tighter or looser in the user's ear, the user can replace the earplug head, etc.

[0045] The novel embodiments described in this application have a hypothetical training phase and an operational phase. Before being released for consumer use, the training phase occurs in a laboratory and / or other controlled environment, and the operational phase occurs when the end consumer wears the earphones including the features described in this application. In the training phase, the training version of the earpiece is implemented using at least two microphones (or one or more microphones at each of the three positions): an error microphone 235 and an eardrum microphone 340. In some cases, the training also includes a reference microphone 105. The eardrum microphone 340 is one or more physical microphones placed at the eardrum, or other audio sensors. As used in this application, phrases such as "at the eardrum" refer to an actual or useful position that is in contact with, or close to, the eardrum, depending on the type of acoustic sensor used. The training version of the earpiece is worn by multiple individuals in one or more ways to obtain data representing different wearers / wearing states. For example, each of the multiple wearers can wear one or more earpieces in one or more ways (e.g., more or less tightly inserted, with different earpiece heads, oriented in slightly different ways, etc.). Wearers with very similar MPEDs will generate a candidate to reduce frequent switching of ANC filters.

[0046] The following description assumes the use of three microphones. For each wearer / wearing state, acoustic information is collected from the three microphones. For example, the reference ambient noise signal is recorded by all three microphones without any ANC function being active (e.g., in some cases, the training version of the earpiece may not even include the same ANC features as the production version). Therefore, the reference microphone 105 receives the "original" ambient noise signal (or directly inputs the reference noise signal without using the reference microphone 105), the error microphone 235 receives the first version of the ambient noise affected by the MPEM, and the eardrum microphone 340 receives the second version of the ambient noise affected by the MPED. The difference between the versions of the same ambient signal can be used to model (e.g., calculate a transfer function representation) the acoustic path effects of the MPEM and MPED, which can be further used to isolate the acoustic path effects of the third path. Based at least on the calculated model, a filter can be designed (e.g., a filter coefficient can be set) to represent the corresponding MPEM (which can substantially correspond to, or can be used to derive the corresponding anti-noise path 120) and the corresponding MPED for each wearer / wearing state. For example, the training phase generates K pre-trained ANC models 425, each with a corresponding one of the K pre-trained ANP models and a corresponding one of the K pre-trained MPED models. K may be any suitable integer, such as 50.

[0047] In the operation phase, the consumer wears the consumer version of the earpiece (in a specific wearer / wearing state), and the ANC function is operable. A first manifestation of ambient noise (in-ear noise signal 125) and a first manifestation of the synthesized anti-noise signal 115 are both present at a conventional target location, which are combined (e.g., modeled as a summing node 130) to generate a residual noise signal 135. The in-ear noise signal 125 is essentially ambient noise because it has passed through the MPEM, and the manifestation of the anti-noise signal 115 is essentially the anti-noise signal 115 because it has passed through the anti-noise path 120. At the same time, at the eardrum, there is a second manifestation of ambient noise (eardrum noise signal 325) and a second manifestation of the synthesized anti-noise signal 115, which are combined (e.g., modeled as a summing node 330) to generate an eardrum residual signal 335. The eardrum noise signal 325 is essentially the ambient noise because it has passed through the MPED (i.e., the MPEM plus the third path), and the second representation of the anti-noise signal 115 is essentially the anti-noise signal 115 because it has passed through the anti-noise path 120 and the ear cavity noise path (ECNP) 320. As described above, in any wearer / wearing state, the error microphone 235 is actually located in the wearer's outer ear, substantially at the beginning of the ECNP 320 (the end of the ECNP 320 is the eardrum). Therefore, when the anti-noise signal 115 and the in-ear noise signal 125 are combined at the conventional target location (when the earpiece is worn by the wearer), the resulting residual noise signal 135 passes through the ECNP 320 and interacts with the ECNP 320, so that the signal received by the error microphone 235 is actually the residual noise signal 135 affected by the ECNP 320. However, conventional ANC implementations are not pre-trained to distinguish between the MPEM effect and the MPED effect, which may undermine the ability of conventional ANC implementations to optimize noise cancellation for different wearers / wearing states.

[0048] If the earpiece provides a good acoustic seal, the frequency responses of the third path (TYSP) and the ECNP 320 should be substantially the same. Therefore, in a particular wearer / wearing state, it can generally be assumed that the difference between the eardrum residual signal 335 and the residual noise signal 135 is based on the same acoustic path effect as the difference between the eardrum noise signal 325 and the internal noise signal 125. For example, assume that the earpiece is currently worn in a particular wearer / wearing state (i.e., in the operating phase), and that the wearer / wearing state is the same as the kth wearer / wearing state pre-trained in the training phase. In the kth wearer / wearing state, any difference between the current residual noise signal 135 measured by the error microphone 235 and the current eardrum residual signal 335 (which cannot be directly measured in the operating phase configuration) should be accurately described by the kth pre-trained ANC model 425k.

[0049] For example, as referenced Figure 2 As described, the in-ear noise signal 125 can be estimated as: n=e+SPi*y, where "SPi" is the pre-calculated impulse response of SP, as implemented by the ANPi filter 220. Therefore, the eardrum noise signal 325 (ned) can be expressed as: ned=ECNP*n, where "ECNP" is the mathematical representation of ECNP 320. Figure 2 As described, the reference input 205w ("x") can be described as: x = SPi * xw, where "xw" is the reference noise signal 205w received by the reference microphone 105. Therefore, the WB-ANC subsystem 110w adaptive filtering algorithm reference signal (xed) can be described as: xed = TYSP * x. The residual noise in the posterior eardrum region of the WB-ANC subsystem 110w (eed) can be described as: eed = ned - WB * xed, where "WB" is the adaptive filtering coefficient of the WB-ANC subsystem 110w. WB can be updated according to the following formula: WB = WB + mu * eed * x, where "mu" is an adaptive coefficient, which can be a variable depending on which adaptive filtering algorithm is used.

