BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention relates to noise reduction/cancellation, and more particularly, to an adaptive active noise control system with unstable state handling and an associated method.
2. Description of the Prior Art
Active noise control (ANC) can cancel the unwanted noise based on the principle of superposition. Specifically, an anti-noise signal of equal amplitude and opposite phase is generated and combined with the unwanted noise signal, thus resulting in cancellation of both noise signals at a local quite zone (e.g. user's ear drum). For example, the adaptive ANC algorithm models the transfer function of noise traveling from point A (e.g. a reference microphone) to point B (e.g. an error microphone or user's ear drum), and then converts the ambient noise picked up by point A into an anti-noise signal which can cancel the noise at point B. However, the adaptive ANC algorithm may derive incorrect transfer functions when the cancellation target is not from the ambient noise that the ANC aims to cancel but other sound sources, such as the voice spoken by the user himself/herself (i.e. near-end speech), the body collision, or the wind noise. This situation can be called “unstable state”, which continues in a relatively short period compared to the normal ambient noise, and makes the corresponding transfer function unstable. The incorrect transfer function may not be able to cancel the ambient noise, and may even increase the noise in a worst case.
Thus, there is a need for an innovative adaptive ANC system with unstable state handling for keeping an adaptive filter from diverging in the presence of an unstable state caused by near-end speech, body collision, wind noise, etc.
SUMMARY OF THE INVENTION
One of the objectives of the claimed invention is to provide an adaptive active noise control system with unstable state handling and an associated method.
According to a first aspect of the present invention, an exemplary adaptive active noise control (ANC) system is disclosed. The exemplary ANC system includes an ANC circuit and a control circuit. The ANC circuit is arranged to generate an anti-noise signal for noise reduction, wherein the ANC circuit comprises at least one adaptive filter. The control circuit is arranged to receive a first input signal derived from a reference signal output by a reference microphone that picks up ambient noise, receive a second input signal derived from an error signal output by an error microphone that picks up remnant noise resulting from the noise reduction, and perform a transfer function variation detection based on the first input signal and the second input signal to control the at least one adaptive filter.
According to a second aspect of the present invention, an exemplary adaptive active noise control (ANC) method is disclosed. The exemplary ANC method includes: generating, by an ANC circuit, an anti-noise signal for noise reduction, wherein the ANC circuit comprises at least one adaptive filter; receiving a first input signal derived from a reference signal that is generated by picking up ambient noise; receiving a second input signal derived from an error signal that is generated by picking up remnant noise resulting from the noise reduction; and performing a transfer function variation detection based on the first input signal and the second input signal to control the at least one adaptive filter.
These and other objectives of the present invention will no doubt become obvious to those of ordinary skill in the art after reading the following detailed description of the preferred embodiment that is illustrated in the various figures and drawings.
BRIEF DESCRIPTION OF THE DRAWINGS
FIG. 1 is a schematic diagram illustrating an adaptive active noise control (ANC) system according to an embodiment of the present invention.
FIG. 2 is a diagram illustrating a first adaptive ANC system with unstable state handling according to an embodiment of the present invention.
FIG. 3 is a diagram illustrating a detection circuit according to an embodiment of the present invention.
FIG. 4 is a diagram illustrating a transfer function estimation circuit according to an embodiment of the present invention.
FIG. 5 is a diagram illustrating a path change degree evaluation circuit according to an embodiment of the present invention.
FIG. 6 is a diagram illustrating a transfer function restoration circuit according to an embodiment of the present invention.
FIG. 7 is a diagram illustrating a second adaptive ANC system with unstable state handling according to an embodiment of the present invention.
FIG. 8 is a diagram illustrating a third adaptive ANC system with unstable state handling according to an embodiment of the present invention.
