Electrical computers: arithmetic processing and calculating – Electrical digital calculating computer – Particular function performed
Reexamination Certificate
1996-11-27
2004-08-03
Malzahn, David H. (Department: 2124)
Electrical computers: arithmetic processing and calculating
Electrical digital calculating computer
Particular function performed
C708S314000
Reexamination Certificate
active
06772182
ABSTRACT:
BACKGROUND OF THE INVENTION
The present invention relates generally to noise filters, and more particularly, to a matched-phase noise filter and associated method which enables a signal which is extremely weak relative to associated noise to be extracted from an extremely noisy channel.
Presently available methods for enhancing the signal-to-noise ratio (SNR) of a noisy recording include, in increasing order of sophistication, bandpass filtering, optimal filtering, averaging, and noise subtraction. Bandpass filtering is effective when the signal and the noise are known to occupy different portions of the frequency spectrum, with the pass band(s) being set to allow passage of the signal and to reject the noise. Optimal filtering (which is a generalization of bandpass filtering) is effective to improve the SNR when the frequency spectrum of the signal is known, with the shape of the filter frequency curve being set to match that of the original spectrum. With the averaging approach, many recordings of the same signal received at different times are averaged. Presumably, the noise in the overlaid recordings is canceled by the mechanism of destructive interference, while the signal is obtained by the mechanism of constructive interference. With this approach, it is necessary to accurately detect a timing mark within the signal time series in order to obtain constructive signal interference among the different recordings. The signal processing gain for this approach is equal to the number of independent recordings averaged. The noise subtraction approach can be utilized when a precise recording of the noise field is made simultaneously with the reception and recording of the noisy signal. With this approach, the noise time series is subtracted from the recording containing the noisy signal to thereby yield a “clean” signal with the noise removed therefrom. An illustrative example of a device which employs this technique in the field of law enforcement works in the below-described manner. First, a conversation between criminal suspects is deliberately obscured by increasing the volume on a nearby radio or television, and law enforcement personnel recording the obscured conversation simultaneously make a “clean” recording of the same radio or television program at another location. The “clean” recording of the “noise” signal (i.e., the recorded radio or television program) is later subtracted from a recording of the obscured conversation, thereby yielding a “clean” signal which can be made to reveal the content of the conversation.
Although the first three noise filtering techniques are adequate for many applications, they each attempt to recover a signal from a noisy background by recognizing features of the signal (i.e., by its expected time signature or spectral domain), and thus, in addition to other limitations and shortcomings, are ineffective when no a prior information about the received signal is available. The final technique, in contrast, becomes ineffective unless a pure recording of the “noise” is available at precisely the time of interest.
In the past, attempts were made to suppress noise in cases in which the strength of the signal is extremely weak relative to that of the noise by utilization of such array processing techniques as matched-field processing and beam forming, which typically exploit the spatial structure of the acoustic field through multiple sensors to effect a gain. Thus, these array processing techniques are also ineffective in systems in which a single phone or sensor, whether moving or stationary, is available. Furthermore, the presently available signal processing techniques do not make full use of all that is known about a background noise field in order to improve the SNRs. For example, the Heard Island Feasibility Test disclosed in an article entitled “The Heard Island Feasibility Test”,
J. Acoust. Soc. Am
. 96, 2343-2352 (1994), averaged many receptions of a repeatedly transmitted sequence as part of a “pulse compression” strategy in order to improve SNR. This strategy relies on being able to align receptions of carefully coded signal segments before averaging them together, achieving gain by virtue of the vanishing long-term time average of the noise field. No attempt was made to “recognize” the noise field by its spectral profile, and thus use the knowledge of what noise “sounds like” in order to reject it.
Based on the above, it can be appreciated that there presently exists a need in the art for a noise filtering method (and noise filter for implementing the same) which improves signal-to-noise ratios (SNRs) in noise dominated data, utilizing information concerning the shape of the noise spectrum, even in cases of very-low-signal-to-noise-ratio (VLSNR) where no a priori information about the signal is known, and no a priori information about the phase of the noise field is known. The present invention fulfills this need in the art.
SUMMARY OF THE INVENTION
Accordingly, an overall object of the present invention is to improve the signal-to-noise ratio (SNR) of a noise-dominated recording made on a single receiver (as opposed to an array) by removing the majority of the noise, which signal component is recognized by the shape of its power spectrum.
An object according to the present invention is to provide an acoustic filter which permits the use of the known shape of the noise spectrum for purposes of noise rejection without a priori knowledge of the relative phases of the different frequency components of the noise itself. NO a priori knowledge of the signal is required. More specifically, when a signal of interest and noise are completely un-correlated the performance of the present invention is maximized.
Another object according to the present invention is to provide a signal processing algorithm for suppressing noise in cases for which the strength of the signal is extremely weak relative to that of the noise. According to one aspect of the algorithm, unlike many signal processing methods (e.g., matched-field processing, beam forming, or other array processing techniques, which typically exploit the spatial structure of the acoustic field through multiple sensors to effect a gain) the algorithm according to the present invention concentrates on the noise content of a recording from a single phone or sensor whether moving or stationary. Consequently, it offers performance enhancement for work using only a single phone such as single-phone localization, as well as multiple-phone array processing.
Yet another object according to the present invention is to provide a method for extracting more information per sensor by using the physical characteristics of the noise field ascertainable from the collected data. The form of this a priori knowledge could in principle be extended to any measurable quantity, but the algorithm is most effective for those cases where the frequency spectrum of the noise field is known approximately. This is a natural choice, since frequency spectrum measurements and/or models of all types of sources (both discrete and ambient) are quite common in the literature. Furthermore, it has been shown recently, for example, that the shape of the underwater noise spectrum from wave breaking may be estimated from properties of the observable ocean surface wave field, thus offering the possibility of incorporating knowledge about the noise spectrum implicitly derived from observations unrelated to direct acoustic measurements.
The method according to the present invention enhances signal detection in cases of very-low-signal-to-noise ration (VLSNR) when the shape of the noise spectrum is known, but all relative phase information of individual frequency components is unknown. The method also has the advantage of being easy to implement.
Many contemporary signal processing strategies do not make full use of all that is known about a background noise field in order to improve SNRs. For example, no attempt is made to “recognize” the noise field by its spectral profile, and thus use the knowledge of what noise “sounds like” in o
Kuperman William A.
McDonald B. Edward
Orris Gregory J.
Karasek John J.
Malzahn David H.
Mills III John Gladstone
The United States of America as represented by the Secretary of
LandOfFree
Signal processing method for improving the signal-to-noise... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Signal processing method for improving the signal-to-noise..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Signal processing method for improving the signal-to-noise... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3361632