Untitled record

full description

Sister's wallpaper

The invention relates generally to the field of signal processing. More specifically, the invention relates to methods, systems, and program code for noise filtering and retrieving spectral components that are attenuated in signals.

<p>5 Signals may be transmitted in the form of a sound wave (acoustic signals), for example, generated by an acoustic wave source through a variety of materials including reservoir and non-reservoir shale, well tubes including drill pipe, and drilling equipment Others including the drill bit Audio signals generally lose their accuracy due to background noise during transmission and recording Background noise consists of two parts, an internal part generated by the measuring system, 10 and an external part emanating from the surroundings.</p>

Audio signals can also be distorted during transmission and recording due to signal attenuation, especially high frequency components. The amplitude spectrum of an audio signal is generally non-uniform. The higher the frequency of the spectral components of the audio signals, the greater the attenuation of the spectral components corresponding to the audio signals.

<p>15th Background noise and irregular attenuation are combined to break up the quality of the audio signals. The frequency component plot of a sample of an audio signal supplied by a microphone is shown in comparison to the frequency component plot of a corresponding sample of a reference signal provided by the accelerometer, recorded at the same time. The frequency components of the sample presented by the accelerometer represent an unattenuated version of the frequency components of a sample of the audio signal; That is, what they should be if not for attenuation. It can be seen here</p>

<p>20 That the high-frequency components of the presented sample are attenuated by the audio signal to the noise level.</p>

٦٥٧٥

-٣-

To increase the quality of the signals, the attenuated signals should be filtered to remove the noise and their attenuated spectral components should be recovered. There are two common methods: frequency filtering and amplitude filtering. Frequency filtering is to remove from a signal some unwanted frequency components using an electronic device or calculation. In this method, any frequency components with a frequency greater and/or lower than the previously selected 5 cut-off values ​​are removed or severely attenuated.

When using a arithmetic operation, the signals in the time domain (for example, graphically shown as signal amplitude over time) are converted into the frequency domain to represent the signals in the amplitude spectrum. This is achieved, for example, through the use of a fast Fourier transform Fast Fourier

Transformation . (FFT).

<p>10 An example shows a pair of audio signals, which are in the time domain, and are converted to the frequency domain. By converting the signal to the frequency band, components of the signal in the amplitude spectrum that have a frequency above and/or below a cut-off value are removed.</p>

The amplitude filtering process is usually a calculation in which components in the amplitude spectrum are removed with an amplitude value above and/or below the cut-off (threshold value). If required, a reverse FFT is then performed on the filtered frequency band signal. To extract the time domain output signal.

In these two methods, the cutoff (b) values ​​are of vital importance. However, this is not always the case, so that there are clear cutoffs that can be used to separate the sound signals from the noise.

One example shows a restored signal (shown by solid line) where the cut-off threshold value of 20 is less than necessary, which results in over-filtration.

One example shows a restored signal (solid line) where the amplitude cut-off is too high, which results in excessive residual and amplified noise in the restored signal.

٦٥٧٥

-٤-

Some relatively complex noise filtering techniques using the "spectral subtraction" method have been proposed, for example in IEEE Trans. No. 27 1997,

Suppression of Acoustic Noise in Speech Using : pp. 113-120 under the title

SF Boll, Spectral Subtraction; US Patent No. 0255560/2007 A1,

<p>5 under the title “Low Complexity Noise Reduction Method”. In this type of method, the noise signals are filtered by subtracting the slope of the spectral noise. In the first example, spectral noise is calculated during the activity without speaking. In the second example, spectral noise is estimated from the “Noisy Activity Detector” procedure. However, this type of method can be difficult to apply to situations where noise characteristics are not</p>

<p>10 known, such as, for example, those associated with drilling operations, to include drilling operations that involve the direction of the drill bit in real time.</p>

To further increase the accuracy of the audio signals, the attenuated spectral components should be restored. Discuss the innocence