DETAILED DESCRIPTION
Certain terms are used throughout the following description and claims, which refer to particular components. As one skilled in the art will appreciate, electronic equipment manufacturers may refer to a component by different names. This document does not intend to distinguish between components that differ in name but not in function. In the following description and in the claims, the terms “include” and “comprise” are used in an open-ended fashion, and thus should be interpreted to mean “include, but not limited to . . . ”. Also, the term “couple” is intended to mean either an indirect or direct electrical connection. Accordingly, if one device is coupled to another device, that connection may be through a direct electrical connection, or through an indirect electrical connection via other devices and connections.
FIG. 1 is a schematic diagram illustrating an adaptive active noise control (ANC) system according to an embodiment of the present invention. The adaptive ANC system 100 may be installed on an earphone such as an earbud. In this embodiment, the adaptive ANC system 100 includes a reference microphone 102, an error microphone 104, an ANC circuit 106, a control circuit 108, and a cancelling loudspeaker 110. The ANC circuit 106 is arranged to generate an anti-noise signal y[n] for noise reduction/cancellation. Specifically, the anti-noise signal y[n] may be a digital signal that is transmitted to the cancelling loudspeaker 110 for playback of analog anti-noise, where the analog anti-noise is intended to reduce/cancel the unwanted ambient noise through superposition. Since an adaptive ANC algorithm is employed by the ANC circuit 106, the ANC circuit 106 includes at least one adaptive filter 112, each being arranged to estimate the unknown transfer function of a primary path from the reference microphone 102 to a position where the noise reduction/cancellation is to be realized. For example, the adaptive filter(s) 112 used by the ANC circuit 106 may be least mean square (LMS) based adaptive filter(s). It should be noted that the number of adaptive filters 112 used by the ANC circuit 106 depends on the adaptive ANC structure employed by the ANC circuit 106. For example, the ANC circuit 106 may employ an adaptive feed-forward ANC structure, an adaptive feedback ANC structure, or an adaptive hybrid ANC structure which is a combination of an adaptive feed-forward ANC structure and an adaptive feedback ANC structure.
The reference microphone 102 is arranged to pick up ambient noise from noise source(s), and generate a reference signal x[n]. The error microphone 104 is arranged to pickup remnant noise resulting from noise reduction/cancellation, and generate an error signal e[n]. One or both of the reference signal x[n] and the error signal e[n] may be used by the ANC circuit 106 for adaptively adjusting filter coefficients of the adaptive filter(s) 112.
In this embodiment, the control circuit 108 is arranged to receive a first input signal derived from the reference signal x[n], receive a second input signal derived from the error signal e[n], and perform a transfer function variation detection based on the first input signal and the second input signal to control the adaptive filter(s) 112.
For better comprehension of technical features of the present invention, the following assumes that the control circuit 108 is used for unstable state handling. When the control circuit 108 is used for unstable state handling, the transfer function variation detection performed by the control circuit 108 is for unstable state detection. However, this is for illustrative purposes only, and is not meant to be a limitation of the present invention. That is, the use of control circuit 108 is not limited to unstable state handling. In practice, any adaptive ANC system using the proposed control circuit 108 for controlling the behavior of adaptive filter(s) falls within the scope of the present invention.
Furthermore, the first input signal used by the control circuit 108 may be directly set by the reference signal x[n], and the second input signal used by the control circuit 108 may be indirectly obtained after the error signal e[n] undergoes certain processing. However, this is for illustrative purposes only, and is not meant to be a limitation of the present invention. In practice, any adaptive ANC system using the proposed transfer function variation detection scheme that is based on signals derived from the reference signal x[n] (which is output by one sensor such as the reference microphone 102) and the error signal e[n] (which is output by another sensor such as the error microphone 104) falls within the scope of the present invention.
