CA2519868C

CA2519868C - Method and system for active noise cancellation

Info

Publication number: CA2519868C
Application number: CA2519868A
Authority: CA
Inventors: Hamid Sheikhzadeh-Nadjar; Todd Schneider; Robert L. Brennan
Original assignee: On Semiconductor Trading Ltd
Current assignee: On Semiconductor Trading Ltd
Priority date: 2004-09-15
Filing date: 2005-09-15
Publication date: 2012-11-06
Anticipated expiration: 2025-09-15
Also published as: CA2519868A1

Abstract

A method and system for active noise cancellation is provided. The system employs subband processing, and preferably implements over-sampled filterbank. The system is applicable to adaptive noise cancellation, adaptive echo cancellation for portable listening devices, such as headsets and other similar listening devices.

Description

Method and System for Active Noise Cancellation FIELD OF INVENTION

[0001 ] The present invention relates to signal processing and modeling technique, and more specifically to signal processing and modeling technique for noise cancellation.
BACKGROUND OF THE INVENTION

[0002] Analog active noise cancellation (ANC) systems suffer from a number of problems. Specifically, they are prone to acoustic feedback, and they do not provide as high a degree of cancellation for periodic or other quasi-stationary signals as can be realized with a digital signal processing (DSP) enhanced analog ANC system.

[0003] Analog ANC systems are also difficult to adjust (or "tune") for different headset designs and also in a production environment where normal production variations in transducers and listening device assembly increase the likelihood of acoustic feedback.

[0004] Further, current analog ANC techniques address only part of the noise cancellation that is needed by users in high noise environments. Specifically, analog ANC provides noise cancellation at predominantly low frequencies (below 1500 Hz to 2000 Hz).

[0005] Fully digital ANC systems are possible. However, group delay or latency is induced by analog to digital (A/D) conversion, digital to analog conversion and digital processing associated with DSP systems. Further, due to power consumption, they are not practical in many portable applications.

[0006] U.S. Patent Nos. 5,475,761, 5,699,436, and 5,815,582 by Noise Cancellation Technologies (NCT) disclose methods of digital ANC using a combination of both feedback and feed-forward methods. The methods employ DSP to perform ANC.
However, due to the inherent delay in the DSP, they are not practical for most applications when low-power, low-cost, and small-size constraints are applied.
There are many other similar DSP-based systems that suffer from the same delay problem, for example, the systems disclosed in U.S. Patent Nos. 6,418,227 B1, 5,991,418 and 5,940,519 by Kuo et al. from Texas Instruments Inc.

[0007] U.S. Patent Nos. 6,069,959 and 6,118,878 by NCT disclose fully analog solutions to the ANC problems. Specifically, as U.S. Patent No. 6,118,878 explains, significant tuning and adaptation of the system parameters are necessary to avoid instability and artifacts. However, the patent suggests that the tuning can be implemented using analog components and methods.

[0008] DSP-controlled ANC systems have tried to address the difficult problem of tuning of the analog ANC systems through the use of CPUs and signal processing methods. For example, U.S. Patent 5,440,642 by Denenberg et al. discloses DSP
techniques that can control ANC system parameters, such as loop gain and loop filter to frequency response. U.S. Patent Application Publication No. 20040037430 Al uses DSP techniques (LMS adaptation) to control the secondary path typically used in the filtered-X LMS algorithm. U.S. Patent No. 4,965,832 uses DSP control of a feed-forward ANC system to control the loop-gain and the loop-filter bandwidth. U.S.
Patent No. 6,278,786 B 1 uses DSP to not only control the loop-gain but also to provide an acoustic signal (to be added to the analog ANC anti-noise) that will cope with tonal noises more effectively.

[0009] Subband adaptive filters (SAFs) become an interesting and viable option for many adaptive systems. The SAF approach uses a filterbank to split the fullband signal input into a number of frequency bands, each serving as input to an adaptive filter. This subband decomposition greatly reduces the update rate and the length of the adaptive filters resulting in much lower computational complexity.

[0010] To be able to employ powerful SAF method for ANC, one has to tackle a processing delay issue.

[0011] To reduce the processing delay in the SAFs, U.S. Patent No. 5,329,587 by Morgan et al. has introduced a method of reconstructing the subband filter back into time-domain. Starting with adapted subband filters, they first transform the SAFs into the frequency-domain (using an FFT), appropriately stack the results, and inverse transform them back into time-domain to obtain a time-domain adaptive filter.
The time-domain filter is then used to implement time-domain adaptive filtering.
The details of their technique are also reported in a research paper Morgan et.
al. ("A

delayless subband adaptive filter structure", IEEE Trans. on Signal Proc., Vol. 43, pp.
1819-1830, Aug. 1995.) that offers a good survey of previous efforts on low-delay adaptive systems. Let us call this method DFT-1 Stacking as disclosed in J.
Huo et al.
("New weight transform schemes for delayless subband adaptive filtering", in Proc. of IEEE Global Telecom. Conf., pp. 197-201, 2001). After analyzing Morgan's method in by J. Huo et al., they offer two variations to the method (known as "DFT-2 Stacking"
and "DFT-FIR Stacking") to improve the performance. These methods are all based on DFT, proper stacking, and inverse DFT. In DFT-FIR, a convolution with a synthesis filter after DFT is also added. Moreover in L. Larson et al. ("A new subband weight transform for delayless subband adaptive filtering structures", in Proc. of IEEE DSP
workshop, pp. 201-206), a Linear Weight Transform method is introduced. The method employs a linear matrix transformation of the subband filters using both analysis and synthesis filters to recover the time-domain adaptive filter. In yet another set of works following Morgan's method, a different method is proposed that employs the Hadamard transform to reconstruct the time-domain filter (N. Hirayama and H.
Sakai, "Analysis of a delayless subband adaptive filter", in Proc. of ICASSP, pp.
2329-2332, 1997; and N. Hirayama et al., "Delayless subband adaptive filtering using the hadamard transform", IEEE Trans. on Signal Proc., Vol. 47, No. 6, pp. 1731-1734, Jun. 1999).

[0012] In a series of research paper presented from 1997 to 1999, Merched et al. present methods of transferring the SAFs to time-domain (R. Merched et al. "A
Delayless alias-free subband adaptive filter structure", in Proc. of IEEE Int. Symp. On Circuits and Systems, pp. 2329-2332, Jun. 9-12, 1997; P. S. R. Diniz et al. "Analysis of a delayless subband adaptive filter structure", in Proc. of ICASSP, pp. 1661-1664, 1998;
R. Merched et al. "A new delayless subband adaptive filter structure", IEEE
Trans. on Signal Proc., Vol. 47, No. 6, pp. 1580-1591, Jun. 1999). Their methods are designed only for maximally decimated (QMF) PR filterbanks, and constraints the filterbank prototype filter to be a Nyquist(K) filter (where K represents number of subbands). As a result, the SAFs become simple fractional delay filters. They also use a polyphase fiterbank to reconstruct the time-domain adaptive filter.

[0013] U. S. Patent No. 6,661,895 B1 by Jin et al. discloses a zero-delay SAF
system.
They discard the initial segment of each SAF to obtain a "forward filter". The estimated (time-domain) echo signal is generated by filtering the reference signal through the subband forward filters and then applying subband reconstruction. The time-domain echo cancelled signal goes through another separate time-domain LMS filter to compensate for the discarded initial segments of subband adaptive filters. The method however has a fundamental problem: the time-domain LMS filter has to model a potentially non-causal filter. This is not practically possible.

[0014] Over-sampled subband adaptive filters (OS-SAF) offer many advantages over time-domain adaptive algorithms. However, OS-SAF systems may introduce a group delay or latency that is too high for some applications. The conversion from analog to digital and back again, use of anti-aliasing and anti-imaging filters, as well as the digital processing introduces this delay. Reducing the delay of OS-SAF systems by increasing the sampling rate is not practical in many applications because of power consumed and specialized hardware required.

[0015] Also, in the conventional OS-SAF systems, the primary input signal goes through analysis and synthesis stages of the over-sampled filterbank. Often, perfect reconstruction (PR) is not practical and only a near PR performance is achieved. As a result, the primary signal may be distorted. To minimize distortions, longer analysis windows have to be employed which further increases the delay and add extra computation cost.

[0016] Further, in the conventional OS-SAF systems, the effect of analysis filter band edges and under-modeling errors limit the system performance.

SUMMARY OF THE INVENTION

[0017] It is an object of the invention to provide a novel method and system that obviates or mitigates at least one of the disadvantages of existing systems.