“American Method for Restoring Spectral No. 0143604/2012, Components in Denoised Speech Signals,”

<p>15th way to do it. However, in this method, it requires training the undistorted grammar obtained from a pure speech signal with full bandwidth. For this reason, these requirements limit the application of their method to sequences of events in which the pure signal is available for full bandwidth, except for the application of their method for those events in which full bandwidth cannot be obtained. Describes US Patent 0122596/2004, entitled “Method for High Frequency.”</p>

<p>20 ”, Restoration of Seismic Data is a method in which the attenuation of high-frequency components is estimated from</p>

Reflected acoustic signals at successive depth levels of the formation boundary. An inverse operator is then defined from the attenuation for each depth level. Specific inverters are applied to the reflected audio signals to restore their attenuated high frequency components. However, this method requires knowledge of the manner in which the high frequency components are attenuated.

٦٥٧٥

-٥-

Each of the above-mentioned methods and techniques has its own advantages and field of application. However, the inventor realizes, that there are many situations in which sound signals cannot be separated from associated noise of some frequency or cut-offs of constant amplitude, pure signal or noise samples, and where a high frequency component attenuation pattern cannot be obtained.

<p>5 As noted above, audio signals can be attenuated during transmission and recording. Under the influence of various conditions, some or all of the high-frequency components of the signals can be attenuated to a level similar to the background noise. For example, the original acoustic signal (sound) from an underwater device is generated by the underlying background noise that varies with time, and is distorted as a result of the attenuation of its high frequency components during transmission through the water.</p>

<p>10 When recorded from a long distance from the source, the recorded sound will have inherent noise and the sound will be significantly distorted due to the attenuated high frequency components.</p>

The inventor realizes that situations are the same when sound signals are recorded from a source some distance in the air or from underground. Based on the foregoing, the inventor realized that common defining characteristics of these situations include: (1) that background noise may not be constant, and (2) that components of

<p>15th The high frequency is generally very high at the time the signal reaches the recording devices. Analogously, the inventor realized that there was a need for systems, computer programs, and computer aided methods to perform both filtering of the unsteady noise, and then recovery of the attenuated high-frequency components of the filtered signals sufficiently to provide a filtered and recovered signal, It closely matches the original raw signal.</p>

<p>20 US Patent No. 071156/2005 relates to methods and systems for processing the audio signal. More specifically, it relates to methods and systems that enhance audio signals and to systems that include these methods and systems.</p>

British Patent No. 2426167 relates to signal processing, specifically a method for noise estimation.

٦٥٧٥

-٦-

General description of the invention

In light of the above, various models of the invention uniquely provide methods, systems, and program code for filtering noise and recovering the spectral components that have been attenuated in the signals. Various embodiments of the invention provide, as a result, the ability to filter and retrieve audio signals sufficient to provide a signal of sufficient quality.

5 Allows "listening" for the drill bit. According to various models, the drill bit sound can also be used to derive real-time petrophysical properties while drilling, and/or allow real-time bit orientation.

Recorded audio signals include background noise and its high frequency components are attenuated. Various embodiments of the invention characteristically provide improved methods for filtering out background noise and recovering

10 High frequency components attenuate the signals, thus recovering more information from the signals. In addition to the above, distinctively various models can be applied to seismic data operations to improve the quality of seismic signals, among other uses.

15

20

25

More specifically, an embodiment of the noise filtering method and the recovery of the spectral components that have been attenuated in the signals may include steps for the reception of audio signals for a preselected time period in order to form one or more records of the audio signals (as typical in a time range) and/or perform one or more of the following steps for each one but more typically a set of audio signal records each record separately for a short period of time. The steps may alternatively include sampling the audio signals within the relevant record, for example, by processing so that numerical data is thus formed from the sample containing a set of raw data samples, for example, if no So already. The steps can also include using a Fast Fourier Transform system to transform a set of unprocessed data samples into a set of unprocessed Fast Fourier Transformation (FFT) data samples. The unprocessed FFT data samples are formed from the sampled selected background noise and audio signal data.