[0050] Embodiments may implement the NB-ANC subsystem 110n after the WB-ANC subsystem 110w, and the filter of the NB-ANC subsystem 110n may be adaptively updated according to an adaptive algorithm. For example, the WB-ANC subsystem 110w is a filter designed to minimize the residual noise at the eardrum (i.e., the eardrum residual signal 335) without using the NB-ANC subsystem 110n, and wherein the reference noise signal 205w is an adaptive filter reference. The NB-ANC subsystem 110n is a filter designed to minimize the residual noise at the eardrum (i.e., the eardrum residual signal 335), and wherein its adaptive filter reference and the noise to be minimized are both based on the residual noise at the eardrum after using the WB-ANC subsystem 110w. Therefore, this adaptive algorithm independently performs two minimizations: the WB-ANC subsystem 110w attenuates the larger bandwidth noise, and the NB-ANC subsystem 110n further reduces the noise near the low frequency range after the WB-ANC subsystem 110w. The residual noise after the WB-ANC subsystem 110w is filtered to account for the anti-noise path 120 and the ECNP 320 can be described as: xned=SPi*TYSP*eed. The formula for xned can be used as a reference for the adaptation of the NB-ANC subsystem 110n. Thus, the output signal from the NB-ANC subsystem 110n, the NB anti-noise component 215n, can be described as: yned=NB*xned. Therefore, the residual noise after implementing the WB-ANC subsystem 110w and the NB-ANC subsystem 110n is eedn=eed-yned. The adaptive filter of the NB-ANC subsystem 110n can be updated with the following adaptive filtering algorithm: NB=NB+mu*eedn*xned.

[0051] As described above, training generates K pre-trained ANC models 425, each corresponding to a particular wearer / wearing state. Any of the K pre-trained ANC models 425 can be "applied" to the ANC 110 by adjusting filters (e.g., filter coefficients, filter selection, etc.) based on parameters defined by the pre-trained ANC model 425. For example, each of the pre-trained ANC models 425 can correspond to a respective ANC filter (e.g., a set of filter coefficients and / or other parameters) configured to implement the pre-trained ANC model 425, and applying the pre-trained ANC model 425 can involve effectively replacing the current ANC filter with the ANC filter corresponding to the pre-trained ANC model 425. Applying any k-th pre-trained ANC model 425k to the ANC 110 causes the ANC 110 to process the error microphone 235 (e.g., respond to feedback from the error microphone 235) as if the error microphone 235 was located on the eardrum corresponding to the k-th wearer / wearing state, even if the error microphone 235 is located at the same general target location. Therefore, by applying the kth pre-trained ANC model 425k to the ANC 110, adaptive noise reduction is performed by the ANC 110 in a manner that considers the acoustic path effects of the entire MPED (e.g., including the third path, ECNP 320, etc.) for the corresponding kth wearer / wearing state.

[0052] Figure 4 4 shows a simplified block diagram of an exemplary ECFR trained ANC architecture 400 in accordance with various embodiments described herein. Figure 1-3 , the ECFR-trained ANC architecture 400 may include an ANC 110 disposed between a reference microphone 105 and an error microphone 235. In general, the ANC 110 generates and outputs an anti-noise signal 115 based on one or more reference signals (e.g., a feedback reference signal generated based on a reference noise signal recorded by the reference microphone 105 and a residual noise signal 135 recorded by the error microphone 235). The reference microphone 105 receives a version of the ambient noise displayed outside the ear (e.g., a reference input 205w), and the error microphone 235 receives another version of the same ambient noise (the in-ear noise signal 125) after passing through the main path. As described above, without any ECFR-based pre-training, the ANC 110 cannot distinguish between the main path effect from the MPEM and the main path effect from the MPED, and therefore the ANC 110 cannot compensate for different wearers / wearing conditions.

[0053] As shown, the ECFR-trained ANC architecture 400 includes a user adaptation engine 410 and an ECFR-trained ANC model storage 420. The ECFR-trained ANC model storage 420 is any suitable data storage on which K pre-trained ANC models 425 are stored. As described above, each of the K pre-trained ANC models 425 corresponds to a specific wearer / wearing state. During consumer use, the user adaptation engine 410 is used to quickly determine and select one of the pre-trained ANC models 425 from the ECFR-trained ANC model storage 425 that is closest to the current wearer / wearing state. Then, an embodiment of the user adaptation engine 410 can apply (or directly apply) the selected pre-trained ANC model 425 to the ANC 110. The determination and selection of the user adaptation engine 410 can be based on the reference input 205w, the anti-noise signal 115, and the residual noise signal 135. After applying the determined pre-trained ANC model 425 to the ANC 110, the ANC 110 may be allowed to perform adaptive noise reduction, and such adaptive noise reduction will automatically take into account the entire MPED, not just the MPEM.

[0054] Depending on the ANC 110 architecture and / or other factors, the pre-trained ANC models 420 may be trained to provide different types of filter configurations. For example, for an ANC 110 that includes only the error microphone 235 (i.e., without the reference microphone 105, such as for a feedback-based ANC architecture), each pre-trained ANC model 420 may include a filter configuration representing a secondary path model and a filter configuration representing a NB-ANC subsystem 110n model. For an ANC 110 that includes both a reference microphone 105 and an error microphone 235 (e.g., a hybrid ANC architecture), each pre-trained ANC model 420 may include a filter configuration representing a secondary path model, a filter configuration representing a NB-ANC subsystem 110n model, and a filter configuration representing a WB-ANC subsystem 110w model. The filter configurations representing the NB-ANC subsystem 110n model and / or the WB-ANC subsystem 110w model are generally considered MPED models in this application.

[0055] Figure 5 FIG. 5 shows an exemplary implementation of a user adaptation engine 500 according to various embodiments. Figure 4 Similar to the user adaptation engine 410, Figure 5The user adaptation engine 500 takes as input the reference input 205w recorded by the reference microphone 105, the anti-noise signal 115 output by the ANC 110, and the residual noise signal 135 recorded by the error microphone 235. In some implementations, as shown, each of these input signals can pass through a corresponding downsampler 510. For example, a downsampled version of the reference input 205w (representing the reference ambient noise) is shown as "x(n)"; a downsampled version of the anti-noise signal 115 (representing the output of the ANC 110) is shown as "y(n)"; and a downsampled version of the residual noise signal 135 (representing the noise reduction error) is shown as "e(n)". For example, the ANC 110 can operate at 768 kHz, and the downsampler 510 can convert each signal to an 8 kHz signal, a 4 kHz signal, etc.