[0018] It is an object of the invention to provide an improved method and system for active noise cancellation.

[0019] The invention relates to the improvements that can be made to well-known analog ANC techniques as well as over-sampled subband adaptive filtering using specialized DSP methods and apparatuses.

[0020] According to an aspect of the present invention, there is provided a system for active noise cancellation, includes: a first over-sampled analysis filterbank for transferring a reference signal in a time-domain, which is associated with noise, into a plurality of subband reference signals in a frequency-domain; a second over-sampled analysis filterbank for transferring a primary signal in the time-domain, which is associated with an acoustic signal and may be contaminated by the noise, into a plurality of subband primary signals in the frequency-domain; a subband processing module for processing the subband reference signals, the subband primary signals or a combination thereof, and implementing one or more than one subband adaptive algorithm in the frequency-domain; and an over-sampled synthesis filterbank for transferring the outputs of the subband processing module into a time-domain output signal.

[0021 ] According to a further aspect of the present invention, there is provided a system for active noise cancellation, includes: an over-sampled analysis filterbank for transferring a primary signal in a time-domain, which is associated with an acoustic signal and maybe contaminated by noise, into a plurality of subband primary signals in a frequency-domain; a subband processing module for the subband primary signals and implementing one or more than one subband adaptive algorithm in frequency-domain;
and an over-sampled synthesis filterbank for transferring the outputs of the subband processing module into a time-domain output signal.

[0022] According to a further aspect of the present invention, there is provided a system for active noise cancellation, includes: a first analysis filter bank for transferring a reference signal in a time-domain into a plurality of subband reference signals in a frequency-domain; a second analysis filter bank for transferring a primary signal in the time-domain, which is associated with an acoustic signal and may be contaminated by noise, into a plurality of subband primary signals in the frequency-domain; a subband estimator for modeling subband acoustic transfer function for the subband reference signals; a subband adaptive filter for providing a plurality of subband output signals in response to the subband reference signals; an adjustor for adjusting the subband adaptive filter in response to the subband primary signals and the modeling for the subband reference signals; and a synthesis filter bank for transferring the subband output signals to a time-domain output signal.

[0023] According to a further aspect of the present invention, there is provided a system for active noise cancellation, includes: an analysis filter bank for transferring a primary signal in a time-domain into a plurality of subband primary signals in a frequency-domain; a subband filter bank for providing a plurality of subband output signals in response to a plurality of subband reference signals; a synthesis filter bank for transferring the subband output signals into a time-domain output signal; and a feed-back loop for generating the subband reference signals, including: a first subband estimator for modeling subband acoustic transfer function for the subband output signals; a signal path for providing the subband reference signals in response to the subband primary signals and the modeling for the subband output signals; a second subband estimator for modeling subband acoustic transfer function for the subband reference signals; and an adjustor for adjusting the subband adaptive filter in response to the subband primary signals and the modeling for the subband reference signals.
[0024] According to a further aspect of the present invention, there is provided a system for active noise cancellation, includes: a first analysis filter bank for transferring a primary signal in a time-domain into a plurality of subband primary signals in a frequency-domain; a time-domain filter bank for providing a time-domain output signal in response to a reference signal in the time-domain; a feed-back loop for generating the reference signal, including: a first subband estimator for modeling subband acoustic transfer function for the time-domain output signal, a signal path for providing the reference signal in the time-domain in response to a primary signal in the time-domain and the modeling for the time-domain output signal, a second subband estimator for modeling subband acoustic transfer function for the subband reference signal, a second analysis filter bank for transferring the primary signal into a plurality of subband primary signals in the frequency-domain, an adjustor for adjusting the subband adaptive filter in response to the subband primary signals and the modeling for the reference signal, and a synthesis filter bank for converting the subband adaptive filter to the time-domain filter bank for filtering the reference signal.

[0025] According to a further aspect of the present invention, there is provided a system for active-noise cancellation, includes: an analog active noise cancellation (ANC) system for performing an active noise cancellation to a primary signal in a time-domain, which is associated with an acoustic signal and may be contaminated by noise;
a first over-sampled analysis filterbank for transferring a reference signal in the time-domain into a plurality of subband reference signals in a frequency-domain, the reference signal in the time-domain being associated with the noise; a second over-sampled analysis filterbank for transferring the primary signal in the time-domain into a plurality of subband primary signals in the frequency-domain; a subband processing module for processing the subband reference signals, the subband primary signals or a combination thereof, and for adjusting one or more parameters of the analog ANC system; an over-sampled synthesis filterbank for performing conversion on the outputs of the subband processing module from the frequency-domain to the time-domain.

[0026] According to a further aspect of the present invention, there is provided a system for active noise cancellation, includes: an analog active noise cancellation (ANC) system for performing an active noise cancellation to a primary signal in a time-domain, which is associated with an acoustic signal and may be contaminated by noise;
an over-sampled analysis filterbank for transferring the primary signal in the time-domain into a plurality of subband primary signals in a frequency-domain; a subband processing module for processing the subband primary signals and for adjusting one or more than one parameter of the analog ANC system; an over-sampled synthesis filterbank for transferring the outputs of the subband processing module into an output signal in the time-domain.

[0027] According to a further aspect of the present invention, there is provided a system for active noise cancellation, comprising: a first WOLA analysis filterbank for transferring a reference signal in a time-domain, which is associated with noise, into a plurality of subband reference signals in a frequency-domain; a second WOLA
analysis filterbank for transferring a primary signal in the time-domain, which is associated with an acoustic signal and may be contaminated by the noise, into a plurality of subband primary signals; a subband adaptive processing module for processing the output of the first WOLA analysis filterbank, the output of the second WOLA analysis filterbank or a combination thereof, and providing a plurality of subband adaptive filters;
and a WOLA synthesis filterbank for synthesizing the subband adaptive filters to provide a time-domain filter for filtering the reference signal in the time-domain.

[0028] According to a further aspect of the present invention, there is provided a method for active noise cancellation implemented by the systems described above.
[0029] This summary of the invention does not necessarily describe all features of the invention.

BRIEF DESCRIPTION OF THE DRAWINGS

[0030] These and other features of the invention will become more apparent from the following description in which reference is made to the appended drawings wherein:
[0031 ] FIGURE 1(a) is a diagram showing a conventional analog ANC system;
[0032] FIGURE 1(b) is a diagram showing a detailed block diagram of Figure 1(a), depicting a transfer function and an acoustic transfer function;

[0033] FIGURE 2 is a diagram showing a conventional DSP-based ANC system;
[0034] FIGURE 3 is a diagram showing a subband ANC system using a subband FX-LMS;

[0035] FIGURE 4 is a diagram showing a subband ANC system using a subband FX-LMS in accordance with an embodiment of the present invention;

[0036] FIGURE 5 is a diagram showing a conventional feedback ANC system using a subband FX-LMS;

[0037] FIGURE 6 is a diagram showing a subband feedback ANC system using a subband FX-LMS in accordance with a further embodiment of the present invention;
[0038] FIGURE 7 is a diagram showing a delayless subband ANC system using a FX-LMS in accordance with a further embodiment of the present invention;

[0039] FIGURE 8 is a diagram showing a delayless subband feedback ANC system using a FX-LMS in accordance with a further embodiment of the present invention;
[0040] FIGURE 9 is a diagram showing an ANC system using over-sampled subband processing in accordance with a further embodiment of the present invention;

[0041] FIGURE 10 is a diagram showing a feedback ANC system using over-sampled subband processing in accordance with a further embodiment of the present invention;
[0042] FIGURE 11 is a diagram showing an ANC system using Weighted Overlap-Add (WOLA) in accordance with a further embodiment of the present invention;

[0043] FIGURE 12 is a diagram showing a feedback ANC system using WOLA in accordance with a further embodiment of the present invention;

[0044] FIGURE 13 is a diagram showing an ANC system using WOLA in accordance with a further embodiment of the present invention;

[0045] FIGURE 14 is a diagram showing a conventional OS-SAF system;

[0046] FIGURES 15(a)-(b) are diagrams showing examples of an adaptive processing block (APB);

[0047] FIGURE 16 is a diagram showing an OS-SAF having WOLA;

[0048] FIGURE 17 is a diagram showing an OS-SAF system using over-sampled subband processing in accordance with a further embodiment of the present invention;
[0049] FIGURE 18 is a diagram showing an OS-SAF system using over-sampled subband processing in accordance with a further embodiment of the present invention;
[0050] FIGURE 19 is a diagram showing an OS-SAF system using WOLA in accordance with a further embodiment of the present invention;

