Electrical audio signal processing systems and devices – Noise or distortion suppression – Using signal channel and noise channel
Reexamination Certificate
1997-04-14
2001-01-23
Chang, Vivian (Department: 2747)
Electrical audio signal processing systems and devices
Noise or distortion suppression
Using signal channel and noise channel
C381S092000, C381S094200
Reexamination Certificate
active
06178248
ABSTRACT:
BACKGROUND OF THE INVENTION
The present invention relates generally to signal processing, and more specifically to an interference cancelling system and method using a combination of adaptive and non-adaptive filter processing. A system using such a combination of adaptive and non-adaptive filter processing is referred to herein as a dual-processing system.
Interference cancelling systems have a wide range of applications such as directional microphones and hearing aids. An interference cancelling system amplifies a target signal originating from a target source (information source) while suppressing interfering signals (“interferences”) originating from interference or noise sources.
Interference cancelling systems using adaptive filters are well-known in the art. An adaptive filter is a filter which can change its characteristics by changing its filter coefficients. The interference cancelling system may be a non-directional system having one or more sensors measuring the signal received from the target to generate a main channel, which has a target signal component and an interference component. The system may include one or more other sensors for measuring the interferences to generate one or more reference channels The adaptive filter uses the reference channels to cancel the interference component present in the main channel.
Alternatively, the system may be a directional system, well-known in the art, which amplifies a target signal originating from a target source at a particular direction relative to the system and suppresses interferences originating from interference sources at all other directions. In such a directional system, the target signal and the interferences may be detected by an array of spatially distributed sensors forming what is called a beamformer.
A beamformer is a form of spatial filter, itself well-known in the art, which takes inputs from an array of spatially distributed sensors and combines them in such a way that it either enhances or suppresses signals coming from certain directions relative to signals from other directions. Thus it can change the direction of receiving sensitivity without physically moving the sensor array. The inputs are combined for this purpose based on filter coefficients as discussed below.
In non-adaptive beamforming, the filter coefficients of a beamformer are predetermined such that the beamformer can exhibit maximum sensitivity or minimum sensitivity (null) in a predetermined direction. Since the coefficient values are fixed in time, a non-adaptive beamformer cannot dynamically place nulls in the directions of strong interferences existing at particular times as the environment changes.
In adaptive beamforming, in contrast, the spatial filter coefficients of a beamformer are continually updated so that directional sensitivity can be dynamically changed depending on the changing locations of a target source and interference sources. For more details on beamforming, see Van Veen & Buckley, Beamforming: A Versatile Approach to Spatial Filtering, IEEE ASSP Magazine, April 1988, pp. 4-24.
An adaptive beamformer can be implemented for example by using tapped delay lines, forming a finite-impulse-response (FIR) filter having time-varying coefficients which are directly changed as the locations of interference sources change.
Alternatively, the adaptive beamformer can be implemented using an adaptive filter (dealing with temporal signals rather than spatial signals). The adaptive beamformer uses fixed-coefficient tapped delay lines, called a main-channel matrix, to obtain a signal received from the direction of a target and other fixed-coefficient tapped delay lines, called a reference-channel matrix, to obtain interferences received from all other directions. An adaptive filter is used to generate cancelling signals resembling the interferences changing in direction. In this manner, instead of directly changing the coefficients of the tapped delay lines, the implementation achieves the same effect by changing the characteristics of the adaptive filter. The adaptive filter generally subtracts the cancelling signals from the main channel and adjusts the filter weights to minimize the mean-square values of the output. When the filter weights settle, the cancelling signals closely track the interferences so that the output has substantially reduced interference.
For some applications, it is important to be able to process a broadband input, that is, one having a relatively large bandwidth. For example, in hearing applications, speech intelligibility is critical to performance. It is well known that the higher frequency portion of the speech spectrum carries much of the information required for speech intelligibility. For applications such as hearing aids or directional microphones for voice activation systems, good intelligibility requires at least 6 Khz of bandwidth. In fact, professional audio systems will not tolerate a bandwidth of less than 12 Khz.
This bandwidth requirement imposes a severe computational burden on the interference cancelling system using adaptive filter processing. Adaptive filter processing is inherently intensive in computation. It involves performing filter operations to produce an output and further updating filter weights based on the output. All these operations must be performed for each new sample.
In order to extend the operation of an adaptive filter in the discrete time domain from any bandwidth to a broader bandwidth, the sampling rate should be increased to maintain comparable quality. According to the well-known sampling theorem, a sampling rate of at least twice the maximum frequency of an incoming analog signal is required in order to represent the signal completely in the discrete time domain. The increased sampling rate increases the number of operations to be performed per unit time.
Increasing the sampling rate alone is not, however, enough to handle the broader bandwidth. An adaptive filter acts on later samples by observing earlier samples within a given period, as feedback. How well the adaptive filter can react depends on how long the filter can observe the earlier samples. This time period is called an effective time delay through an adaptive filter. The delay is proportional to the number of filter stages, each storing a filter coefficient, divided by the sampling frequency. If the sampling frequency is increased, the number of filter stages should be increased in order to maintain the same effective time delay. The increased number of filter stages also increases the number of operations that must be performed per unit time.
The combination of increasing sampling rate and increasing the number of required filter stages sharply increases the number of operations to be performed by a processor. Thus a simple extension of adaptive filter processing to a broader bandwidth places a disproportionately large computational burden on the system and hence is not desirable.
The simple extension of adaptive filter processing presents another problem for an interference cancelling system using adaptive filter processing. Adaptive interference cancelling systems suffer from signal leakage. The system works well when the reference channel is uncorrelated to the main channel. However, in practice, the reference channel contains some signals correlated to the main channel due to signal leakage from the main channel itself. Adaptive filter processing may then partly cancel the target signal as well the inferences. The signal leakage is more likely to occur at higher frequencies for the following reason.
The reference-channel matrix produces reference channels by creating a null in the target direction (by suppressing signals from the target direction). In order to suppress the signals from the target direction effectively, the null should be as deep as possible in the target direction. The null should also be wide enough to provide some tolerance to those signals slightly off the target direction. It turns out that the null is much wider at lower frequencies than at higher frequencies. Therefor
Andrea Electronics Corporation
Chang Vivian
Frommer Lawrence & Laug LLP
Kowalski Thomas J.
LandOfFree
Dual-processing interference cancelling system 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 Dual-processing interference cancelling system and method, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Dual-processing interference cancelling system and method will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2440882