[0056] An embodiment of the user adaptation engine 500 may try each of the K pre-trained ANC models 425 to determine which pre-trained ANC model 425 most closely represents the current wearer / wearing state. In some embodiments, each of the K pre-trained ANC models 425 is tried iteratively (e.g., sequentially), one at a time. In other embodiments, multiple instances of the user adaptation engine 500 are used to parallelize the processing pipeline so that multiple of the K pre-trained ANC models 425 are tried in parallel in each iteration. In some such embodiments, the K parallel instances of the user adaptation engine 500 are used to try all K pre-trained ANC models 425 simultaneously. As described above, each k-th pre-trained ANC model 425k may include a k-th MPED model 530k, which defines filters, transfer functions, etc. of the MPED for the k-th wearer / wearing state, and a k-th ANP model 520k, which defines filters, transfer functions, etc. of the ANP (or MPEM) for the k-th wearer / wearing state. In this way, Figure 4 The ECFR-trained ANC model storage 425 may be considered to include an MPED model storage 535 having K MPED candidate models stored thereon and an ANP model storage 525 having K ANP candidate models stored thereon.

[0057] In each k-th trial, the user adaptation engine 500 seeks to estimate the actual main path noise (i.e., the actual performance of the ambient noise at the target location, including acoustic effects across the actual MPED) for the current wearer / wearing state; if the ambient noise is passing through the modeled MPED corresponding to the k-th pre-trained ANC model 425k, then estimate the candidate noise representing the main path noise; and determine the candidate residual noise r(n) as the difference between the estimated actual main path noise and the estimated candidate noise. The closer the actual MPED of the current wearer / wearing state is to the modeled MPED of a particular pre-trained ANC model 425, the smaller r(n) will be. In this way, the user adaptation engine 500 can select the pre-trained ANC model 425 that generates the smallest r(n). In some embodiments, the noise signal can be equalized to one of several pre-configured "typical" noise types before filter selection. This helps to take into account the huge differences in ambient noise from one location to another.

[0058] Embodiments of the user adaptation engine 500 can estimate the actual main path noise for the current wearer / wearing state by substantially removing the anti-noise signal 115 from the residual noise signal 135. For example, if there is no ANC 110, there will be no anti-noise signal 115, and the error microphone 235 will record the actual noise at the target location (i.e., the in-ear noise signal 125). However, since the ANC 110 is outputting the anti-noise signal 115, the error microphone 235 records the residual noise signal 135, which is a combination of the noise at the target location and the anti-noise signal 115. In addition, the anti-noise signal 115 as output at the target location is actually a version of the anti-noise signal 115 after passing through the anti-noise path 120. Therefore, in order to obtain an estimate of the actual noise at the target location in the kth trial, the anti-noise signal 115 (or a downsampled version, y(n)) is passed to the kth ANP model 520k to generate a path-compensated estimate of the anti-noise signal 115 as an output at the target location. At summing node 540 (e.g., implemented as an adder, subtractor, or in any suitable manner), the path compensated estimate of the anti-noise signal 115 is subtracted from the residual noise signal 135 (or downsampled version, e(n)) to effectively leave an estimate of the actual noise d(n). For example, d(n) can be described as e(n)-y(n)*S(k), where S(k) is a functional representation of the kth ANP model 520k.

[0059] To estimate the candidate noise, an embodiment may pass the reference input 205w (or a downsampled version, x(n)) to the kth MPED model 530k, thereby converting the ambient noise (as represented by the reference input 205w) into an estimated eardrum noise signal based on the modeled acoustic path effects of the entire MPED passing through the wearer / wearing state corresponding to the kth pre-trained ANC model 425k. For example, the output of the kth MPED model 530k is x(n)*W(k), where W(k) is a functional representation of the kth MPED model 530k. At a summing node 550 (e.g., implemented as an adder, subtractor, or in any suitable manner), an embodiment combines the output of the kth MPED model 530k with the actual noise estimate d(n) to generate a candidate residual noise, r(n).

[0060] Some embodiments estimate the actual noise (d) and the candidate residual noise (r) according to an algorithm, such as the following:

[0061]

[0062]

[0063] K trials (e.g., performed serially, partially in parallel, or fully in parallel) may result in the generation of Kr(n) signals. Each of the Kr(n) signals is associated with an index (e.g., K) corresponding to which of the K pre-trained ANC models 425 generated the r(n). An embodiment of the user adaptation engine 500 includes an index selector 560 to output a selected index 565, which indicates the index of one of the K pre-trained ANC models 425 selected to be applied to the ANC 110 based on determining which of the K pre-trained ANC models 425 generated the minimum r(n). The minimum r(n) can be determined by determining which of the Kr(n) signals represents the smallest minimum mean square error. As described above, the selected index 565 corresponds to any pre-trained ANC model 425 having a modeled MPED that most closely matches the MPED of the current wearer / wearing state. By applying the selected pre-trained ANC model 425 (corresponding to the selected index 565) to the ANC 110, as Figure 4 As shown, the ANC 110 can achieve improved performance in both the noise reduction depth in the low frequency range and the width of the noise reduction frequency.

[0064] The user adaptation engine 500 may apply the selected pre-trained ANC model 425 in any suitable manner, or directly to the ANC 110. In some implementations, the user adaptation engine 500 sends one or more control signals to the ANC 110 (e.g., to the WB-ANC subsystem 110w and / or the NB-ANC subsystem 110n) to adjust the ANC filter coefficients and / or otherwise configure the ANC filter based on the selected pre-trained ANC model 425. In other implementations, the user adaptation engine 500 sends one or more control signals to the ANC 110 to indicate the selected index 565, and the ANC 110 maps the selected index 565 to a set of stored filter coefficients and / or other filter configurations accordingly. For example, the ANC 110 includes a lookup table that maps each of the K indexes to a corresponding one of the K sets of ANC filter configuration parameters.

[0065] The above-described features of the user adaptation engine 500 may generally be referred to as "filter selection". In some embodiments, the user adaptation engine 500 performs filter selection as part of the initial setup. For example, the wearer inserts one or both earpieces into their ears and begins playing the desired audio. Just before or after starting playback of the desired audio, the user adaptation engine 500 may perform index selection to find the best initial ANC filter settings for the ANC 110. The ANC 110 may then dynamically adapt to changes in ambient noise. In some embodiments, the user adaptation engine 500 performs filter selection in response to an explicit request from the user. For example, the user presses a button on the headset, triggering filter selection.

[0066] In some embodiments, the user adaptation engine 500 continuously performs filter selection throughout the use of the headset, for example continuously or periodically. The use of the downsampler 510 can significantly reduce the amount of power, time, computing resources and / or other resources required to perform filter selection, so that the embodiment can continuously re-execute filter selection without placing a significant burden on the headset system. During use, the wearer may consciously adjust the position of the earpiece in the ear (or close the earplug head, etc.), or the position of the earpiece in the ear may move inadvertently (for example, become loose during use, etc.). By continuously re-executing filter selection, the user adaptation engine 500 can continue to ensure that any of the pre-trained ANC models 425 that serve as the basis for the ANC 110 filter configuration remains optimal when the wearer / wearing state changes.