[0051] FIGURE 20 is a diagram showing a closed-loop OS-SAF system using over-sampled subband processing in accordance with a further embodiment of the present invention;

[0052] FIGURE 21 is a diagram showing a closed-loop OS-SAF system using WOLA
in accordance with a further embodiment of the present invention;

[0053] FIGURE 22 is a diagram showing a delayless subband feedback ANC system using WOLA in accordance with a further embodiment of the present invention;

[0054] FIGURE 23 is a diagram showing a delayless SAF system using oversampled filterbanks, employing weight transform for time-filter reconstruction;

[0055] FIGURE 24 is a diagram showing a Reconstruction of TAF through WOLA
synthesis of the SAFs in accordance with a further embodiment of the present invention;

[0056] FIGURE 25 is a diagram showing details of time-filter reconstruction using the WOLA process;

[0057] FIGURE 26 is a diagram showing an oversampled SAF system applied to echo cancellation;

[0058] FIGURE 27 is a graph for showing the simulation results of Figures 24-26;
[0059] FIGURE 28 is a graph showing the ERLE results;

[0060] FIGURE 29 is a graph showing an example of real and imaginary parts of SAFs;
[0061] FIGURE 30 is a graph showing an example of periodic extension of a WOLA
synthesis;

[0062] FIGURE 31 is a graph showing an example of a synthesis window;

[0063] FIGURE 32 is a graph showing the result of widow application of Figure 31;
[0064] FIGURE 33 is a graph showing an example of time sample for synthesizing a time-domain filter;

[0065] FIGURE 34(a) is a graph showing an example of time sample for a synthesized time-filter super-imposed on the ITUT plant;

[0066] FIGURE 34(b) is a graph showing an example of time-domain difference between the synthesized time-filter and the ITUT plant; and [0067] FIGURE 35 is a graph showing echo attenuation of a time-domain filter and a conventional SAF.

DETAILED DESCRIPTION

[0068] The embodiments of the present invention are described for a headset.
However, the embodiments of the present invention are applicable to any other listening devices, such as portable listening devices. The embodiments of the present invention are described mostly for active noise cancellation and echo cancellation. However, the embodiments of the present invention can be employed for other applications, including adaptive noise cancellation.

[0069] The embodiments of the present invention relate to over-sampled subband adaptive filtering using specialized DSP techniques and analog ANC techniques.
Thus, DSPs are relevant to the technology disclosed below. Because the applications of these techniques are in listening devices and the cancellation relies on acoustic summation, the embodiments of the present invention relates to acoustics.

[0070] Figure 1(a) illustrates a conventional analog ANC system 2 for a headset. As shown in Figure 1(a), a primary noise signal x(t) is sensed by a microphone 6.
The microphone 6 is usually located within the earcup of the headset. For example, the primary noise signal x(t) is a signal outside the earcup of the headset. An analog ANC
circuitry 4 receives the microphone signal e(t), and generates an electric signal z(t).
The electric signal z(t) is added at 8 with a local audio signal s(t) (possibly speech) to generate an electric speaker signal y(t). The electric speaker signal y(t) is played through a loudspeaker 10 for the listener. The loudspeaker 10 is located within the earcup of the headset. The ANC system 2 tries to cancel the effect of a noise signal for the listener through estimating, generating the signal z(t) to be played through the loudspeaker 10 together with the local audio signal s(t).

[0071 ] Figure 1(b) illustrates modeling of a transfer function and an acoustic transfer function of Figure 1(a). P(s) 12 models the transfer function for the acoustic noise signal x(t) to be converted to an electric signal d(t). Q(s) 14 models the acoustic transfer function for the loudspeaker signal y(t) to reach the microphone (6).
Usually Q(s) 14 is assumed to be more known than P(s) 12 since the locations of the loudspeaker (10) and the microphone (6) are fixed and known. A through review of the conventional analog ANC system 2 is provided in Kuo-Morgan99 (Sen M. Kuo and Dennis R. Morgan, "Active Noise Control: A Tutorial Review", Proceedings of the IEEE, Vol. 87, June 1999, pp. 943-973).

[0072] ANC systems may also provide one or more microphones to measure the ambient noise (e.g. signal x(t) outside of the earcup), however, single microphone systems are generally preferred as discussed later.

[0073] Figure 2 illustrates a conventional DSP-based ANC system 20. The system of Figure 2 is a feed-forward ANC system, and employs the FX LMS algorithm for active noise cancellation with two microphones. The two microphone signals x(t), e(t) are converted to digital signals x(n), e(n) by analog/digital (A/D) converters 22 and 24, and processed in discrete-time by an algorithm to generate an anti-noise signal z(n).
The anti-noise signal z(n) is converted back to an analog signal z(t) by a digital/analog (D/A) converter 32, and played through the loudspeaker together with the signal s(t).
The method might employ adaptive methods, such as the Normalized Least Mean Square (NLMS) 30 or similar techniques to adapt an adaptive filter W(z) 28. A
rough estimate of the loudspeaker to the error microphone transfer function Q(s) is also required. In Figure 2, this is depicted by the discrete-time estimated transfer function Q(z) 26. Various methods for off-line on-line estimation of Q(s) have been proposed in the prior art and reviewed in Kuo-Morgan99.

[0074] Subband ANC methods have been presented in Kuo-Morgan99 to achieve lower computation cost and faster convergence. Figure 3 illustrates a conventional subband ANC system for two microphones, employing a subband FX-LMS.

[0075] The system 40 of Figure 3 includes three Analysis Filter Bank (AFB) components 42. The AFB components 42 decompose the time-domain signals e(n), x(n), x' (n) into K (possibly complex) subband signals e; (m) x i (m), x;' (m), i = 0,1,..., K -1 that might be also decimated in time. There exist K (possibly complex) subband adaptive filers (SAFs) Wi(z) 44 (W; (z), i = 0,1,..., K -1) which generate subband output signals z; (m), i = 0,1,..., K -1. All of the adaptive processing is done in a subband adaptive processing (SAP) block 90 in Figure 3. A Synthesis Filter Bank (SFB) 46 then combines the subband output signals to obtain the time domain signal z(n). The D/A
converter 32 converts the time domain, digital signal z(n) into a time domain, analog signal z(t).

[0076] Figure 4 illustrates a subband ANC system 50a for two microphones in accordance with an embodiment of the present invention. In Figure 4, adaptive processing is implemented in a SAP block 91. The system 50a of Figure 4 employs suband FX-LMS, and includes a block 54 that includes a subband estimate of Q(s) depicted as Q; (z), i = 0,1,..., K -1. Subband estimation and implementation of Q(s) allows for faster computation due to parallel subband processing by filters Q;
(z). This allows the system 50a to include only two AFBs 52 for two microphones. One AFB
is provided to x(n) , while the other is provided to e(n). K (possibly complex) subband adaptive filers (SAFs) W; (z) 56 (W; (z), i = 0,1,..., K -1) generate subband output signals z; (m), i = 0,1,..., K -1, based on x; (m), i = 0,1,..., K -1. A block (such as NLMS) 58 is provided to adapt the subband adaptive filters Wi(z). A SFB 60 combines the subband output signals z; (m), i = 0,1,..., K -1, to obtain the time domain signal z(n). The D/A converter 32 converts the time domain, digital signal z(n) into a time domain, analog signal z(t).

[0077] It is possible to implement FX-LMS with only one microphone as illustrated in Figure 5. Figure 5 shows a conventional feedback ANC system 70. The system 70 is disclosed in Kuo-Morgan99. In Figure 5, the reference signal is reconstructed in the system (signal r(n)) via a discrete-time estimated transfer function Q(z) 72 and a summation node 74.

[0078] The system 50a of Figure 4 may be implemented using oversampled filterbank as shown in Figure 9. AFBs 52 of Figure 4 may be implemented by over-sampled analysis filterbanks 112, 114 of Figure 9. WOLA implementation offers a low-delay, flexible, and efficient implementation of the over-sampled filterbanks as described in U.S. Patent Nos. 6,236,731, 6,240,192 and 6,115,478.

The system 50a of Figure 4 may be implemented using WOLA filterbank as shown in Figures 11 and 16. AFBs 52 of Figure 4 may be implemented by WOLA

analysis filterbanks 132, 134 of Figures 11 and 16, and SFB 60 of Figure 4 may be implemented by a WOLA synthesis filterbank 138 of Figures 11 and 16.