FIG. 2 is a diagram illustrating a first adaptive ANC system with unstable state handling according to an embodiment of the present invention. The adaptive ANC system 200 includes an ANC circuit 206 and a control circuit 208. The ANC circuit 106 shown in FIG. 1 may be realized by the ANC circuit 206. The control circuit 108 shown in FIG. 1 may be realized by the control circuit 208. The transfer function of an acoustic channel, also called the primary path, between the reference signal x[n] (which is the ambient noise picked up by the reference microphone 102) and a noise signal d[n] at a position where noise reduction/cancellation occurs is represented by P(z). The transfer function of an electro-acoustic channel, also called the secondary path, between the anti-noise signal y[n] (which is an output of the ANC circuit 206) and the error signal e[n] (which is the remnant noise picked by the error microphone 104) is represented by S(z). Hence, regarding the acoustic superposition in the space from the ANC circuit 206 to the error microphone 104, there is a signal y′[n] resulting from passing the anti-noise signal y[n] through the secondary path transfer function S(z). In this embodiment, the ANC circuit 206 employs an adaptive feed-forward ANC structure having a filtered-x LMS (Fx-LMS) based adaptive filter 212. The Fx-LMS based adaptive filter 212 has a transfer function W(z) defined by filter coefficients that are adaptively adjusted through the Fx-LMS algorithm. Hence, the ANC circuit 206 further includes a filter 214 having a transfer function Ŝ(z) which is an estimation of the second path transfer function S(z). The present invention focuses on the control scheme of the adaptive filter 212. Since the adaptive feed-forward ANC using the Fx-LMS algorithm is known to those skilled in the pertinent art, further description is omitted here for brevity.
Regarding the control circuit 208, it includes a filter 222, a combining circuit 224, an unstable state detection circuit (labeled by “unstable state detection”) 226 and a transfer function restoration circuit (labeled by “transfer function restorer”) 228. In this embodiment, the filter 222 has a transfer function Ŝ(z) which is an estimation of the second path transfer function S(z), and the combining circuit 224 is arranged to subtract an output of the filter 222 from the error signal e[n] to generate an estimated signal {circumflex over (d)}[n] that is an estimation of d[n] (d[n]=P(z)*x[n], where P(z) is unknown). The unstable state detection circuit 226 is arranged to perform unstable state detection according to a first input signal S1 derived from the reference signal x[n] and a second input signal S2 derived from the error signal e[n], and generate a flag signal FL that indicates if an unstable state occurs due to near-end speech, body collision, wind noise, etc. In this embodiment, the first input signal S1 is set by the reference signal x[n], and the second input signal S2 is set by the estimated signal {circumflex over (d)}[n] output from the combining circuit 224.
FIG. 3 is a diagram illustrating a detection circuit according to an embodiment of the present invention. Due to inherent characteristics of the primary path, the transfer function P(z) between x[n] and d[n] is relatively stable under a normal condition with no undesired sound source. However, when there is undesired near-end speech, body collision, wind noise, etc., an incorrect transfer function learned by an adaptive algorithm may deviate from the relatively stable transfer function P(z) and may have large variation. Based on such observations, the unstable state detection circuit 226 may be implemented using the detection circuit 300 shown in FIG. 3. In this embodiment, the detection circuit 300 includes a transfer function estimation circuit 302, a path change degree evaluation circuit 304, and a comparison circuit 306. The transfer function estimation circuit 302 is arranged to estimate a previous transfer function Fn-1(z) of a path between the first input signal S1 (e.g. x[n]) and the second input signal S2 (e.g. {circumflex over (d)}[n]) and a current transfer function Fn(z) of the path between the first input signal S1 (e.g. x[n]) and the second input signal S2 (e.g. {circumflex over (d)}[n]). For example, the transfer function estimation circuit 302 may be implemented using the transfer function estimation circuit 400 shown in FIG. 4, where the transfer function estimation circuit 400 includes an LMS-based adaptive filter 402, a set of filter coefficients w′[n−1] previously used by the LMS-based adaptive filter 402 is output as a representation of the previous transfer function Fn-1 (z), and a set of filter coefficients w′[n] currently used by the LMS-based adaptive filter 402 is output as a representation of the current transfer function Fn(z).