[0067] In some such embodiments, whenever the filter selection determines that the currently selected index 565 is not optimal, a newly selected index 565 is applied to the ANC 110. For example, R(k) represents the candidate residual noise for a particular index (k) and D(k) represents the actual noise estimate for the same index (k). The filter selection of the user adaptation engine 500 seeks to minimize the residual error rate i(k), where i(k)=R(k) / D(k). Assume that the currently selected index 565 is 'p' and the index in each re-executed trial is 'n'; so that the current residual error rate is i(p)=R(p) / D(p) and the residual error rate for each new attempt is i(n)=R(n) / D(n). Some embodiments may select a newly attempted index n when i(n) is less than i(p). For example, a lower residual error rate indicates that the pre-trained ANC model 425 corresponding to the newly attempted index is closer to the actual MPED for the current wearer / wearing state than the MPED represented by the current ANC filter configuration. Other embodiments may select a new index n to try whenever i(n) is less than i(p) by at least a predetermined threshold. As an example, assume that the ANC 110 is currently configured according to a pre-trained ANC model 425 with index '5', and re-performing the index selection results in the pre-trained ANC model 425 with index '10' now presenting a better candidate residual noise. An embodiment may determine to switch the selected index 565 from '5' to '10' and may update the ANC 110 accordingly.

[0068] In some implementations, updating the index may be performed abruptly. In the above example, the ANC 110 filter configuration associated with the selected index 565 '5' is abruptly replaced by the ANC 110 filter configuration associated with the selected index 565 '10'. In other implementations, updating the index may be performed in a gradual and / or smooth manner. Figure 6 An example operational block diagram 600 is shown for switching from one ANC filter configuration to another ANC filter configuration based on updating a selected index 565 in response to a filter selection by a user adaptation engine 500. As described above, the filter coefficient sets (e.g., generally intended to include any suitable filter configuration parameters) are defined by pre-trained ANC models 425, and each pre-trained ANC model 425 is associated with an index. Figure 6The context assumes that a previous filter selection was performed by the user adaptation engine 500, thereby generating a current (old) selected index 565 corresponding to a selected one of the pre-trained ANC models 425. Prior to any change in filter selection, the ANC 110 filter configuration was set according to at least the set of filter coefficients defined by the “old” selected index 565. The context further assumes that the user adaptation engine 500 has now performed a subsequent filter selection, thereby generating a newly selected index 565 corresponding to one of the newly selected pre-trained ANC models 425, and that it is desired to update the ANC 110 filter configuration according to the newly selected pre-trained ANC model 425.

[0069] As shown, the divergence estimator block may generate a divergence flag (div_flag) 615 based on inputs from the error microphone 235 (e.g., the residual noise signal 135) and the reference microphone 105 (e.g., the reference input 205w). As shown, one of the inputs may be a "leakage indicator," which may be a ratio between the error microphone 235 and the reference microphone 105 inputs. For example, when the residual noise signal 135 increases to a threshold level (err_th) such that the leakage indicator is above a different threshold level (li_th), the div_flag 615 may be asserted (e.g., set to "1," set to HIGH, etc.), indicating that the noise correction being performed by the ANC 110 is tending to diverge. When the residual noise signal 135 drops below err_th such that the leakage indicator is below li_th, the div_flag 615 may be de-asserted (e.g., set to "0," set to LOW, etc.), indicating that the noise correction being performed by the ANC 110 is tending to stabilize.

[0070] The old filter coefficient releaser block 620 may take the current (old) ANC filter coefficients (defined by the current pre-trained ANC model 425p), the new ANC filter coefficients (defined by the new pre-trained ANC model 425n), the divergence flag 615, and the release time as input. The old filter coefficient releaser block 620 may operate differently depending on the state of the divergence flag 615 received from the divergence estimator block 610. If there is no divergence, an embodiment may replace the poles and zeros in the old ANC filter configuration as long as the poles and zeros of the new ANC filter configuration are close to the poles and zeros of the old ANC filter configuration. Other poles and zeros of the old ANC filter configuration may be replaced with zeros. After replacement, the ANC 110 may maintain this state until the release time has passed. If there is divergence, an embodiment may replace all poles and zeros in the old ANC filter configuration with zeros. After replacement, the ANC 110 may maintain this state until the release time has passed.

[0071] For example, once the divergence is detected by the divergence estimator block 610, the current ANC filter coefficients are released and the adaptation is frozen to prevent divergence howling. Ideally, the ANC filter state can be set to zero, especially when using an infinite impulse response (IIR) filter, because the zero input response may persist. If the filter state is not accessible, the ANC 110 can wait until the zero input response decays. When the noise pump decays so that the leakage indicator is less than the threshold, the ANC adaptation can be started or restarted. When the filter selection is completed and there is a new selected index 565, the ANC 110 can be updated with a new filter configuration corresponding to the new selected index 565. When it is uncertain whether the state of the ANC 110 is released for the new ANC filter configuration, the input of the ANC 110 can be gradually increased to a normal level. The divergence caused by the ANC filter switching is typically a product of the NB-ANC subsystem 110n, which filters based on feedback and performs noise reduction primarily on low frequencies. Embodiments of the NB-ANC subsystem 110 may be designed such that the filter is a high performance low pass filter that has a stop band in any frequency region where noise is not attenuated.

[0072] The new filter adder and smoother block 630 may take as input the new ANC filter coefficients and the smoothing time. In one implementation, the new filter adder and smoother block 630 may write the new coefficients to the ANC filter positions, gradually increasing the ANC input signal level to a normal level over the smoothing time. In another implementation, the new filter adder and smoother block 630 maintains a release period at the smoothing time, after which the new filter coefficients are available to the ANC 110. In another implementation, the new filter adder and smoother block 630 gradually amplifies the ANC filter coefficients to a specified value over the smoothing time.

