Pulse or digital communications – Miscellaneous
Reexamination Certificate
1999-11-26
2004-05-18
Corrielus, Jean B. (Department: 2631)
Pulse or digital communications
Miscellaneous
C381S094200
Reexamination Certificate
active
06738445
ABSTRACT:
FIELD OF THE INVENTION
This invention relates to alteration of frequency components of an input signal spectrum to reduce the perceptibility of a component of the input signal. One application includes changing noise in an audio signal.
BACKGROUND OF THE INVENTION
Microphones such as those used during commercial broadcasting, motion picture filming and communications, for example, are often susceptible to unwanted sounds made by background objects such as air conditioning systems, propellers, engines, fans, computer disk drives or shutter mechanisms on motion picture cameras, for example. This can result in the unwanted sounds being converted by the microphone into an undesirable portion of the electrical signal produced by the microphone. Undesirable signals can also be introduced through a variety of other sources such as quantization noise and electromagnetic Interference, for example.
Techniques exist for reducing the undesirable portion of a signal (i.e. noise), however, such techniques often result in distortion of the desirable portion of the signal or add audible artifacts to the signal, detracting from the effect of the technique. Consequently, such techniques only produce satisfactory results when the unwanted component of the signal is well below the desired portion of the signal, or in other words, when the signal to noise ratio is relatively high.
Current noise reduction techniques include a spectral subtraction method in which a noise spectrum is subtracted from the spectrum of the input signal. In this technique the processing is done entirely on the spectral magnitude of the input signal and the phase of the input signal is left unchanged. While this technique can offer a perceived reduction of the unwanted signal, it has been found that as the signal-to-noise ratio decreases, audible artifacts tend to become more and more noticeable in the output signal. Such artifacts include musical noise, incomplete or variable cancellation of the noise which can produce modulation of the noise floor, timbral effects and/or loss of frequency components of the signal, missing sounds, loss of low level signal (speech) components, phase distortions, time aliasing, and pre-echoes and post echoes which can result in temporal smearing. For input signals having a relatively high signal-to-noise ratio these artifacts may not be too disturbing and may go unnoticed by the listener. However, at lower signal-to-noise ratios these artifacts are often more perceptually disturbing than the original unwanted signal thus negating the value of the technique.
Each of these artifacts is a result of two underlying assumptions upon which spectral subtraction is based, namely that the spectral magnitude of the noise is equal to the expected value of the noise and that the phase of the signal can be approximated by the phase of the noisy signal.
In addition, spectral subtraction is done with a constant subtraction spectrum (or adjustment frequency spectrum) which typically consists of the average (expected) value of the unwanted signal as measured over a short interval of time. This can result in overprocessing of the signal which further leads to the artifacts listed above. This can be particularly limiting to noise reduction in signals in which the noise in the input signal changes over time, especially if the noise changes in a repetitive way, such as the noise produced by a shutter mechanism of a motion picture camera. In such a case, the periods of the input signal that have a lower noise level will be processed in the same manner as the periods where the noise level is high. As a result, the lower noise periods will be overprocessed and the artifacts will be more noticeable.
Thus there is a need to alter frequency components of an input signal spectrum to reduce the perceptibility of a component of the Input signal, while minimizing artifact production, particularly where the input signal has an undesirable repetitive component.
SUMMARY OF THE INVENTION
In accordance with one aspect of the invention, the present invention addresses the above need by providing a method and apparatus for changing the frequency content of an input spectrum, which involves adjusting frequency components of the input spectrum In response to a time varying adjustment frequency spectrum to produce an output frequency spectrum including adjusted frequency components of the input spectrum.
By using a time varying adjustment frequency spectrum to adjust the frequency content of the input spectrum, optimum adjustment frequency spectra can be used during corresponding periods of the input signal such that only the required amount of signal processing is done to adjust the frequency components of the input signal, for each segment of the input signal. This is particularly useful where the undesired component is a repetitive noise, such as sounds caused by a shutter mechanism of a motion picture camera, for example. The noise spectrum for the period during which the shutter is actuated is different than the noise spectrum for the period during which the shutter is inactive.
During periods when the shutter is actuated, the noise component may have a relatively high loudness level and may have a relatively strong high frequency content. When the shutter is inactive, the; noise spectrum may merely consist of noise from a motor in the motion picture camera, for example, and may therefore have a greatly reduced loudness level and may have a greater low frequency content By selectively using one or the other of these spectra for adjustment of the spectrum of the input signal during respective corresponding periods of the input signal, only frequency components of the input signal which correspond to those of the noise during that segment are affected, resulting in processing only frequency components in the input signal which coincide with noise frequency components in existence during that period. Thus, effectively a time varying adjustment spectrum is generated and used to adjust the input spectrum. The number of spectra used to produce such a time varying adjustment spectrum for any one signal will depend upon the number of segments, or frames, into which the input signal is divided. There is no limit on this number.
In addition, the method and apparatus involve dividing the frequency components of the input spectrum into a plurality of frequency bands and applying a separate time varying adjustment spectrum to each band. This allows the adjustment process to be matched to the time-frequency distribution of the noise, such that frequencies of the input signal in some bands may be adjusted more greatly than frequencies in other bands, while the time varying adjustment frequency spectrum provides the best adjustment spectrum for the instant period of the input signal. In other words, the processing is divided in time and frequency so as to match the processing to the particular characteristics of the noise. This allows the noise reduction processing to track the cyclical or repetitive characteristics of the noise. This inherently reduces all formns of audible artifacts since the average amount of processing applied to the input signal is minimized. At time intervals and frequencies where the noise is loudest, more aggressive processing is employed. Elsewhere, less aggressive processing is applied. In general, this allows processing to occur with the best signal to noise ratios in each frequency band, resulting in fewer artifacts being produced.
The method and apparatus may also include a perceptual model. The purpose of the perceptual model is to determine which portions of the unwanted signal are being masked by the desired signal and which are not being masked. Masking is the phenomenon that occurs in the human auditory system by which a signal which would otherwise be audible is rendered inaudible by the presence of another signal. By including a perceptual model in the processing, only the audible portion of the unwanted signal is removed, and thus the overall adjustment of frequencies applied to the input signal is furth
Corrielus Jean B.
IVL Technologies Ltd.
LandOfFree
Method and apparatus for changing the frequency content of... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Method and apparatus for changing the frequency content of..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Method and apparatus for changing the frequency content of... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3213077