Regarding the path change degree evaluation circuit 304, it is arranged to determine a path change degree value DV according to difference between the previous transfer function Fn-1(z) and the current transfer function Fn(z). Ina case where the previous transfer function Fn-1(z) is represented by filter coefficients w′[n−1] of the LMS adaptive filter 402 and the current transfer function Fn(z) is represented by filter coefficients w′ [n] of the LMS adaptive filter 402, the path change degree evaluation circuit 304 may be implemented using the path change degree evaluation circuit 500 shown in FIG. 5. The path change degree evaluation circuit 500 includes a frequency-domain processing circuit 502 and a smoothing filter 504. The frequency-domain processing circuit 502 is arranged to apply frequency-domain processing to the previous transfer function Fn-1(z) (e.g. w′[n−1]) and the current transfer function Fn(z) (e.g. w′[n]) to generate a first processing result D1 and a second processing result D2, respectively. For example, the frequency-domain processing may include a band-pass filter (BPF) or a fast Fourier transform (FFT). In addition, the frequency-domain processing circuit 502 is further arranged to output difference DF between the first processing result D1 and the second processing result D2 (i.e. DF=|D2−D1|). The smoothing filter 504 is arranged to process the difference DF between the first processing result D1 and the second processing result D2 to generate and output the path change degree value DV. Specifically, the path change degree value DV is an indicator of the degree of transfer function variation.
After the path change degree value DV is available, the comparison circuit 306 is arranged to compare the path change degree value DV with a pre-defined threshold TH to generate a comparison result, set a flag signal FL according the comparison result, and output the flag signal FL to at least the Fx-LMS based adaptive filter 212. When the path change degree value DV reaches the pre-defined threshold TH (e.g. DV>TH), the comparison circuit 306 judges that an unstable state occurs at this moment, and sets the flag signal FL by a first logic level (e.g. FL=1). When the path change degree value DV does not reach the pre-defined threshold TH (e.g. DV≤TH), the comparison circuit 306 judges that there is no unstable state at this moment, and sets the flag signal FL by a second logic level (e.g. FL=0).
As shown in FIG. 2, the Fx-LMS based adaptive filter 212 is controlled by the flag signal FL. When the flag signal FL has the first logic level (e.g. FL=1) for indicating the presence of the unstable state, the Fx-LMS based adaptive filter 212 may be instructed to freeze coefficient adaptation. That is, the transfer function W(z) estimated by the Fx-LMS based adaptive filter 212 keeps unchanged when the flag signal FL is asserted by the unstable state detection circuit 226. When the flag signal FL has the second logic level (e.g. FL=0) for indicating the absence of the unstable state, the Fx-LMS based adaptive filter 212 is instructed to resume coefficient adaptation. That is, the transfer function W(z) estimated by the Fx-LMS based adaptive filter 212 is allowed to be updated by the Fx-LMS algorithm when the flag signal FL is deasserted by the unstable state detection circuit 226. Since the coefficient adaptation is frozen during a period in which the unstable state detection circuit 226 detects occurrence of the unstable state, the Fx-LMS based adaptive filter 212 is protected from diverging in the presence of the unstable state.
Generally, the unstable state detection requires certain processing time, such that the flag signal FL is asserted later than the start time of the unstable state. At the time an unstable state is detected by the unstable state detection circuit 226, a set of filter coefficients w[n] currently used by the Fx-LMS based adaptive filter 212 may be already affected by the unstable state and may represent an incorrect transfer function. To address this issue, the present invention proposes using the transfer function restoration circuit 228 to buffer one or more sets of filter coefficients w[n−i] previously used by the Fx-LMS based adaptive filter 212. The transfer function restoration circuit 228 is also controlled by the flag signal FL set by the unstable state detection circuit 226, and can be used to correct the transfer function (i.e. filter coefficients) misled by the undesired sound source.