[0073] The above methods generally obtain and utilize a pre-trained ANC model 420, including using downsampled filter selection to reduce the cost and complexity of filter adaptation of ANC 110. Other methods may use some similar techniques. A more costly alternative is to train ANC 110 directly. As described in the present application, noise reduction performance can be optimal when the ANC filter configuration is trained for a wearer / wearing state that substantially matches the current wearer / wearing state. If not, the ANC filter configuration can be selected as the initial best option relative to other candidate configurations, and adaptive filtering running in real time can be applied to further improve the performance of ANC 110. The method can apply all ANC 110 parameters and filters to ANC 110 input data without using the pre-trained ANC model 420, while disabling ANC 110. For example, if there are K potential filter configurations, each configuration can be tried in ANC 110 to generate K corresponding residual noise signals, and the various residual noise signals can be compared to find the configuration that provides the best ANC 110 performance. As described above, performance may be evaluated based on factors such as optimizing the depth of noise reduction near low frequencies, the width of the noise reduction frequencies, and the total noise reduction. Using this approach, it is possible to minimize the time that the ANC is disabled for filter selection to limit the impact on the wearer's comfort and listening experience. Because different filter configurations are being tried at full rate (i.e., at the first sampling rate of the ANC 110), trying a large number of different candidate configurations may be expensive (i.e., may require significant power, computational, and / or time resources).

[0074] Another alternative approach is to train the ANC 110 while the wearer is wearing the device. Using this approach, it may be necessary to ensure that the training time is very short, such as 1 or 2 seconds. With this approach, the ANC 110 anti-noise signal 115 output can be disabled in order to train the secondary path impulse response (e.g., ANPi220). Instead, the ANC 110 can transmit a piece of music, pink noise, or other signal within a specified first training time window (e.g., approximately 1 second). Specific data segments can also be used to train filter coefficients to represent a secondary path model (e.g., similar to the ANP model 520). In order to train the ANC filter impulse response (e.g., WB and / or NB filters, similar to the MPED model 530), the error microphone 235 and / or the reference microphone 105 can be used to capture ambient noise within a second training time window (e.g., approximately 1 second) while disabling the ANC 110 anti-noise signal 115 output.

[0075] Some such methods may train the ANC 110 in a frequency bend. The frequency bend is designed to have a higher resolution for low frequency components, while the high frequency components have a lower resolution. In this way, a few frequency bins may be sufficient to generate the ANC filter. For example, the frequency response of each data block (e.g., 4ms) with a selected frequency bin may be calculated in real time. A time domain window may be applied when calculating the frequency response to reduce response variations. Such block-based calculations may tend to distribute the computational burden over the blocks.

[0076] Other such methods can train the ANC 110 directly in the time domain. The obtained time domain data can be downsampled and input into the adaptive algorithm. For each data, the adaptive algorithm is repeatedly applied to ensure the timing of the adaptive filtering is normal. A very good low sampling frequency finite impulse response ANC filter can be obtained. The ANC 110 can include an algorithm module that converts the ANC 110 from a low sampling frequency to a high sampling frequency, and also has an algorithm module that converts the FIR filter to an IIR filter. Such an approach can generate a filter that is relatively smooth and reliable in the frequency domain, because the adaptive filtering algorithm can reduce the interference components in the input signal under appropriate conditions.

[0077] Other such methods may train the ANC 110 by converting the input data into frequency warped time domain data. The amount of data generated using this method may be much smaller, and the warped frequency FIR ANC length may be very short. These methods are compatible with the same type of adaptive algorithms as the other methods. The ANC 110 may have an algorithm module for converting the ANC 110 from a warped frequency to a high sampling frequency, and also have an algorithm module for converting a warped frequency FIR to an IIR.

[0078] Figure 7 A flowchart of an exemplary method 700 for pre-training an ANC model based on ear cavity frequency response according to various embodiments described herein is shown. As described herein, some pre-training can be performed in a laboratory environment using a training configuration of an earphone earpiece that includes a physical eardrum microphone for being positioned at the eardrum (or an artificial simulation of the outer ear and eardrum) of an individual wearing the device. An embodiment of method 700 begins at stage 704, positioning the training configuration of the device in a wearer / wearing state. Assume that method 700 will be performed for K (e.g., 50) wearers / wearing states, so that the stages of method 700 represent the kth execution of the method corresponding to the kth wearer / wearing state.

[0079] At stage 708, when the device is positioned at the kth wearer / wearing state, the ANC filter can be trained to compensate for the entire MPED (i.e., the kth MPED generated by the kth wearer / wearing state). Such training can involve training the anti-noise path filter, as well as training the feedback (narrowband) ANC filter and / or the feedforward (wideband) ANC filter. At stage 712, an embodiment can use the trained ANP filter to generate the kth ANP filter model. At stage 716, an embodiment can use the trained feedback and / or feedforward ANC filter model to generate the kth MPED filter model. At stage 720, the kth ANP filter model and the kth MPED filter model are stored in association with the kth wearer / wearing state (e.g., using an index). For example, in subsequent production use, any specific ANC filter configuration defined by a corresponding one of the K pre-trained ANC models (including its Kth ANP filter model and its Kth MPED filter model) can be identified by its index and applied to set the filter configuration of the ANC system of the earpiece.

[0080] Figure 8 A flowchart of an exemplary method 800 for adaptive noise reduction (ANC) of an ear cavity frequency response (EFCR) with path compensation on the entire main path (MPED) to the eardrum according to various embodiments described in the present application is shown. An embodiment of method 800 operates to display the current wearer / wearing state relative to the user's outer ear when the user wears the earphone. For example, the user has placed the earphone in at least one of his or her ears, and the user may or may not be currently playing the desired audio. The embodiment starts at stage 804 by receiving a set of input signals. One input signal is a reference noise signal (e.g., reference input 205w), which represents the ambient noise of the earphone toward the outside. Another input signal is an anti-noise signal (e.g., anti-noise signal 115) generated by the ANC system of the earphone, which uses the current ANC filter configuration to eliminate the in-ear noise signal corresponding to the ambient noise, which is manifested at a target position toward the inside of the earphone due to the acoustic path effect of the MPED. Another input signal is a residual noise signal (e.g., residual noise signal 135) representing a combination of the anti-noise signal and the in-ear noise signal. In some embodiments, the residual noise signal is received by an error microphone disposed on the inner side of the earphone. In some embodiments, the reference noise signal is received by a reference microphone disposed on the outer side of the earphone.

[0081] An embodiment of method 800 continues filter selection at stages 808-816. Filter selection may be performed on K trials corresponding to K pre-trained ANC models, where K is a positive integer greater than 1. Typically, K is much greater than 1, for example, where K is 50, 100, 1000, etc. In some implementations, stages 808-816 are performed sequentially K times, once for each of the K pre-trained ANC models. In other implementations, stages 808-816 are performed in parallel on all K pre-trained ANC models, so that a single iteration of stages 808-816 includes K concurrent executions of stages 808-816. In other implementations, each iteration of stages 808-816 is performed in parallel on multiple but not all pre-trained ANC models, so that more than 1 and less than K iterations are performed.