[0079] Figure 6 illustrates a subband feedback ANC system 50b in accordance with a further embodiment of the present invention. A subband implementation of feedback FX-LMS system is shown in Figure 6. In Figure 6, adaptive processing is implemented in a SAP block 92. The reference signal is reconstructed in the system 50b via Q; (z), i = 0,1,..., K -1 (referenced by.80) and a summation node 82. The block 80 includes a subband estimate of Q(s) depicted as Q; (z), i = 0,1,..., K -1.

[0080] As discussed in the prior art (Kuo-Morgan99), the use of AFBs in subband implementations may impose delays on the signal that are often prohibitively large for the operation of the system.

[0081 ] The system 50b of Figure 6 may be implemented using oversampled filterbank as shown in Figure 10. AFB 52 of Figure 6 may be implemented by an over-sampled analysis filterbank 114 of Figure 10. The system 50b of Figure 6 may be implemented using WOLA filterbank as shown in Figure 12. AFB 52 of Figure 6 may be implemented by a WOLA analysis filterbank 134 of Figure 12, and SFB 60 of Figure 6 may be implemented by a WOLA synthesis filterbank 138 of Figure 12.

[0082] Delayless subband ANC systems associated with the systems of Figures 4 and 6 are shown in Figures 7 and 8. Figure 7 illustrates a delayless feed-forward subband ANC system 50c in accordance with a further embodiment of the present invention.
Figure 8 illustrates a delayless subband feedback ANC system 50d in accordance with a further embodiment of the present invention. In Figure 7, adaptive processing is implemented in a SAP block 93. In Figure 8, adaptive processing is implemented in a SAP block 94. Each of the systems 50c and 50d includes the components of the system 50a of Figure 4, and further includes a single time-domain filter W(z) 84. The time-domain filter W(z) 84 is an FIR adaptive filter synthesized from subband adaptive filters 56 by SFB 60 and applied for adaptive filtering in time-domain.

[0083] In each of Figures 7 and 8, SFB 60 is provided to convert the SAFs W; (z), i = 0,1,..., K -1 into the single time-domain filter W(z) 84.
Thus, delays due to AFBs are not seen in the signal path. The method to obtain the time-filter from the SAFs is disclosed below in a further embodiment(s) associated with delayless SAF.
Using this method, the processing delay of the filterbank is eliminated from the adaptive processing. As a result, more non-stationary components of the noise can also be cancelled through the digital ANC part of the system.

[0084] The system 50c of Figure 7 may be implemented using oversampled filterbank as shown in Figure 17. AFBs 52 of Figure 7 may be implemented by over-sampled analysis filterbanks 112,114 of Figure 17. The system 50c of Figure 7 may be implemented using WOLA filterbank as shown in Figures 13 and 19. AFB 52 of Figure 7 may be implemented by WOLA analysis filterbanks 132,134 of Figures 13 and 19, and SFB 60 of Figure 7 may be implemented by a WOLA synthesis filterbank 138 of Figures 13 and 19.

[0085] The system 50c of Figure 8 may be implemented using oversampled filterbank.
The system 50c of Figure 8 may be implemented using WOLA filterbank as shown in Figure 22. AFBs 52 of Figure 8 may be implemented by WOLA analysis filterbanks 132,134 of Figure 22, and SFB 60 of Figure 8 may be implemented by a WOLA
synthesis filterbank 138 of Figure 22.

[0086] U.S. Patent Application Publication No. 20030198357 (Serial No.
10/214,056), entitled "Sound Intelligibility Enhancement Using a Psychoacoustics Model and an Over-sampled Fitterbank", discloses the use of ANC in combination with other techniques to improve the intelligibility of audio signals. The sound intelligibility enhancement disclosed in this U.S. application is applicable to the ANC
systems of Figures 3, 4 and 6-8.

[0087] Convergence improvement techniques such as whitening by decimation (WBD), whitening by spectral emphasis (WBS), and whitening by decimation and spectral emphasis (WBDS), disclosed in U.S. Patent Application Publication Nos.
20030108214 and 20040071284 (Serial Nos. 10/214,057 and 10/642,847), can be employed in combination with all methods and systems described in Figures 3, 4, and 6-8.

[0088] A combination of an analog ANC and subband processing is now described in detail. In Figures 9-13, the analog ANC and the subband processing are combined to achieve a higher performance as described below.

[0089] Figure 9 illustrates an ANC system 100a in accordance with a further embodiment of the present invention. The system 100a of Figure 9 includes an analog ANC 105 and subband processing. The analog ANC 105 may be the analog ANC 2 of Figure 1.

[0090] Comparing Figure 9 with Figure 1(b), it can be seen that the analog ANC
system is entirely embedded in the system 100a. The system 100a further includes a second (Reference) microphone that is possibly located outside of the headset earcup to sample the noise. The two microphone signals (x(t), e(t)) are converted to digital, discrete-time signals by the A/D converters 22 and 24 to obtain the signals x(n) and y(n).
The signals are next processed by two (identical) over-sampled analysis filterbanks 112 and 114. The subband processing block 116 processes the over-sampled subband signals (x; (m) and yi (m), i = 0,1,..., K -1) output from the over-sampled analysis filterbanks 112 and 114, and detects various system and signal conditions including the brink of instability of the analog ANC 105. Accordingly, it can tune and adjust parameters of the analog ANC system 105 (such as loop-gain, and loop-filter bandwidth) and/or turn on or off certain features (such as feedback loop) of the system 100a. The outputs of the subband processing block 116 are synthesized by an over-sampled synthesis filterbank 118, which produces the time domain, digital signal z(n).

[0091] The embodiments of the present invention may also employ over-sampled subband processing to provide improved cancellation of periodic or other quasi-stationary signals. Here, the ambient noise (measured by a reference where a microphone is possibly outside the earcup for a feedforward system) is analyzed using subband techniques to determine if there are any stationary (ideally periodic) or quasi-stationary components in the ambient noise. If these components are detected, the DSP will generate a delayed version of this stationary or quasi-stationary signal (shown as z(n) in digital form and z(t) in analog) in Figure 9 and supply it to the analog ANC subsystem (105) for subtraction.

[0092] Adaptive techniques, such as subband adaptive filters on over-sampled filterbanks (OS-SAFs) similar to those disclosed in U.S. Patent Application Publication Nos. 20040071284 and 20030108214 can be employed. The FX-LMS algorithm may be employed in subband methods described in Figures 3, 4, and 6-8, where OS-SAFs and SAPs are employed. This method of combining the analog and digital ANC
provides improved noise cancellation compared to a system that does not employ this technique.

[0093] It is noted that in Figure 9, the reference microphone could have dual usages: 1) to provide information to the subband processing block about the ambient noise in order to control the parameters of the analog ANC system, 2) to provide a reference signal for digital ANC part of the system (employing algorithms such as the FX-LMS). In the system 100a of Figure 9, various versions of the subband FX-LMS, such as feed-forward, feedback, and a combination of the two, may be used for the digital ANC
part of the system.

[0094] In some systems, only one microphone is available (such as feedback ANC).
The system 100a of Figure 9 can be modified to obtain another embodiment disclosed in Figure 10. Figure 10 illustrates a feedback ANC system 100b in accordance with a further embodiment of the present invention. In the system 100b, the subband signals for controlling the analog ANC system 105 are provided only by the error microphone.
The SAP block 92 of Figure 6 can be employed as a part of the subband processing block 116 of Figure 10 to do adaptive processing. Similarly, the SAP block 91 of Figure 4 can be employed in the subband processing block 116 of Figure 9.

[0095] The over-sampled filterbanks may be efficiently implemented using WOLA
analysis and synthesis. Figure 11 illustrates an ANC system 100c in accordance with a further embodiment of the present invention. The system 100c of Figure 11 includes WOLA analysis filterbanks 132 and 134 and a WOLA synthesis filterbank 138.
Efficient hardware realizations of the WOLA have been disclosed in U.S. Patent Nos.
6,236,731 and 6,240,192.

[0096] The embodiments of Figures 7 and 10 may be efficiently implemented using the WOLA filterbanks as shown in Figures 12 and 13, respectively. Figure 12 illustrates a feedback ANC system 100d in accordance with a further embodiment of the present invention. Figure 13 illustrates an ANC system 100e in accordance with a further embodiment of the present invention. The subband processing block 116 of Figure 12 may employ feed-forward FX-LMS strategies, such as the SAP block 94 of Figure 8.
The subband processing block 116 of Figure 13 may employ feed-forward FX-LMS
strategies such the SAP block 93 of Figure 7.