٦٥٧٥

-٧-

The steps of the method may also include dynamically filtering each set of unprocessed FFT data samples to remove or attenuate hidden noise specific to the sample within it and thus produce a corresponding set from each of the clean FFT data samples. Sample selected background noise is removed or attenuated by a special selected dynamic filter

5 record-specific dynamic filter to produce corresponding FFT data samples. The dynamically modulated filter is at least partially determined by the cut-off portion of the dynamic abundance noise selected from each set of unprocessed FFT data samples. The dynamic amplitude noise threshold is determined by a value chosen from a value selected from the percentage base noise and a given value selected from a limit variable 10. Clean FFT data samples can include audio signal data that has significantly attenuated high-frequency components.

The steps of the method can include recovering the high-frequency attenuated components from clean data samples and thus producing clean and recovered data samples that are in the frequency range. The restore step can be performed by implementations made from a registry-specific recovery wizard that has been predefined at least 15 with a portion of clean data samples and a Gain function

Function. The steps can also include a transformation so that the cleaned and restored data samples are transformed into the restored and cleaned data samples in time range data.

The steps of the method may also or alternatively include a primary recorder dynamic filter at the raw dynamic abundance fraction determined at least in part by 20 percent of the primary recorder specified base noise and a specified value by the raw recorder

from a Threshold Parameter variable in order to form a dynamic filter that has been fine-tuned (and selected) to perform the previous filtering step. The dynamic pre-filter can include specifying the percentage of basic noise identified by a recorder as the percentage of Kth within a specified frequency range of the register from The abundance range of each group of samples of the relevant register 25 below which each frequency component in the frequency range of the relevant abundance spectrum of each group is

٦٥٧٥

-٨-

In the extent of the log that is processed in the form of background noise sample with greatly confirmation. “Noise floor” is the background noise in a signal or a level of noise that is fed through the system below where the signal is caught and which cannot be isolated from the noise.

5 The alignment step may include setting a specific value that is initially recorded for a variable of the defined threshold value in the form of either a generic threshold value attenuated with the percentage of base noise for the first register or an augmentation method for the threshold value that can be added at a percentage of The basic noise of the first register in order to determine a value for a noise cutoff, the dynamic amplitude that has been

10 selected, which can be applied separately to each other from each set of unprocessed data samples.

The matching step includes steps for receiving or reviewing the sub-value from a set of samples for each relevant record of one or more records. The relevant record is recorded and the alignment step may not include repetition of the subset of samples of processed data

15th Recorded at substantially different times with different background noise levels thus defining a set of representative FFT data samples and thus defining a set of representative FFT data samples. If the relevant record frequently is a direct record that can be processed then all the unprocessed data samples cannot be selected at significantly different times The alignment step involves receiving a subset of the data samples at the start of a related record and thus determining

20 A set of representative FFT data samples.

Regardless of the matching step, that step may include a specified frequency range for a related record from one or more records. A specific frequency range can be determined by a range of common frequencies common to each sample from a set of samples of FFT data with frequency components that can be identified by background noise or if not

25 . A range of frequencies is predominant through background noise or a range of frequencies is common with

٦٥٧٥

-٩-

All samples from a solid-state FFT data set having a higher percentage or background noise substantially greater than the range values ​​with respect to frequencies at or near the upper end of the range for a representative FFT data set.

. The alignment step may also include the selection of the percentage of basic noise in relation to

5 the relevant record from one or more records. This selection step could include: Determining an apparent partitioned abundance under which at least about all of the frequency components fall within a particular selected frequency range which is sample-specific background noise within a set of relevant FFT data samples and selecting an initial value from a log-specific boundary value variable with respect to a log with Relevancy and determination of the cut-off noise portion of the dynamic abundance of a relevant log 10 is determined by selecting a percentage of the base noise and selecting a particular log-specific value from the threshold value variable. The alignment step could include evaluating the results of a first dynamic filter that is determined at least in part by a noise portion that is dynamically cut from one or more valleys of samples within two representative data samples drawn from a set of unprocessed data samples and thus creating a filter 15 dynamic alignment.