[0082] Each K-th pre-trained ANC model includes a K-th anti-noise path (ANP) filter model and a k-th MPED filter model previously trained in the K-th trained wearer / wearing state. As described in the present application, some embodiments of the ANC system include a feedback ANC filter subsystem that generates a narrow-band (NB) component of the anti-noise signal by converting a feedback reference signal based on a set of NB filter coefficients. The feedback reference signal is generated from a residual noise signal detected by an error microphone of an earpiece of an earphone substantially located at a target position. In such an embodiment, each k-th MPED filter model includes a k-th feedback ANC filter model that defines a k-th group of values ​​for configuring a set of NB filter coefficients. Other embodiments of the ANC system also include a feedforward ANC filter subsystem that generates a wide-band (WB) component of the anti-noise signal by converting a reference noise signal based on a set of WB filter coefficients. In such an embodiment, each k-th MPED filter model also includes a k-th feedforward ANC filter model that defines a k-th group of values ​​for configuring the set of WB filter coefficients. For example, in the context of a hybrid ANC architecture, each of the K pre-trained ANC models may define an ANP filter model and an MPED filter model, including a feedback ANC filter model and a feedforward ANC filter model.

[0083] As described in the present application, each of the K pre-trained ANC models is pre-trained based on a training configuration using an earphone in each of the K training wearers / wearing states. In the operating environment of method 800, the user wears a production configuration of an earphone that does not include an eardrum microphone (e.g., it typically includes an error microphone, and it may also include a reference microphone). Instead, the training configuration of the earphone includes an error microphone at a target position and an eardrum microphone for positioning at the eardrum of the training wearer in each k-th trained wearer / wearing state. In some cases, the training wearers include one or more human users, and the training configuration of the earphone is placed in the outer ear of one or more human users to display each training wearer / wearing state. In other cases, the training wearer is a true simulator of the outer ear that closely mimics the outer ear anatomy and physiology (including the anatomy and physiology of the eardrum) of one or more actual human users.

[0084] Since the training configuration of the earphone earpiece includes an error microphone and an eardrum microphone, each k-th MPED filter model is trained based on direct concurrent measurements of the same reference noise performance of the error microphone at a target position and the eardrum microphone at the eardrum. For example, in each k-th trained wearer / wearing state, the error microphone is positioned at a corresponding training target position of the outer ear of the corresponding training wearer, and the eardrum microphone is positioned in the eardrum of the corresponding training wearer. This can define a k-th ear cavity noise path between the corresponding training target position and the eardrum of the corresponding training wearer, so that each k-th MPED filter model is trained to compensate for the ECFR effect of the k-th ear cavity noise path.

[0085] At stage 808, an embodiment may calculate the kth estimated actual noise signal using the kth ANP filter model to convert the anti-noise signal into a path-compensated anti-noise signal, and an embodiment may remove the path-compensated anti-noise signal from the residual noise signal. At stage 812, an embodiment may calculate the kth estimated eardrum noise signal using the kth MPED filter model to convert the reference noise signal. At stage 816, an embodiment may calculate the kth candidate residual noise between the kth estimated actual noise signal and the kth estimated eardrum noise signal.

[0086] The ANC system of the earphone operates at a first sampling rate. In some embodiments, the method 800 also includes downsampling the input signal at stage 806 before performing the filter selection. In particular, such an embodiment includes downsampling the reference noise signal from the first sampling rate of the ANC system to a second sampling rate to generate a downsampled reference noise signal; downsampling the anti-noise signal from the first sampling rate to the second sampling rate to generate a downsampled anti-noise signal; and downsampling the residual noise signal from the first sampling rate to the second sampling rate to generate a downsampled residual noise signal. The second sampling rate is much smaller than the first sampling rate, such as less than five percent of the first sampling rate. For example, the first sampling rate is around 800 kHz and the second sampling rate is around 8 kHz (about 1%), or even 4 kHz (about 0.5%). In these embodiments, every k-th ANP filter model and every k-th MPED filter model operate at the second sampling rate. The kth estimated actual noise signal can be calculated using the kth ANP filter model to convert the downsampled anti-noise signal into a path-compensated anti-noise signal, and the path-compensated anti-noise signal can be removed from the downsampled residual noise signal. The kth estimated eardrum noise signal can be calculated using the kth MPED filter model to convert the downsampled reference noise signal.

[0087] At stage 820, an embodiment may select one of the pre-trained ANC models as generating the lowest k-th candidate residual noise. At stage 824, an embodiment may guide the ANC system of the earphone to replace the current ANC filter configuration with a new ANC filter configuration based on one of the selected pre-trained ANC models. In some embodiments, at stage 818, method 800 may calculate a current residual error ratio based on a ratio of the current candidate residual noise to the current estimated actual noise signal, the current candidate residual noise and the current estimated actual noise signal being associated with the current ANC filter configuration. In these embodiments, the selection at stage 820 may include, for each k-th pre-trained ANC model ratio, calculating the k-th residual error ratio based on the ratio of the k-th candidate residual noise to the k-th estimated actual noise signal, and selecting one of the pre-trained ANC models as the model generating the lowest residual ratio. In addition, in such embodiments, the guidance of stage 824 may be performed in response to detecting that the lowest residual error ratio is less than the current residual error ratio. In some such embodiments, the detection of stage 824 may be performed in response to detecting that the lowest residual error ratio is less than the current residual error ratio by at least a predetermined threshold amount.

[0088] In some embodiments, the directing at stage 824 includes freezing the adaptation of the ANC system for a predetermined release time; before the release time elapses, releasing the current filter coefficients defined by the current ANC filter configuration; and after the release time elapses, writing the new filter coefficients to the ANC system defined by the new ANC filter configuration, and gradually unfreezing the adaptation of the ANC system according to the new ANC filter configuration within a predetermined smoothing time. In some such embodiments, the gradual unfreezing adaptation may be performed by gradually increasing the ANC input signal level to a normal level (the normal level is the ANC input signal level before freezing) within the smoothing time, or by gradually scaling from the released filter coefficient values ​​to the values ​​defined by the new filter coefficients within the smoothing time. In some such embodiments, releasing the current filter coefficients may include determining whether a divergence condition exists by determining whether a residual noise signal currently exceeds a predetermined error threshold level, and whether a leakage indicator currently exceeds a predetermined leakage threshold level, the leakage indicator being a ratio signal between the residual noise signal and a reference noise. If a divergence condition exists, these embodiments may replace the old poles / zeros defined by the current ANC filter configuration with all zeros. If a divergence condition does not exist, then for each old pole / zero point, such an embodiment may: replace the old pole / zero point with one of the corresponding new poles / zero points defined by the new ANC filter configuration if the old pole / zero point is within a predetermined threshold distance of one of the corresponding new poles / zero points; and replace the old pole / zero point with a zero point if the old pole / zero point is not within a predetermined threshold distance of one of the corresponding new poles / zero points.

