Electrical computers: arithmetic processing and calculating – Electrical digital calculating computer – Particular function performed
Reexamination Certificate
2000-11-24
2004-02-03
Malzahn, David H. (Department: 2124)
Electrical computers: arithmetic processing and calculating
Electrical digital calculating computer
Particular function performed
Reexamination Certificate
active
06687723
ABSTRACT:
FIELD OF INVENTION
The invention relates to an adaptive filter, and in particular, to the adaptive filter having a hybrid structure and providing a tri-mode operation.
BACKGROUND OF THE INVENTION
Adaptive filtering is a digital signal processing technique that has been widely used in numerous signal processing applications such as echo cancellation, noise cancellation, channel equalization and system identification. Characteristics of an adaptive filter are largely determined by its adaptation algorithm, and the choice of the adaptation algorithm in a specific adaptive filtering system directly affects the performance of the system.
Being simple and easily stable, the normalized least mean square (NLMS) adaptation algorithm is now widely used in the industry with a certain degree of success. However, because of its intrinsic weakness, this algorithm converges slowly, especially with colored training signals, e.g. speech signals. As a result, the performance of systems incorporating NLMS adaptive filters often suffers from the slow convergence nature of the algorithm.
To solve the problem of slow convergence, it has been proposed to use a modification of an Affine Projection algorithm which is known as Fast Affine Projection (FAP) algorithm, see e.g. publication by Steven L. Gay and Sanjeev Tavathia (Acoustic Research Department, AT&T Bell Laboratories), “The Fast Affine Projection Algorithm,” pp. 3023-3026, Proceedings of the International Conference on Acoustics, Speech, and Signal Processing, May 1995, Detroit, Mich., U.S.A. and U.S. Pat. No. #5,428,562 to Gay. The FAP converges several times faster than NLMS, with only a marginal increase in implementation complexity. However, a stability issue has been preventing FAP from being used in the industry.
A remedy for the instability problem associated with the FAP algorithm has been proposed in two U.S. patent applications to Heping Ding Ser. Nos. 09/218,428 and 09/356,041 filed Dec. 22, 1998 and Jul. 16, 1999 respectively, wherein a new Stable FAP (SFAP) filter and method have been invented. The cited patent applications are incorporated herein by reference. The SFAP filter solves the instability problem of FAP caused by error accumulation in an inversion process of an auto-correlation matrix while providing the advantages of other known methods.
Although the development of fast convergence adaptive methods may be considered as a giant leap forward, there is still a problem remaining which is common to all existing adaptive filters while most severely influencing the performance of the fast convergence adaptive filters. The problem is that the adaptive filter performs satisfactory only in the absence of the near-end signal or when the near end signal is negligibly small.
To describe this problem in more detail, let us refer to
FIG. 1
, which illustrates a typical echo-cancellation system
10
with an embedded adaptive filter
100
. A digitallysampled far-end reference input signal x(n) is supplied to the adaptive filter
100
and to an echo path
14
producing an unwanted signal u(n), the signal being an echo of the input signal x(n) through the echo path
14
. The unwanted signal u(n) is mixed up with the wanted near-end signal s(n) in a summer
16
to produce a response signal d(n). The response signal d(n) is sent to another summer
18
together with an echo estimate signal y(n) generated by the adaptive filter
100
. The summer
18
subtracts y(n) from d(n) producing an output signal e(n) to be transmitted to the far-end. The echo path
14
can be constantly changing, and the adaptive filter
100
must continuously adapt to the new echo path. Therefore the echo estimate signal y(n) must be as close to u(n) as possible, so that the latter is largely cancelled by the former to ensure that e(n) best resembles s(n). The output signal e(n), called the error signal, is then transmitted to the far-end and at the same time sent to the adaptive filter
100
which uses the signal to adjust the filter coefficients.
If the near-end signal s(n) is not small enough to ensure that the response signal d(n) is purely an echo signal u(n), the response signal d(n) contains components of the near end signal s(n) which can be short-term correlated with x(n). This is because both x(n) and s(n) are speech signals, which often show short-term correlations with each other even if they come from different sources. As a result, the adaptive filter
100
may mis-converge, and the near-end signal s(n) component in e(n) may be distorted, because the adaptive filter attempts to track and cancel components in the near-end signal s(n) that are short-term correlated to the input far-end signal x(n). The faster the convergence of the adaptive filter's adaptation algorithm, the stronger the influence of the near end signal s(n) on the performance of the filter operation, causing the filter to mis-converge and to distort the near-end signal itself in the output e(n).
Accordingly, there is a need for development of an adaptive filter and method which would provide high performance operation in the presence of the near-end signal.
SUMMARY OF THE INVENTION
It is therefore an object of the invention to provide an adaptive filter and a method of adaptive filtering which would avoid the afore-mentioned problems.
According to one aspect of the present invention there is provided a method of adaptive filtering, comprising the steps of:
providing operation of an adaptive filter in one of the three modes, the modes being a fast converging mode, a slow converging mode and a freezing mode;
measuring non-convergence of the filter and a near-end signal, and generating a feedback signal based on the measurements; and
switching between the modes of operation of the filter in response to the feedback signal.
Convenietly, the step of measuring non-convergence and near-end signal comprises comparing the measured values with predetermined threshold values for non-convergence and near end signals respectively.
The step of switching between the modes of operation of the filter comprises switching the adaptive filter to the freezing mode when the near-end signal is above the corresponding threshold value, and non-convergence is below the corresponding threshold value. Alternatively, when non-convergence is above the threshold value and the near-end signal is below the threshold value, the filter is switched to the fast converging mode. Yet alternatively, when non-convergence and near-end signal are both below the corresponding threshold values, the adaptive filter is switched to the slow converging mode. When the non-convergence and near-end signal are both above the corresponding threshold values, the filter may be switched to either one of the three modes depending on the system requirements.
Advantageously, operation of the adaptive filter in the slow converging mode comprises adaptive filtering in accordance with a normalized least mean square (NLMS) method, and operation of the filter in the fast converging mode comprises providing adaptive filtering in accordance with a fast affine projection (FAP) method or Stable fast affine projection method (SFAP). Beneficially, the adaptive filtering in accordance with the SFAP method comprises using a steepest descent method or a conjugate gradient method.
According to another aspect of the invention there is provided a method of adaptive filtering, comprising the steps of:
(a) providing initial adaptive filtering for an incoming signal by one of adaptive filtering methods;
(b) measuring non-convergence of the adaptive filtering method and a near-end signal;
(c) comparing the measured values of the non-convergence and near-end signal with pre-determined threshold values;
(d) when the near-end signal is above the corresponding threshold value and non-convergence is below the corresponding threshold value, freezing coefficients in the currently used adaptive filtering method;
(e) when non-convergence is above the corresponding threshold value and the near-end signal is below the corresponding threshold value, providing adaptive
de Wilton Angela C.
Malzahn David H.
Nortel Networks Limited
LandOfFree
Tri-mode adaptive filter and method does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Tri-mode adaptive filter and method, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Tri-mode adaptive filter and method will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3304684