Data processing: speech signal processing – linguistics – language – Speech signal processing – For storage or transmission
Reexamination Certificate
2002-10-08
2004-08-17
Dorvil, Richemond (Department: 2654)
Data processing: speech signal processing, linguistics, language
Speech signal processing
For storage or transmission
C381S071600, C381S317000
Reexamination Certificate
active
06778955
ABSTRACT:
The present invention relates generally to a system and method for processing signals with low signal-to-noise ratios (SNRs), and particularly to physiological signals with low SNRs.
BACKGROUND OF THE INVENTION
In certain applications it is necessary to analyze physiological signals which are contaminated with noise. These signals often have low amplitudes, which results in a poor signal-to-noise ratio (SNR). A poor SNR causes difficulty in signal processing and requires complex, lengthy algorithms for processing the signals with accuracy. In some cases, not only does the physiological signal occur with poor SNRs, but also the stimuli that elicit such a physiological signal are of the same mode, or nature, as the signal. Such stimuli can affect the data acquisition process or contaminate the signal.
Such a problem is demonstrated in current methods used to test an individual's hearing. It is known that the introduction to the ear canal of an acoustic stimulus results in the production of numerous audible intermodulation distortion products. The acoustic stimulus comprises two single frequency sinusoidal tones, called primaries, at frequencies f
1
and f
2
with the levels of about 30 to 75 dB Sound Pressure Level (SPL). A normal inner ear will then produce sinusoidal mechanical responses at additional frequencies, the stronger component of which occurs at frequency 2f
1
−f
2
(the cubic Distortion Product Otoacoustic Emission, DPOAE). This energy is transferred by the middle ear back into the ear canal where it appears as an acoustic signal. The origin of DPOAE lies in the mechanical non-linearity of the cochlea due to internal active processes, associated with the motility of the outer hair cells. The phenomenon is intrinsic to the normally functioning inner ear. Thus, the presence or absence of DPOAE provides strong evidence of inner ear function (or dysfunction), making it a valuable diagnostic and screening tool.
However, detection of a DPOAE signal is difficult because its level is very low (that is the sound is very soft), and is typically between minus 15 and plus 10 dB SPL. As a consequence of background physiological, acoustic, and instrumentation noise, which is typically about 30 to 50 dB SPL, the signal-to-noise ratio is very poor.
Several solutions have been proposed thus far, two of which are described in U.S. Pat. No. 5,413,114 (which is a divisional of U.S. Pat. No. 5,267,571) and U.S. Pat. No. 5,664,577 (which is a continuation of U.S. Pat. No. 5,526,819). The contents of these references are incorporated herein by reference. U.S. Pat. No. 5,413,114 teaches a system and method for testing hearing by presenting multiple single frequency tones to an individual. The multiple frequencies are used for preventing numerous intermodulation products. However, the invention does not provide any way of reducing other noise influences. The signal-to-noise ratio, while improved, is still low and therefore many of the problems remain unchanged. U.S. Pat. No. 5,664,577 teaches a system and a method for reducing the noise levels in the system by collecting multiple readings for the intermodulation products and taking the average value. Also, two microphones are used with a differential amplifier for reducing the noise.
These and other solutions are plagued by many technical and clinical disadvantages. At present, most instruments for detection of signals in noise use signal processing methods which employ averaging in the time domain and Fast Fourier Transforms (FFT).
Because of the need to average several time segments in these methods, there is a time delay before the signal is known. This delay is even larger in the presence of artifacts. In the case of DPOAE artifacts can arise from irregular breathing, patient or operator movements, and environmental noise such as shutting of doors, sounds of equipment, steps of personnel and the like. Further, the averaged time signal contains artifacts due to the time segmentation. These artifacts are to be rejected from averaging, and therefore increase the delay. Also, the FFT data does not allow the signal to be monitored and output (or played back) in real time. The aforementioned technical factors cause clinical disadvantages, which decrease the clinical value of the present-day methods. Because the signal can not be directly output, for example, DPOAEs cannot be output to a speaker. Therefore, they cannot be detected and/or monitored by an operator. Because the signal can not be quickly analyzed when the frequencies of stimuli are varying in time, it is very time consuming to obtain a frequency response of the signal, that is, its amplitude as a monotonous function of the stimulus frequency. This can be important, for example, for DPOAEs because their amplitude varies significantly with very little change in the frequencies of primary tones.
The use of averaging techniques allows the clinician to obtain the signals only at certain times, and does not allow him/her to continually monitor the signal's level in time. In certain situations, this is critical. For example, during an operation on the acoustic nerve, DPOAE level can indicate the physiological state of the cochlea and help prevent a cochlear catastrophe caused by interruption of blood supply. Another example is in titrating ototoxic drugs, DPOAE level monitoring can help prevent drug-induced cochlear injury.
Another example of physiological signal significantly contaminated by noise is Auditory Steady State Response (ASSR). ASSR is an electric sinusoidal signal, supposedly originating in the brainstem, elicited by a modulated sinusoidal stimulus. The stimulus is typically a carrier tone of audible frequency range the amplitude or frequency of which is modulated with low modulation frequency, typically between 40 and 100 Hz. The ASSR signal has exactly the frequency of such modulation, and very low amplitude, which causes difficulty for reliably extracting it from noise.
The principle of ASSR measurement is described as follows. A modulated pure tone is presented to the ear. The carrier frequencies are usually conventional audiometric tones, from 125 to 8000 Hz. The levels of the frequencies are at or higher than 20 dB SPL. The modulation frequencies are typically 40 Hz or within the 70 to 100 Hz range (usually 80 Hz if they are in the 70 to 100 Hz range), with a 0.95 modulation index.
At the time of stimulation, a sinusoidal electric signal, which has the frequency equal to the modulation frequency of the stimulus, appears on the surface of the skull. This signal, supposedly produced by the brainstem, is called the ASSR. The ASSR can be recorded from the surface of the skull with three electrodes, typically on the vertex, on the temporal bone, and on the lobule. This electric signal, whose magnitude is typically from 40 to 400 nV, is then amplified with a typical gain of approximately 10,000 V/V. It is then passed through a band pass filter, with a typical lower frequency cutoff at 10 to 30 Hz and a higher frequency cutoff at 100 to 300 Hz. It is converted into its digital form and processed.
Techniques for ASSR detection suffer from the same drawbacks as those for DPOAE; however, a particular disadvantage of ASSRs is that their detection with current signal processing techniques requires long recording times.
It is an object of the present invention to obviate or mitigate at least some of the disadvantages discussed above.
SUMMARY OF THE INVENTION
The present invention provides a system for use in a real time system and for processing a signal with a low signal-to-noise ratio (SNR). The system comprises a model for modeling an expected signal and a filter that uses the model for filtering the signal. The filter is used for generating a prediction of the signal and an error variance matrix. The system further comprises an adaptive element for modifying the error variance matrix such that the bandwidth of the filter is widened, wherein the filter behaves like an adaptive filter.
The present invention further provides a system for processing a signal with a
Kunov Hans
Li Xinde
Sokolov Yuri
Dorvil Richemond
Fasken Martineau DuMoulin LLP
Henderson Neil
Lerner Martin
Vivosonic Inc.
LandOfFree
System and method for processing low signal-to-noise ratio... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with System and method for processing low signal-to-noise ratio..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and System and method for processing low signal-to-noise ratio... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3270564