[0089] The scope of the embodiments disclosed in the present application is not limited by the specific embodiments described in the present application. According to the foregoing description and the accompanying drawings, in addition to what is described in the present application, various modifications to the embodiments of the present invention are apparent to those of ordinary skill in the art. In addition, although some embodiments of the present invention have been described in the context of specific implementations for specific purposes in specific environments, those of ordinary skill in the art will recognize that their use is not limited thereto, and embodiments of the present invention can be beneficially implemented in any number of environments for any number of purposes.

Claims

1. An adaptive active noise reduction method for mixed environment noise reduction with ear cavity frequency response compensation, wherein: The method comprises: receiving, when a user wears an earphone to display a current wearer / wearing state relative to an outer ear of the user, a reference noise signal representing ambient noise outwardly directed toward the earphone, an anti-noise signal generated by an ANC system of the earphone using a current ANC filter configuration to cancel an in-ear noise signal corresponding to the ambient noise, the ambient noise being displayed at a target position inwardly directed toward the earphone due to an acoustic path effect of a main path MPED to an eardrum, and a residual noise signal representing a combination of the anti-noise signal and the in-ear noise signal; For each K-th pre-trained ANC model among K pre-trained ANC models, each having a K-th anti-noise path ANP filter model and a k-th MPED filter model previously trained on the K-th trained wearer / wearing state, K is a positive integer greater than 1: calculating a k-th estimated actual noise signal using a k-th ANP filter model to convert the anti-noise signal into a path-compensated anti-noise signal, and removing the path-compensated anti-noise signal from the residual noise signal; calculating a kth estimated eardrum noise signal using the kth MPED filter model to convert the reference noise signal; and Calculating a kth candidate residual noise between the kth estimated actual noise signal and the kth estimated eardrum noise signal; selecting one of the pre-trained ANC models as the model that generates the lowest k-th candidate residual noise; and The ANC system of the earpiece is directed to replace the current ANC filter configuration with a new ANC filter configuration based on the selected one of the pre-trained ANC models.

2. The method according to claim 1, wherein: The ANC system of the earphone operates at a first sampling rate and further includes: downsampling the reference noise signal from a first sampling rate of the ANC system to a second sampling rate that is less than five percent of the first sampling rate to generate a downsampled reference noise signal; downsampling the anti-noise signal from the first sampling rate to the second sampling rate to generate a downsampled anti-noise signal; and downsampling the residual noise signal from the first sampling rate to the second sampling rate to generate a downsampled residual noise signal, Wherein, each kth ANP filter model and each kth MPED filter model operate at the second sampling rate, the kth estimated actual noise signal is calculated using the kth ANP filter model to convert the downsampled anti-noise signal into the path-compensated anti-noise signal, and the path-compensated anti-noise signal is removed from the downsampled residual noise signal, and the kth estimated eardrum noise signal is calculated using the kth MPED filter model to convert the downsampled reference noise signal.

3. The method according to claim 1, wherein: The user wears a production configuration of an earphone receiver that does not include an eardrum microphone; The K pre-trained ANC models were previously trained using a training configuration of the earphone earpiece, wherein the training configuration of the earphone earpiece includes an error microphone at the target position and an eardrum microphone positioned at the eardrum of the corresponding training wearer for each k-th trained wearer / wearing state, such that each k-th MPED filter model is trained based on direct concurrent measurements of the same reference noise representation of the error microphone at the target position and the eardrum microphone at the eardrum.

4. The method according to claim 3, wherein: In every k-th training wearer / wearing state, the error microphone is positioned at the corresponding training target position of the corresponding training wearer's outer ear, and the eardrum microphone is positioned in the eardrum of the corresponding training wearer, thereby defining the k-th ear cavity noise path between the corresponding training target position and the eardrum of the corresponding training wearer, so that every k-th MPED filter model is trained to compensate for the ECFR effect of the k-th ear cavity noise path.

5. The method according to claim 1, wherein: The ANC system of the earpiece includes a feedback ANC filter subsystem that generates a narrowband NB component to the anti-noise signal by transforming a feedback reference signal based on a set of NB filter coefficients, the feedback reference signal being generated by the residual noise signal detected by an error microphone of the earpiece substantially located at the target location; and Every k-th MPED filter model includes a k-th feedback ANC filter model defining a k-th set of values ​​for configuring the set of NB filter coefficients.

6. The method according to claim 5, wherein: The ANC system of the earphone also includes a feed-forward ANC filter subsystem that generates a wideband WB component to the anti-noise signal by transforming the reference noise signal based on a set of WB filter coefficients; and Every k-th MPED filter model also includes a k-th feedforward ANC filter model, which defines a k-th set of values ​​for configuring the set of WB filter coefficients.

7. The method according to claim 1, further comprising: calculating a current residual error ratio based on a ratio of a current candidate residual noise to a current estimated actual noise signal, the current candidate residual noise and the current estimated actual noise signal being associated with the current ANC filter configuration, wherein selecting one of the pre-trained ANC models as the model that generates the lowest k-th candidate residual noise comprises, for each k-th pre-trained ANC model, calculating a k-th residual error ratio based on a ratio of the k-th candidate residual noise to the k-th estimated actual noise signal, and selecting one of the pre-trained ANC models as the model that generates the lowest error ratio, and wherein directing the ANC system of the earpiece to replace the current ANC filter configuration with a new ANC filter configuration based on the selected one of the pre-trained ANC models is performed in response to detecting that a lowest residual error ratio is less than the current residual error ratio.

8. The method according to claim 7, wherein: in, The directing the ANC system of the earpiece to replace the current ANC filter configuration with a new ANC filter configuration based on the selected one of the pre-trained ANC models is performed in response to detecting that the lowest residual error ratio is less than the current residual error ratio by at least a predetermined threshold amount.