[0097] The subband processing block 116 of Figures 9-13 may model one or more than one acoustic transfer function, a transfer function for a microphone, a transfer function for a loudspeaker, or combinations thereof in accordance with an application.

[0098] Monitoring the amplitude to the noise to be cancelled may save the battery life.
The over-sampled filterbank can also perform this monitoring more accurately using subband processing, since more accurate decision making is possible by monitoring energies of the signals in various subbands. For example, different energy thresholds may be employed for different subbands according to the effectiveness of ANC
in various frequency bands.

[0099] Delayless SAF through over-sampled synthesis/WOLA synthesis is now described in detail.

[00100] Figure 14 illustrates a conventional OS-SAF system 140a. The OS-SAF system 140a includes the over-sampled analysis filterbanks 112 and 114, subband adaptive processing blocks (APBs) 142, and the over-sampled synthesis filterbank 118. The OS-SAF system 140a has two inputs, i.e., a primary input e(n) (e.g., the error microphone signal in the ANC systems of Figures 2-5), and a reference input x(n) (e.g., the reference signal in the ANC systems of Figures 2-5). The reference inputs leaks into the primary input for example by going through an acoustic plant P(s) in Figures 2-5. As a result, the primary and reference inputs become correlated. The OS-SAF system 140a tries to eliminate the portion of the primary input that is correlated with the reference input through adaptive filtering.

[00101] Figures 15(a)-15(b) illustrate examples of adaptive processing. The APBs of Figures 15(a)-15(b) are applicable to APB 142 of Figure 14. The APB of Figure 15(b) contains a summation node 148, while the APB of Figure 15(a) lacks the summation node 148.

[00102] The APB of Figure 15(a) is applicable to the ANC applications, such as the FX-LMS method of Figures 2-6. The summation node 146 is rather transferred to the acoustic domain as shown in the ANC systems of Figures 2-6. The APB of Figure 15(b) is applicable to applications, such adaptive interference (echo or noise) cancellation. Here the interference signal is cancelled in the digital domain through the adaptive algorithms 144. Both of the two APBs in Figures 15(a)-(b) could be equally used in the embodiment of the present invention disclosed here. For brevity, it is assumed that in the description below, the APB 142 has the form of Figure 15(b), unless otherwise stated.

[00103] An example of the disclosed time-filter reconstruction through WOLA
synthesis is described for an echo cancellation application as follows.

[00104] As disclosed in U.S. Patent Application Publication Nos. 20030108214 and No. 20040071284 (Serial Nos. 10/214,057 and 10/642,847), the over-sampled analysis and synthesis filterbank operations can be efficiently implemented using the WOLA analysis and synthesis, respectively. WOLA analysis/synthesis for oversampled filterbank analysis/synthesis is disclosed in U.S. Patent Nos.
6,236,731 and 6,240,192. Figure 16 illustrates an OS-SAF system 140b having WOLA
analysis filterbanks 132 and 134 and a WOLA synthesis filterbank 138.

[00105] Figure 17 illustrates an OS-SAF system 150a in accordance with a further embodiment of the present invention. In the system 150a, the subband adaptive filters Wk (n) are combined together through the filterbank synthesis process to obtain a time-domain adaptive filter W(z) 154 of W (n). The adaptive filter W(z) 154 receives the reference input x(n). The output of the adaptive filter W(z) 154 and the reference input e(n) are summed at 156.

[00106] Figure 18 illustrates an OS-SAF system 150b in accordance with a further embodiment of the present invention. In the AFBs 142 of the system 150b, the AFB of Figure 15(b) is employed.

[00107] Figure 19 illustrates an OS-SAF system 150c in accordance with a further embodiment of the present invention. The system 150c corresponds to the system 150a of Figure 17, and utilizes the WOLA filterbank. As illustrated in Figure 19, the WOLA synthesis could be used to efficiently synthesize the time-domain adaptive filter. The WOLA synthesis process includes steps disclosed in U.S.
Patent Nos. 6,236,731, 6,240,192 and 6,115,478.

[00108] Closed-loop versions of the OS-SAF systems disclosed above are also possible. Figure 20 illustrates a closed-loop version of the system 150a in Figure 17, and Figure 21 discloses a closed-loop version of the system 150c in.Figure 19.
The feedback systems 150d and 150e of Figures 20-21 offer steady-state performance, since the final error signal is sensed back by the system and optimally eliminated.

[00109] Figure 22 illustrates a delayless WOLA-based subband feedback ANC
system using a FX-LMS in accordance with a further embodiment of the present invention. The system 150h of Figure 22 corresponds to the system 50d of Figure 8, and utilizes the WOLA filterbank.

[00110] A. Filter reconstruction for Low-resource delayless subband adaptive filter using WOLA

[00111] Method and system for reconstruction of a time-domain adaptive filter (TAF) by using WOLA synthesis method (implemented with an oversampled filterbank) is described in detail. An inverse fast Fourier transform (IFFT) of length K
is employed in the method. Due to the nature of the WOLA synthesis, the reconstruction process is distributed in time rendering it suitable for real-time implementation. The method is arranged such that segments of the TAF may be used as they become available in time. This makes the method a perfect match for sequential partial update algorithms (described later) that are often integral parts of low-resource implementations. The WOLA synthesis of the subband filters described below is efficiently implemented on an oversampled filterbank that also benefits from the WOLA implementation for its analysis stage. The system is designed and described for an echo cancellation set up though it could be used for other adaptive applications such active noise cancellation or adaptive feedback cancellation.

[00112] A-1. WOLA Synthesis for Filter Reconstruction [00113] Figure 23 illustrates a delayless SAF structure with echo plant. At the output of APBs of Figure 23, the sub-band adaptive filters Pk (z), k = 0,1, =
= =, K - I are obtained. Rather than reconstructing the output signal as in typical SAF
systems, the adaptive filters are passed to a weight transformation stage to obtain the time-domain adaptive filter P(z) to be used to filter the reference signal x(n) in the time-domain.
Assuming a synthesis filter set, in the DFT-FIR approach the TAF is obtained by passing the SAFs through a synthesis filterbank as described by the J. Huo reference and the L. Larson reference:

[00114] P(z) = YFk(z)Pk(ZR) zLs12 k=0 [00115] where R is the filterbank decimation factor, and F0(z) is the prototype filter of the filterbank, bandlimited to 7E/R. Synthesis filters are obtained through discrete Fourier transform (DFT) modulation of the prototype filter as Fk (z) = Fo (z W k) where W = e- j 211 / K . The term zls 12 is added to compensate for delay of the synthesis filter of length Ls [00116] The DFT-FIR uses a polyphase-FFT structure to reconstruct the TAF
through a weight transform of the SAF set. By contrast in this section, a different approach is described whereby the TAF is reconstructed through a WOLA
synthesis of the SAFs. This method is more amenable to a block processing approach and more compatible to hardware implementation. Basically, this method treats the SAFs Pk (m), k=0,1,-.-,K-1, m = 0,1, = = = , M -1 as a set of K subband signals, and passes them through an oversampled filterbank synthesis stage as shown in Figure 24. To efficiently implement the oversampled synthesis stage, we use the WOLA implementation as depicted in Figure 25. For further explanation, consider the SAFs all included in an SAF matrix P, with elements defined as P(k, m) = Pk (m), m = 0,1, = , Nt -1 , k = 0,1,= = =, K -1 . The matrix is set as input to the synthesis stage, one column at every subband time-tick. We call this method of TAF reconstruction "sequential synthesis".
As depicted in Figure 25, the WOLA synthesis starts with taking an IFFT of each column (of length K) of the SAF matrix. After the IFFT 166, and proper circular shifting 168, the vectors of length K are periodically extended to obtain a vector as long as the synthesis window. Next this result is multiplied by the synthesis window followed by an overlap-add operation. Both evenly-stacked FFT and oddly-stacked FFTs may be used. Odd stacking requires an extra sign sequencer to be employed at the final stage. The WOLA synthesis is well described in U.S. Patent Nos.
6,236,731, 6,240,192 and 6,115,478. Assuming aliasing is low in the analysis stage, the SAFs can be shown to converge to the Wiener solution. As a result, the solution will be almost independent of the analysis filter design. Thus the synthesis filter set Fk (z), k = 0,1, = = =, K - I

is designed independently of the analysis filter set to constitute a near perfect-reconstruction set. To obtain the TAF, the output buffer 168 in Figure 25 is first zeroed out. After reading in the input SAF matrix (one column per subband clock tick), the first Ls / 2 samples of the output are discarded. The next L output samples produced a block at a time (R time-samples) constitute the TAF. Thus it takes L/R
input (subband) clock ticks to obtain the TAF. Through optimized implementation it became possible to avoid the initial Ls / 2 sample delays between consecutive filter reconstructions. The total input-output delay for the TAF filter reconstruction is thus (La + Ls) / 2 samples where La denotes the analysis window length. This delay is not seen in the signal path; rather the optimal filter for the reference and primary signals is delayed by this amount. When the echo plant varies slowly (relative to this reconstruction delay), this delay does not degrade the system performance. It is possible to minimize the delay by choosing shorter analysis and synthesis windows as long as distortions in the time-filter due to the reconstruction process are kept within a tolerable range.