An initial dynamic filter evaluation step on one or more samples within a set of related FFT data samples can include data evaluation of the abundance position of a portion of the dynamic abundance noise from one or more samples within a set of data samples of the related FFT and/ OR Evaluate the results of a raw dynamic filter which is initially determined by at least 20 segment segments of the raw dynamic abundance noise on one or more samples within a set of relevant FFT annotation samples. This step can include defining a first dynamic filter and performing an initial dynamic filtering from one or more samples within a set of FFT annotation data, thus producing one or more corresponding FFT data samples and directly graphically examining one or more FFT clean data samples. By comparing 25 each clean related FFT data sample with the unprocessed FFT data sample.

٦٥٧٥

-١٠-

The step of evaluating the results of a preliminary dynamic filter on one or more samples within a set of related FFT data samples may alternatively include selecting a preliminary dynamic filter, and performing the first dynamic filtering through one or more samples within a set of relevant FFT data. Relevance and thus produce one or more FFT clean data samples and examine 5 one or more speech data samples corresponding to one or more FFT clean data samples.

This step can include performing a reverse FFT on one or more clean FFT data samples and thus converting the clean FFT data to a time band format and thus producing one or more time band data samples and producing sounds corresponding to one or more band data samples using a method to listen.

10 If the results of the first dynamic filter are not acceptable, the steps of the method may include restoring the tuning steps and the Threshold factor and thus offsetting the part

A cut-off from noise of dynamic abundance in the relevant direction and an evaluation of the results of replacing a primary dynamic filter until it is acceptable. If the results of the evaluation of the first dynamic candidate are acceptable, the steps of the method may include evaluation of a first dynamic candidate on a second set

15th from the FFT illustrative data samples.

20

If clean FFT data samples are stored such that a subset of a set of clean FFT data samples can be selected at substantially different time intervals, the steps of the method may also or alternatively include performing the iteration step of a subset of Clean FFT data samples that are samples of signals recorded at substantially different times with different baseline noise levels and thus define a set of illustrative clean FFT data samples used for particular job correlation or selection Acquisition and formation of a recovery wizard for the registrar. If the clean FFT data sampling is a directly appropriate step such that a subset of the FFT data sampling set may not be selected at different time intervals, the method steps may also or alternatively include

25 Perform a procedure for receiving a subset of clean FFT data samples at the start of the recorder

٦٥٧٥

-١١-

relevant and thus define a set of clean FFT data samples used to build or choose the acquisition function and form a recovery processor.

According to an example from the model for steps, described above, the restore trigger for the registry is a recovery trigger for the trial log. According to one embodiment, 5 method steps can be selected from a primary retrieval operator of a relevant record from one or more records. The step can include selecting a set of illustrative clean FFT data samples from a set of illustrative clean FFT data samples and building or selecting an acquisition function or choosing an acquisition function from a set of responsive database of illustrative clean FFT data samples and tuning variants of the acquisition function thus forming a recovery wizard and perform 10 primary recovery processing of one or more samples within a set of illustrative clean FFT data samples through a primary recovery processor defined in part by an acquisition function and thus producing One or more samples within a set of recovered FFT data samples and an initial recovery processor evaluation.