9. The method according to claim 1, wherein: Directing the ANC system of the earpiece to replace the current ANC filter configuration with a new ANC filter configuration based on the selected one of the pre-trained ANC models comprises: freezing the adaptation of the ANC system to a predetermined release time; before the release time elapses, releasing current filter coefficients defined by the current ANC filter configuration; and After the release time has elapsed, new filter coefficients are written to the ANC system defined by the new ANC filter configuration, and adaptation is gradually unfrozen with the ANC system according to the new ANC filter configuration within a predetermined smoothing time.

10. The method according to claim 9, wherein: The gradual thawing adaptation includes one of the following: gradually increasing the ANC input signal level to a normal level within the smoothing time, the normal level being the ANC input signal level before the freezing; or Gradually scaling from the released filter coefficient values ​​to the values ​​defined by the new filter coefficients within the smoothing time.

11. The method according to claim 9, wherein: Release the current filter coefficients including: determining whether a divergence condition exists by determining whether the residual noise signal currently exceeds a predetermined error threshold level and whether a leakage indicator currently exceeds a predetermined leakage threshold level, the leakage indicator being a ratio signal between the residual noise signal and the reference noise; If a divergence condition exists, replacing old poles / zeros defined by the current ANC filter configuration with all zeros; and If there is no divergence condition, for each of the old poles / zeros: replacing the old pole / zero with one of the corresponding new poles / zeros defined by the new ANC filter configuration if the old pole / zero is within a predetermined threshold distance of one of the corresponding new poles / zeros; and If the old pole / zero point is not within a predetermined threshold distance of one of the corresponding new pole / zero points, the old pole / zero point is replaced with a zero point.

12. The method according to claim 1, wherein: The residual noise signal is received by an error microphone arranged on the inner side of the earphone, or by a reference microphone arranged on the outer side of the earphone.

13. The method according to claim 1, wherein: K is at least 50.

14. An adaptive active noise reduction system for mixed environment noise reduction with ear cavity frequency response compensation, wherein: The adaptive active noise reduction system comprises: an ANC filter system for outputting an anti-noise signal at a target location based on a current ANC filter configuration, such that the anti-noise signal output at the target location is an inverse estimate of an in-ear noise signal that corresponds to ambient noise when the user wears the headphone earpiece in a current wearer / wearing state relative to the user's outer ear, as indicated at the target location, the target location being located on an inwardly facing side of the headphone earpiece; an error microphone for recording a residual noise signal representing a combination of the anti-noise signal and the in-ear noise signal for the current wearer / wearing state; A model data memory having K pre-trained ANC models stored thereon, each K-th pre-trained ANC model having a K-th anti-noise path ANP filter model and a k-th MPED filter model previously trained under a K-th trained wearer / wearing state, K being a positive integer greater than 1; A user adaptation engine is coupled to the ANC filter system, the error microphone and the model data memory for: For each k-th pre-trained ANC model, calculating a k-th estimated actual noise signal using a k-th ANP filter model to convert the anti-noise signal into a path-compensated anti-noise signal, and removing the path-compensated anti-noise signal from the residual noise signal; For each k-th pre-trained ANC model, calculating a k-th estimated eardrum noise signal using the k-th MPED filter model to convert a reference noise signal; and For each k-th pre-trained ANC model, calculating a k-th candidate residual noise between the k-th estimated actual noise signal and the k-th estimated eardrum noise signal; selecting one of the pre-trained ANC models as the model that generates the lowest k-th candidate residual noise; and The ANC filter system is directed to replace the current ANC filter configuration with a new ANC filter configuration based on the selected one of the pre-trained ANC models.

15. The adaptive active noise reduction system according to claim 14, further comprising: A reference microphone is disposed on the outer side of the earphone receiver and is used to measure the reference noise signal.

16. The adaptive active noise reduction system according to claim 14, wherein: The ANC filter system operates at a first sampling rate; The user adaptation engine further comprises a downsampler for: downsampling the reference noise signal from the first sampling rate to a second sampling rate that is less than five percent of the first sampling rate to generate a downsampled reference noise signal; downsampling the anti-noise signal from the first sampling rate to the second sampling rate to generate a downsampled anti-noise signal; as well as downsampling the residual noise signal from the first sampling rate to the second sampling rate to generate a downsampled residual noise signal; Every k-th ANP filter model and every k-th MPED filter model runs at the second sampling rate; The user adaptation engine calculates the kth estimated actual noise signal using the kth ANP filter model to convert the downsampled anti-noise signal into the path-compensated anti-noise signal, and removes the path-compensated anti-noise signal from the downsampled residual noise signal; as well as The user adaptive engine calculates the kth estimated eardrum noise signal using the kth MPED filter model to convert the downsampled reference noise signal.

17. The adaptive active noise reduction system according to claim 14, wherein: The ANC filter system further includes a feedback ANC filter subsystem that generates a narrowband NB component to the anti-noise signal by transforming a feedback reference signal based on a set of NB filter coefficients; and Every k-th MPED filter model includes a k-th feedback ANC filter model defining a k-th set of values ​​for configuring the set of NB filter coefficients.

18. The adaptive active noise reduction system according to claim 17, wherein: The ANC system of the earphone also includes a feed-forward ANC filter subsystem that generates a wideband WB component to the anti-noise signal by transforming the reference noise signal based on a set of WB filter coefficients; and Every k-th MPED filter model also includes a k-th feedback ANC filter model, which defines a k-th set of values ​​for configuring the set of WB filter coefficients.

19. The adaptive active noise reduction system according to claim 14, wherein: The user adaptation engine is further configured to calculate a current residual error ratio based on a ratio of a current candidate residual noise to a current estimated actual noise signal, the current candidate residual noise and the current estimated actual noise signal being associated with the current ANC filter configuration; The user adaptation engine is used to select one of the pre-trained ANC models as the model that generates the lowest k-th candidate residual noise, for each k-th pre-trained ANC model, calculate a k-th residual error ratio based on a ratio of the k-th candidate residual noise to the k-th estimated actual noise signal, and select one of the pre-trained ANC models as the model that generates the lowest error ratio; and The user adaptation engine directs the ANC filter system to replace the current ANC filter configuration with a new ANC filter configuration based on a selected one of the pre-trained ANC models in response to detecting that a lowest residual error ratio is less than the current residual error.

20. The adaptive active noise reduction system according to claim 19, wherein: The user adaptation engine directs the ANC filter system to replace the current ANC filter configuration with a new ANC filter configuration based on a selected one of the pre-trained ANC models in response to detecting that the lowest residual error ratio is less than the current residual error ratio by at least a predetermined threshold amount.

Images