[00117] In the conventional SAF systems shorter analysis/synthesis windows will lead to increased output signal degradation since all signals pass through the complete filterbank as described in H. Sheikhzadeh et al. ("Performance Limitations of a New Subband Adaptive System for Noise and Echo Reduction", Proc. Of IEEE
ICECS, 2003). By contrast, in the delayless SAF system of Figure 24, signals are passed through the analysis filterbank only to obtain the adaptive filter. As a result, the adverse effects of shorter analysis/synthesis windows on output signal quality is much less pronounced. It is also possible to further reduce the TAF reconstruction delay to only La/ 2 samples if one is ready to perform filterbank reconstruction of the whole SAF
matrix P for every output block. We call this method of TAF reconstruction "batch synthesis" as opposed to the "sequential synthesis" described in this section.
Batch synthesis will increase the computation cost from a single WOLA synthesis (per output block) in the sequential synthesis to M WOLA synthesis operations. It is also possible to sequentially reconstruct more than one (but less than M) columns of matrix P at every subband time-tick if the computational resources permit.

[00118] A-2 System Evaluation [00119] We evaluate the WOLA filter reconstruction process for a filterbank set up of: analysis and synthesis window lengths of La = 64, and Ls = 128 samples with K =16 subbands and decimation rate of R = 4. The analysis filter is shorter in length (and wider in frequency domain) compared to the synthesis filter. This was chosen to provide better excitation in the analysis filter transition region leading to better convergence behavior as reported by P. L. De Leon II et al ("Experimental Results with Increased Bandwidth Analysis Filters in Oversampled, Subband Acoustic Echo Cancellers," IEEE Sig. Proc. Letters, Jan. 1995, vol. 2, pp 1-3). Each SAF
pk(m) is of length M = 32 resulting in a 16 x 32 SAF matrix P . For comparison, the same filterbank setup was also employed for the oversampled SAF system with the WOLA
implementation depicted in Figure 26. The echo plant was the eighth plant of the ITUT
G. 168 standard (Recommendation ITU-T G. 168, Digital Network Echo Cancellers, Int'l Telecommunication Union, 2000), the Echo Return Loss (ERL) was 10 dB, and random white noise was used at the reference input without any near-end disturbance.
[00120] For subband adaptation, the Gauss-Seidel Pseudo-Affine Projection (GS-PAP) (Albu et al., "The Gauss-Seidel Pseudo Affine Projection Algorithm and its Application for Echo Cancellation," 37`h Asilomar Conf Sig. Sys. & Comp., Pacific Grove, Calif., Nov. 2003) with an affine order of two was employed. The method provides fast convergence and is simple enough to be targeted for a low-resource real-time implementation. GS-PAP may be superior to the Fast Affine Projection Algorithm (FAPA) since unlike the FAPA that operates on a transformed adaptive filter, the GS-PAP directly provides the adaptive filter itself (the Albu et al. reference).

To demonstrate the capability of the WOLA filter reconstruction algorithm in matching sequential update algorithms, sequential update GS-PAP (SGS-PAP) described in H.
Sheikhzadeh et al. ("Partial update subband implementation of complex pseudo-affine projection algorithm on oversampled filterbanks", ICASSP, 2005 IEEE
International Conference on Acoustics, Speech, and Signal Processing, March 18-23, 2005, US) was also used for subband adaptation.

[00121] SGS-PAP employs sequential update of the adaptive filter in the framework of GS-PAP to reduce computation cost. The sequential decimation factor of the SGS-PAP was chosen to be eight (D = 8 ). This means only one polyphase component (of length 32/8 = 4 taps) out of a total of 8 components of each SAF
is adapted at each subband clock tick. For D =1, the SGS-PAP is obviously the same as GS-PAP. In the delayless algorithm, a block of R new samples of the TAF is available every subband tick. This new block is used to update the TAF as soon as it becomes available. This way a smooth and continuous filter reconstruction is achieved.

[00122] Figure 27 shows (A) Time-domain echo plant, and the reconstructed plant with SGS-PAP adaptation, for D=1,8, and (B) error in reconstructed TAFs for D=1 and (C) error for D=8 [00123] Figure 27 illustrates the ERLE results for the SAF system of Figure 26 as well as the delayless WOLA filter synthesis algorithm (Figures 24 and 25), employing SGS-PAP with D =1,8 . As expected, the delayless method achieves a greater ERLE compared to the SAF system, partially due to the fact that the input signals are not passed through the WOLA analysis/synthesis stages. This is consistent with the result and analysis presented in the J. Huo reference and the L.
Larson reference. The SGS-PAP method shows a slight performance degradation for D = 8 (for both the SAF and the delayless methods) since the subband adaptive filters are updated at a much lower rate. Adaptation cost is, however, reduced by a factor of D=8 , [00124] Figure 28 illustrates the ERLE results for SGS-PAP adaptation with D = 1,8 , for the SAF system of Figure 26 and the delayless SAF algorithms of Figure 24.

[00125] Using the WOLA synthesis, the TAF was synthesized. Figure 27-A
shows the synthesized time-filter for SGS-PAP adaptation with D =1, 8 , super-imposed on the ITUT plant. As shown the three impulse responses are almost identical.
To observe the differences, Figures 27-B and 27-C depict time-domain differences between each of the synthesized time-filters and the ITUT plant for D =1, 8 .
As shown, the differences are negligible in both cases, and higher for D = 8 as predicted by the ERLE results.

[00126] The WOLA structure is more amenable with a block processing system (R.E. Crochiere et al. "Multirate digital signal processing", Prentice-Hall, NJ, 1983).
As illustrated in Figure25, for every input vector, all the processes (after the IFFT) occur sequentially to generate R samples of output. All various blocks of Figure 25 can thus operate synchronously with a single subband clock, and there is no need to extra buffering. Since the WOLA filterbank is a block processing system, every slice generated from the IFFT block that must be overlap-added to the previous results has a common exponent. This simplifies the architecture, reducing power and enhancing throughput as described in U.S. Patent Nos. 6,236,731, 6,240,192 and 6,115,478. The same does not hold for straightforward polyphase implementation of the J. Huo reference since it is a stream processing method. In the polyphase filterbank synthesis of the J. Huo reference, a separate convolution of one of polyphase synthesis filters (with one of the K subband signals) has to occur for each sample of output. To do this, K different data buffers have to be updated. To summarize, the WOLA synthesis of Figure 25 provides a simpler, more modular structure for real-time hardware implementation as described in U.S. Patent Nos. 6,236,731, 6,240,192 and 6,115,478.
[00127] The method, based on WOLA synthesis of the SAFs, is efficient and is well mapped to a low-resource hardware implementation. Also the WOLA adaptive filter reconstruction may easily be spread out in time simplifying the necessary hardware. This time-spreading may be easily combined with partial update adaptive algorithms to reduce the computation cost for low-resource real-time platforms.
[00128] A-3 Filter Reconstruction and Evaluation [00129] An example for reconstruction by WOLA is described, using WOLA
setup of analysis window length of L = 128, with K = 32 subband, and decimation rate of R = 8. Assume (without loss of generality) that APB's are of the form illustrated in Figure 15(b) employed for an echo cancellation application. The collection of K SAFs for subbands k = 0,1,..., K -1 is shown in Figure 25. Each SAF Wk (m) is of length M, n = 0,1,..., M -1 resulting in a K by M SAF matrix. The system 150c of Figure 19 for echo cancellation is employed using the eighth plant of the ITUT G. 168 standard for echo generation. The reconstruction process of Figure 25 was applied to this simulation.
[00130] Figure 29 shows an example of real and imaginary parts of SAFs for subbands k = 0,1,2,3 for the subband adaptive system after convergence.
Treating each SAF as a subband signal, the process of WOLA synthesis starts with taking the IFFT of each column (of length K) of the SAF matrix. There are M columns to be processed.
After IFFT, and proper circular shifting, the vectors of length K are periodically extended to obtain vectors as long as the synthesis window. The result of periodic extension is shown in Figure 30 for the SAFs shown in Figure 29, where a total length of 256 samples (8 periodic extensions) is used. Then a synthesis window (of length Ls = 256 samples) is applied to the periodically extended signals. A typical window is shown in Figure 31 and the result of window application is shown in Figure 32.
The last step is the overlap-add of the vectors (a total of M vectors involved). Using a properly designed synthesis window, the time-domain filter is synthesized in time as shown in Figure 33. Figure 34(a) shows the synthesized time-filter super-imposed on the ITUT
plant. As shown, the synthesized time-filter and the ITUT plant are almost identical.
Figure 34(b) illustrates the time-domain difference between the synthesized time-filter and the ITUT plant. As shown, the difference is negligible.