If the RWW results are not acceptable, the steps of the method can include iterating 15 steps of the build or picking out a new acquisition function and tuning variables of the Acquisition and Replacement function and evaluating the RWP results of the replacement until those results are acceptable. If the evaluation results for the primary recovery processing are acceptable, the steps of the method can include evaluating a primary recovery processor on a second subset of a set of clean FFT data samples. The initial recovery processing assessment step can include a graphical comparison

<p>20 For each sample of a set of retrieved FFT data samples with the corresponding clean FFT data sample, and/or one or more time-domain data samples corresponding to one or more samples of the restored FFT data samples. This method can include performing a reverse FFT on one or more restored FFT data samples and thus converting the restored FFT data to a band format and thus creating one or more timeband data samples</p>

٦٥٧٥

-١٢-

one or more and the production of sounds corresponding to one or more band data samples using a listening device.

Usefully, one or more embodiments of the present invention may include a system for filtering noise and recovering attenuated spectral components in audio signals, which are configured to perform specific operations.

<p>5 by one or more combinations of one or more steps of the computer executable method, described above. The system may include a dynamic noise filtering and signal recovery system comprising one or more processors and memory in communication with one or more processors; Dynamic noise filtering and signal recovery software stored in the device memory to filter out dynamic noise and signal recovery to provide filtering noise and component recovery</p>

<p>10 Attenuated spectral or noise filtering and recovery of attenuated spectral components in audio signals, including instructions given when performing dynamic noise filtering and signal regeneration, causing the computer to perform the operations defined by the executable computer method steps, described above.</p>

It is additionally useful, that one or more renders also include a noise codec

<p>15th The dynamic and dynamic noise recovery program and the signal restoration program for noise recovery, recovery of attenuated spectral components, or both filtering noise and recovery of attenuated spectral components in audio signals, and computer software are transmitted in a temporary delay, or stored on non-deletable media via computer to media and includes a set of . An instruction that, when executed by one or more processors, causes one or more processors to perform operations specified by</p>

<p>20 One or more combinations of one or more steps of the method described above.</p>

It is also useful for one or more embodiments to include media that cannot be deleted by computer

Contains addressable code that is non-transitory computer readable medium

in which to provide filtering noise, recover attenuated spectral components, or both filter noise and recover attenuated spectral components in audio signals, the readable processor and embed code

<p>25 on a set of instructions, which, when executed by one or more processors, cause</p>

٦٥٧٥

-١٣-

The execution of one or more processors defined by the combination of one or more steps of one or more methods, described above.

Interestingly, according to one or more models and in contrast to traditional filtering techniques, Dynamic Amplitude Noise Cutoff techniques allow a fraction of the 5 best noises to be evaluated and then applied to each individual sample. As a result, one or more of the models provided here provide better solutions in terms of background noise and/or recovery of attenuated components of audio signals. One or more models have been used in real-world projects with immediate experimental uses. Additionally, one or more of the models can be sub-used and can be used in seismic scanning to retrieve high-attenuated frequency signals10 and thus be used to increase the variance of seismic scanning types.

Brief explanation of the drawings

In order to understand in more detail the manner in which the features and advantages of an invention, as well as other features and advantages, become apparent, a more specific description of the invention briefly summarized above can be obtained by cross-referencing its embodiments which are illustrated in the attached figures, which constitute Part 15 of this specification. However, it should be noted that the figures only show various examples of

Invention and for this reason it should not be considered as limiting the scope of the invention as it may include other effective models as well.

Figure 1a is a graph that provides an example comparison between an audio signal recorded by a microphone and a vibration signal recorded by an accelerometer for the same audio sample to illustrate the attenuation of the audio signal20.

Figure 1b represents a graph showing the overfiltration of high frequency components.

Figure 1c represents a graph showing the lower filtering of the high frequency components.

٦٥٧٥

-١٤-

Figure 1D represents a graph showing a comparison of the capacitance value of a conventional fixed boundary value and capacitance values ​​of a dynamic boundary value according to one of the models of the invention.

shapes. 1e-1f represents a pair of graphs showing the results of signal filtering and recovery of high frequency components using the amplitude values ​​of the dynamic limit value according to one of the models of the invention.