FIG. 6 is a diagram illustrating a transfer function restoration circuit according to an embodiment of the present invention. The transfer function restoration circuit 228 shown in FIG. 2 may be realized by the transfer function restoration circuit 600 shown in FIG. 6. The transfer function restoration circuit 600 has a transfer function pool 602 that can be implemented using a storage device such as a memory, and is arranged to periodically buffer a set of filter coefficients w[n] currently used by the Fx-LMS based adaptive filter 212. When the flag signal FL has the first logic level (e.g. FL=1) for indicating the presence of the unstable state, the transfer function restoration circuit 600 (particularly, transfer function pool 602 of transfer function restoration circuit 600) may be instructed to output a set of filter coefficients w[n−i] previously used by the Fx-LMS based adaptive filter 212 to the Fx-LMS based adaptive filter 212 for updating a set of filter coefficients w[n] currently used by the Fx-LMS based adaptive filter 212. Since the set of filter coefficients w[n−i] previously used by the Fx-LMS based adaptive filter 212 may be determined by the Fx-LMS algorithm in the absence of the unstable state, the transfer function restoration applied to the Fx-LMS based adaptive filter 212 can enhance the ANC performance greatly during a period in which an unstable state is detected by the unstable state detection circuit 226.
FIG. 7 is a diagram illustrating a second adaptive ANC system with unstable state handling according to an embodiment of the present invention. The adaptive ANC system 700 includes an ANC circuit 706 and a control circuit 708. The ANC circuit 106 shown in FIG. 1 may be realized by the ANC circuit 706. The control circuit 108 shown in FIG. 1 may be realized by the control circuit 708. In this embodiment, the ANC circuit 706 employs an adaptive feedback ANC structure having one Fx-LMS based adaptive filter 712. The Fx-LMS based adaptive filter 712 has a transfer function W(z) defined by filter coefficients that are adaptively adjusted through the Fx-LMS algorithm. Hence, the ANC circuit 706 further includes a filter 714 having a transfer function S(z) which is an estimation of the second path transfer function S (z). In this feedback structure, the ANC circuit 706 further includes a filter 716 and a combining circuit 718 jointly used for generating an estimated signal {circumflex over (d)}[n] from the measured error signal e[n], wherein the estimated signal {circumflex over (d)}[n] represents an estimation of d[n] (d[n]=P(z)*x[n], where P(z) is unknown). It should be noted that the reference signal x[n] (which is the ambient noise picked by the reference microphone 102) is used by the control circuit 708, but is not used by the ANC circuit 706 with the adaptive feedback ANC structure. The present invention focuses on the control scheme of the adaptive filter 712. Since the adaptive feedback ANC using the Fx-LMS algorithm is known to those skilled in the pertinent art, further description is omitted here for brevity.
Regarding the control circuit 708, it includes an unstable state detection circuit (labeled by “unstable state detection”) 726 and a transfer function restoration circuit (labeled by “transfer function restorer”) 728. The unstable state detection circuit 726 is arranged to perform unstable state detection according to a first input signal S1 derived from the reference signal x[n] and a second input signal S2 derived from the error signal e[n], and generate a flag signal FL that indicates if an unstable state occurs due to near-end speech, body collision, wind noise, etc. In this embodiment, the first input signal S1 is set by the reference signal x[n], and the second input signal S2 is set by the estimated signal {circumflex over (d)}[n] output from the combining circuit 518 included in the adaptive feedback ANC structure. However, this is for illustrative purposes only, and is not meant to be a limitation of the present invention.
In this embodiment, the unstable state detection circuit 726 may be realized by the prediction circuit 300 shown in FIG. 3 for unstable state detection, and the transfer function restoration circuit 728 may be implemented by the transfer function restoration circuit 600 shown in FIG. 6 for transfer function restoration of the Fx-LMS adaptive filter 712. Since a person skilled in the art can readily understand the principles of unstable state detection circuit 726 and transfer function restoration circuit 728 after reading above paragraphs directed to FIGS. 3-6, further description is omitted here for brevity.