[00131] The time-domain filter can be used for adaptive filtering. As illustrated in Figure 35, for random noise input and the same ITUT plant, the echo attenuation of the time-domain filter is superior to the SAF filter using the NLMS algorithm.
This is partially due to the fact that the input signal to the time-domain filter is not passed through the WOLA analysis stage (hence not experiencing distortions), as shown in Figure 4.

[00132] It is noted that as for the method of adaptation of SAFs (e.g., NLMS
30, 58), the Normalized Least Mean Square (NLMS), the Affine Projection Algorithm (APA) and its variants such as Fast APA (FAPA), or the Recursive Least Squares (RLS) may also be used.

[00133] The corresponding Canadian Patent Application No. 2,481,629 also describes reconstruction of TAF by WOLA.

[00134] The embodiments of the present invention reduces acoustic feedback and provide as high a degree of cancellation for periodic or other quasi-stationary signals.

[00135] The embodiments of the present invention automatically adapts to the different acoustic situations presented by different headsets and to the normal variations encountered in production parts.

[00136] As described above, over-sampled filterbank processing and system architecture disclosed in U.S. Patent Nos. 6,240,192 B 1, 6,236,731 B 1, and 6,115,478 provide low group delay, low power and small size. The embodiments of the present invention can be efficiently implemented on the system architecture disclosed in U.S.
Patent Nos. 6,240,192 B 1, 6,236,731 B 1, and 6,115,478.

[00137] As described above, an analog ANC provides noise cancellation at predominantly low frequencies (below 1500 Hz to 2000 Hz). The embodiments of the present invention can extend this frequency range. The introduction ofa DSP
processor also permits additional processing to be incorporated, including techniques that can improve speech intelligibility at frequencies where even DSP enhanced analog ANC
ceases to provide benefit, as described in U.S. Patent Application Publication No.
20030198357 (Serial No. 10/214,056). U.S. Patent Application Publication No.
20030198357 discloses a sound intelligibility enhancement (SIE) using a psychoacoustic model and preferably an oversampled filterbank. When combined with ANC, the SIE performs better since ANC provides more benefit at low frequencies while SIE provides more benefit at high frequencies. Using the ANC , undesired dynamic range reduction is avoidable due to low-frequency noise.

f [00138] The embodiments of the present invention disclosed above prevents acoustic feedback in a manner that retains high fidelity and good performance in the presence of typical disturbances. Subband acoustic feedback reduction is employed to extend the operating frequency range by permitting more loop gain to be introduced before the onset of acoustic feedback. This type of feedback cancellation introduces fewer artifacts than full band approaches for feedback cancellation. It also provides better performance in the presence of coloured noise (disturbances). Finally the reduced group delay (compared to a full band system with similar feedback cancellation performance) provides a faster response time for feedback cancellation.

[00139] As detailed in the U.S. Patent No. 6,118,878, the brink of instability can be detected by finding the ratio of external to internal noise at high frequencies.
Alternatively, the residual signal can be monitored at various frequencies to detect and prevent an impending instability. Employing an over-sampled filterbank provides more accurate and reliable prediction of impending instabilities. Generally, the overall control of the ANC system employing an over-sampled subband approach is more efficient, and accurate.

[00140] The delayless OS-SAF in accordance with the embodiments of the present invention considerably reduces the delay introduced in the primary signal path.
Also, since the actual echo cancellation occurs in time-domain, inevitable distortions due to over-sampled filterbank analysis/synthesis is avoided. As the results demonstrate, the delayless SAF method proposed outperforms the traditional OS-SAFs in terms of adaptive noise and echo cancellation performance. This is possible due to avoiding errors due to band edges in the OS-SAF system.

[00141] The embodiments of the present invention provide a combination of OS-SAF and time-domain filtering. Comparing to a full band system with similar adaptive processing, this combination provides a faster convergence and response time of the adaptive system.

[00142] The embodiments of the present invention may be implemented by hardware, software or a combination of hardware and software having the above described functions. The software code, either in its entirety or a part thereof, may be stored in a computer readable medium. Further, a computer data signal representing the software code which may be embedded in a carrier wave may be transmitted via a communication network. Such a computer readable medium and, a computer data signal and carrier wave are also within the scope of the present invention, as well as the hardware, software and the combination thereof.

[00143] The present invention has been described with regard to one or more embodiments. However, it will be apparent to persons skilled in the art that a number of variations and modifications can be made without departing from the scope of the 1o invention as defined in the claims.

Claims

1. A system for active noise cancellation, comprising:

a first analysis filterbank for transferring a reference signal in a time-domain into a plurality of subband reference signals in a frequency-domain, the reference signal being associated with noise;

a second analysis filterbank for transferring a primary signal in the time-domain, into a plurality of subband primary signals in the frequency-domain, wherein the primary signal is associated with an acoustic signal, a noise-cancelling signal and may be contaminated by the noise;

a subband processing module for processing the plurality of subband reference signals, the plurality of subband primary signals or a combination thereof, the subband processing module comprising:

a subband estimator for estimating a subband acoustic transfer function, the subband acoustic transfer function modelling conversion of an electrical signal into an acoustic domain signal, the subband estimator being applied to the plurality of subband reference signals;

a subband adaptive filter for providing a plurality of subband output signals based on the plurality of subband reference signals; and an adjustor for adapting the subband adaptive filter based on the plurality of subband primary signals and the plurality of estimated subband reference signals; and a synthesis filterbank for transferring the plurality of subband output signals to a time-domain output signal, the time-domain output signal being used to generate the noise-cancelling signal.

2. A system as claimed in claim 1, wherein each of the analysis filterbanks comprises:

an oversampled analysis filterbank.

3. A system as claimed in claim 2, wherein the oversampled analysis filterbank comprises:

an WOLA analysis filterbank.

4. A system as claimed in any one of claims 1 to 3, wherein the synthesis filterbank comprises:

an oversampled synthesis filterbank.

5. A system as claimed in claim 4, wherein the oversampled synthesis filterbank comprises:

an WOLA synthesis filterbank.

6. A system as claimed in any one of claims 1 to 5, wherein the estimator outputs the transfer function from a speaker to a microphone.

7. A system as claimed in any one of claims 1 to 6, wherein the primary signal includes a microphone signal associating with a local audio signal and the time-domain output signal.

8. A system as claimed in any one of claims 1 to 6, comprising:

an analog active noise cancellation system receiving the primary signal.

9. A system as claimed in claim 8 wherein the primary signal includes a microphone signal associating with a local audio signal, the time-domain output signal and an output from the analog active noise cancellation system.

10. A system as claimed in any one of claims 1 to 6, comprising:

a time-domain adaptive filter synthesized from the subband adaptive filter, for filtering the reference signal.

11. A system as claimed in claim 10, wherein the primary signal includes a microphone signal associating with a local audio signal and an output from the time-domain adaptive filter.

12. A system as claimed in claim 10, comprising:

an analog active noise cancellation system receiving the primary signal.

13. A system as claimed in claim 12, wherein the primary signal includes a microphone signal associating with a local audio signal, the time-domain output signal and an output from the analog active noise cancellation system.

14. A system for active noise cancellation, comprising:

an analysis filterbank for transferring a primary signal in a time-domain into a plurality of subband primary signals in a frequency-domain, the primary signal being associated with an acoustic signal, a noise-cancelling signal and may be contaminated with noise;

a subband processing module for processing the plurality of subband primary signals, the subband processing module comprising:

a subband adaptive filter for providing a plurality of subband output signals based on a plurality of subband reference signals;

a first subband estimator for estimating a subband acoustic transfer function, the subband acoustic transfer function modelling conversion of an electrical signal into an acoustic domain signal, the first subband estimator being applied to the plurality of subband output signals;

a signal path for generating the plurality of subband reference signals based on the plurality of subband primary signals and the plurality of estimated subband output signals;

a second subband estimator for estimating the subband acoustic transfer function, the second subband estimator being applied to the plurality of subband reference signals; and an adjustor for adapting the subband adaptive filter based on the plurality of subband primary signals and the plurality of estimated subband reference signals; and a synthesis filterbank for transferring the subband output signals into a time-domain output signal, the time-domain output signal being used to generate the noise-cancelling signal.