FIG. 8 is a diagram illustrating a third adaptive ANC system with unstable state handling according to an embodiment of the present invention. The adaptive ANC system 800 includes an ANC circuit 806 and a control circuit 808. The ANC circuit 106 shown in FIG. 1 may be realized by the ANC circuit 806. The control circuit 108 shown in FIG. 1 may be realized by the control circuit 808. In this embodiment, the ANC circuit 806 employs an adaptive hybrid ANC structure which is a combination of an adaptive feed-forward ANC structure shown in FIG. 2 and an adaptive feedback ANC structure shown in FIG. 7, and has one Fx-LMS based adaptive filter 812_1 for the adaptive feed-forward ANC structure and another Fx-LMS based adaptive filter 812_2 for the adaptive feedback ANC structure. The Fx-LMS based adaptive filter 812_1 has a transfer function WFF(z) defined by filter coefficients that are adaptively adjusted through the Fx-LMS algorithm. Hence, regarding the adaptive feed-forward ANC structure (which is a part of the adaptive hybrid ANC structure), the ANC circuit 806 includes a filter 814_1 having a transfer function Ŝ(z) which is an estimation of the second path transfer function S(z). In addition, the Fx-LMS based adaptive filter 812_2 has a transfer function WFB(z) defined by filter coefficients that are adaptively adjusted through the Fx-LMS algorithm. Hence, regarding the adaptive feedback ANC structure (which is another part of the adaptive hybrid ANC structure), the ANC circuit 806 includes a filter 814_2 having a transfer function Ŝ(z) which is an estimation of the second path transfer function S(z), and further includes a filter 816 and a combining circuit 818 jointly used for generating an estimated signal {circumflex over (d)}[n] from the measured error signal e[n], where the estimated signal {circumflex over (d)}[n] is an estimation of d[n] (d[n]=P(z)*x[n], where P(z) is unknown). The present invention focuses on the control scheme of the adaptive filters 812_1 and 812_2. Since the adaptive hybrid ANC using the Fx-LMS algorithm is known to those skilled in the pertinent art, further description is omitted here for brevity.
Regarding the control circuit 808, it includes an unstable state detection circuit (labeled by “unstable state detection”) 826 and a transfer function restoration circuit (labeled by “transfer function restorer”) 828. The unstable state detection circuit 826 is arranged to perform unstable state detection according to a first input signal S1 derived from the reference signal x[n] and a second input signal S2 derived from the error signal e[n], and generate a flag signal FL that indicates if an unstable state occurs. In this embodiment, the first input signal S1 is set by the reference signal x[n], and the second input signal S2 is set by the reference signal {circumflex over (d)}[n] output from the combining circuit 818 included in the adaptive hybrid ANC structure. However, this is for illustrative purposes only, and is not meant to be a limitation of the present invention.
In this embodiment, the unstable state detection circuit 826 may be realized by the prediction circuit 300 shown in FIG. 3 for unstable state detection, and the transfer function restoration circuit 828 may be implemented by the transfer function restoration circuit 600 shown in FIG. 6 for transfer function restoration of each of the Fx-LMS adaptive filters 812_1 and 812_2. For example, the transfer function restoration circuit 828 is arranged to periodically buffer a set of filter coefficients currently used by the Fx-LMS based adaptive filter 812_1 and a set of filter coefficients currently used by the Fx-LMS based adaptive filter 812_2. In addition, when the flag signal FL is asserted by the unstable state detection circuit 826 in response to the detected unstable state, the transfer function restoration circuit 828 is further arranged to output a set of filter coefficients previously used by the Fx-LMS based adaptive filter 812_1 to update a set of filter coefficients currently used by the Fx-LMS based adaptive filter 812_1, and output a set of filter coefficients previously used by the Fx-LMS based adaptive filter 812_2 to update a set of filter coefficients currently used by the Fx-LMS based adaptive filter 812_2. Since a person skilled in the art can readily understand the principles of unstable state detection circuit 826 and transfer function restoration circuit 828 after reading above paragraphs directed to FIGS. 3-6, further description is omitted here for brevity.
Those skilled in the art will readily observe that numerous modifications and alterations of the device and method may be made while retaining the teachings of the invention. Accordingly, the above disclosure should be construed as limited only by the metes and bounds of the appended claims.
Patent Application
20230298559