15. A system as claimed in claim 14, wherein the analysis filterbank comprises:
an oversampled analysis filterbank.

16. A system as claimed in claim 15, wherein the oversampled analysis filterbank comprises:

an WOLA analysis filterbank.

17. A system as claimed in any one of claims 14 to 16, wherein the synthesis filterbank comprises:

an oversampled synthesis filterbank.

18. A system as claimed in claim 17, wherein the oversampled synthesis filterbank comprises:

an WOLA synthesis filterbank.

19. A system as claimed in any one of claims 14 to 18, wherein each of the estimator outputs the corresponding transfer function from a speaker to a microphone.

20. A system as claimed in any one of claims 14 to 19, wherein the primary signal includes a microphone signal associating with a local audio signal and the time-domain output signal.

21. A system as claimed in any one of claims 14 to 19, comprising:

an analog active noise cancellation system receiving the primary signal.

22. A system as claimed in claim 21 wherein the primary signal includes a microphone signal associating with a local audio signal, the time-domain output signal and an output from the analog active noise cancellation system.

23. A system for active noise cancellation, comprising:

a first analysis filterbank for transferring a primary signal in a time-domain into a plurality of subband primary signals in a frequency-domain, the primary signal being associated with an acoustic signal, a noise-cancelling signal and may be contaminated with noise;

a second analysis filterbank for transferring a reference signal in the time-domain into a plurality of subband reference signals in the frequency-domain;

a time-domain adaptive filter for providing a time-domain output signal based on the reference signal in the time-domain, the time-domain output signal being used to generate the noise-cancelling signal;

a time-domain estimator for estimating an acoustic transfer function, the acoustic transfer function modelling conversion of an electrical signal into an acoustic domain signal, the time-domain estimator being applied to the time-domain output signal;

a signal path for providing the reference signal in the time-domain based on the primary signal in the time-domain and the estimated time-domain output signal;

a subband processing module for processing the plurality of subband reference signals, the plurality of subband primary signals or a combination thereof, the subband processing module comprising:

a subband estimator for estimating the acoustic transfer function, the subband estimator being applied to the plurality of subband reference signals;

and an adjustor for adapting a subband adaptive filter based on the plurality of subband primary signals and the plurality of estimated subband reference signals; and a synthesis filterbank for synthesizing the time-domain adaptive filter from the subband adaptive filter.

24. A system as claimed in claim 23, wherein each of the analysis filterbanks comprises:

an oversampled analysis filterbank.

25. A system as claimed in claim 24, wherein the oversampled analysis filterbank comprises:

an WOLA analysis filterbank.

26. A system as claimed in any one of claims 23 to 25, wherein the synthesis filterbank comprises:

an oversampled synthesis filterbank.

27. A system as claimed in claim 26, wherein the oversampled synthesis filterbank comprises:

an WOLA synthesis filterbank.

28. A system as claimed in any one of claims 1 to 27, wherein the adjustor implements NMSL adaptation algorithm.

29. A method for active noise cancellation, comprising:

transferring a reference signal in a time-domain into a plurality of subband reference signals in a frequency-domain, the reference signal being associated with noise;

transferring a primary signal in the time-domain, into a plurality of subband primary signals in the frequency-domain, wherein the primary signal is associated with an acoustic signal, a noise-cancelling signal and may be contaminated by the noise;

estimating a subband acoustic transfer function for the plurality of subband reference signals, the subband acoustic transfer function modelling conversion of an electrical signal into an acoustic domain signal;

providing a plurality of subband output signals based on the plurality of subband reference signals by a subband adaptive filter;

adapting the subband adaptive filter based on the plurality of subband primary signals and the plurality of estimated subband reference signals; and transferring the plurality of subband output signals to a time-domain output signal, the time-domain output signal being used to generate the noise-cancelling signal.

30. A method as claimed in claim 29, wherein the step of modeling comprises:
outputting the transfer function from a speaker to a microphone.

31. A method as claimed in claim 29 or 30, wherein the primary signal includes a microphone signal associating with a local audio signal and the time-domain output signal.

32. A method as claimed in claim 29 or 30, comprising:

implementing an analog active noise cancellation to the primary signal.

33. A method as claimed in claim 32 wherein the primary signal includes a microphone signal associating with a local audio signal, the time-domain output signal and an output from the analog active noise cancellation step.

34. A method as claimed in any one of claims 29 to 30, comprising:

synthesizing a time-domain adaptive filter from the subband adaptive filter, the time-domain adaptive filter filtering the reference signal.

35. A method as claimed in claim 34, wherein the primary signal includes a microphone signal associating with a local audio signal and an output from the time-domain adaptive filter.

36. A method as claimed in claim 34, comprising:

implementing an analog active noise cancellation to the primary signal.

37. A system as claimed in claim 36, wherein the primary signal includes a microphone signal associating with a local audio signal, the time-domain output signal and an output from the analog active noise cancellation step.

38. A method for active noise cancellation, comprising:

transferring a primary signal in a time-domain into a plurality of subband primary signals in a frequency-domain, the primary signal being associated with an acoustic signal, a noise-cancelling signal and may be contaminated with noise;

providing a plurality of subband output signals based on a plurality of subband reference signals by a subband adaptive filter;

transferring the plurality of subband output signals into a time-domain output signal, the time-domain output being used to generate the noise-cancelling signal;
estimating a subband acoustic transfer function for the plurality of subband output signals, the subband acoustic transfer function modelling conversion of an electrical signal into an acoustic domain signal;

generating the plurality of subband reference signals based on the plurality of subband primary signals and the plurality of estimated subband output signals;

estimating the subband acoustic transfer function for the plurality of subband reference signals; and adapting the subband adaptive filter based on the plurality of subband primary signals and the plurality of estimated subband reference signals.

39. A system as claimed in claim 38, wherein each step of modeling comprises:
outputting the corresponding transfer function from a speaker to a microphone.

40. A system as claimed in claim 38 or 39, wherein the primary signal includes a microphone signal associating with a local audio signal and the time-domain output signal.

41. A system as claimed in claim 38 or 39, comprising:

implementing an analog active noise cancellation to the primary signal.

42. A system as claimed in claim 41 wherein the primary signal includes a microphone signal associating with a local audio signal, the time-domain output signal and an output from the analog active noise cancellation step.

43. A method for active noise cancellation, comprising:

transferring a primary signal in a time-domain into a plurality of subband primary signals in a frequency-domain, the primary signal being associated with an acoustic signal, a noise-cancelling signal and may be contaminated with noise;

transferring a reference signal in the time-domain into a plurality of subband reference signals in the frequency-domain;

providing a time-domain output signal based on the reference signal in the time-domain by a time-domain adaptive filter, the time-domain output signal being used to generate the noise-cancelling signal;

estimating an acoustic transfer function for the time-domain output signal, the acoustic transfer function modelling conversion of an electrical signal into an acoustic domain signal;

providing the reference signal in the time-domain based on the primary signal in the time-domain and the estimated time-domain output signal;

estimating the acoustic transfer function for the plurality of subband reference signals;

adapting a subband adaptive filter based on the plurality of subband primary signals and the plurality of estimated reference signals; and synthesizing the time-domain adaptive filter from the subband adaptive filter.

44. A method as claimed in any one of claims 28 to 43, wherein the step of adapting comprising implementing NMSL adaptation algorithm.

45. A system as claimed in claim 1, further comprising:

a time-domain adaptive filter for receiving the time-domain output signal and the reference signal in the time-domain, for generating the noise-cancelling signal.

46. A system as claimed in any one of claims 8, 12, and 21, wherein the subband processing module adjusts one or more parameters of the analog active noice cancellation system based on the processing of the plurality of subband reference signals, the plurality of subband primary signals, or a combination thereof.

47. A method as claimed in claim 29, further comprising:

receiving the time-domain output signal and the reference signal in the time-domain to generate the noise-cancelling signal.

48. A method as claimd in any one of claims 32, 36, and 41, further comprising:
adjusting one or more parameters of the analog active noise cancellation based on the processing of the plurality of subband reference signals, the plurality of subband primary signals, or a combination thereof.