5 Figure 2 represents a framework flow diagram showing the basic system components of a system to provide dynamic noise filtering and recover a spectral component that has been attenuated according to one embodiment of the invention.

shapes. 3a-3c represent a set of graphs illustrating the amplitude spectrum of a sound sample recorded with an accelerometer and with a microphone.

shapes. 4a-4d represents a set of graphs that show the values ​​of the amplitude spectrum of two samples to illustrate

<p>10 The background noise level varies over time.</p>

shapes. 5a-5c represent a set of graphs that show the amplitude spectrum values ​​from a sample to illustrate the correct selection of the dynamic amplitude noise limit for use in noise filtering according to one of the models of the invention.

Figure 6 is a schematic diagram of a high-level flow illustrating the steps for filtering noise

<p>15th Background and recovery of the attenuated high-frequency components of the audio signals using “Dynamic Amplitude Noise Cutoff” filtering technology according to an embodiment of the present invention.</p>

Figure 7 represents a schematic flow diagram illustrating the steps for forming fast instant transform data to be applied to the dynamic filter according to one of the embodiments of the invention.

<p>20 Figure 8 represents a high level flow schematic diagram illustrating a process for dynamic filter synthesis according to one of the embodiments of the invention.</p>

٦٥٧٥

-١٥-

Figure 9 represents a flow schematic diagram showing the examination of purified FFT data, or purified and recovered FFT data according to one of the embodiments of the invention.

shapes. 10a-10b represent a pair of graphs showing the derived dynamic amplitude noise limit values ​​for microphone and accelerometer records, respectively, according to one embodiment of the invention, 5 compared to a fixed noise limit line.

Figure 11 represents a schematic flow diagram showing the steps for recovering the attenuated high-frequency components of an audio audio signal and for defining, selecting and tuning the Gain Function according to one embodiment of the invention.

10

15

20

Figure 12 represents a graph showing a typical gain function used to recover the attenuated high-frequency components of an audio signal according to one embodiment of the invention.

shapes. 13A-13E represent a set of graphs showing raw data for a sample microphone and FFT accelerometer and the recorded and/or filtered results of a pair of samples recorded by a microphone and accelerometer, respectively, during an identical audio time frame, according to an embodiment of the invention.

shapes. 14a-14e represents a set of graphs illustrating raw data for a sample microphone and accelerometer FFT and the recorded and/or filtered results of a pair of samples recorded by a microphone and accelerometer, respectively, on an identical audio time frame, according to one of the embodiments of the invention.

shapes. 15A-15D represents a set of graphs that illustrate a comparison between the results processed using a typical dynamic amplitude noise limit process described in the current application, according to one of the models from the invention, and a traditional two-sample fixed amplitude noise limit methodology.

shapes. 16A-16B represents a set of graphs that show the raw data that consists of several samples and the candidate result, respectively, for a part of the accelerometer record, according to one of the models of the invention.

shapes. 17a and 17b represent a set of graphs illustrating the raw data that comprises

٦٥٧٥

-١٦-

From several samples and the filtered and recovered results, respectively, of a portion of the microphone register, according to one of the embodiments of the invention.

Detailed description:

The present invention will now be described in more detail later with reference to the attached figures, which are

<p>5 Show examples of the invention. However, this invention can be embodied in many different forms and should not be considered as limiting the embodiments described in the present application. Instead, such models are presented so that this disclosure is comprehensive and complete, and fully conveys the scope of the invention to those experienced in the field. The identical numbers refer to similar elements throughout the patent. The main disclaimer, if used, refers to similar elements in alternative forms.</p>

<p>10 Disclaimer: Two terms, “record” and “sample,” are clarified for their specific meaning in this specification. A record (for example, for audio signals) represents a set of data recorded or otherwise aggregated for a specified period of time, from the same source in the same medium. The record can be digitized into serial segments of data over a period of time running within the bounds of the period temporal, yi</p>

So that each slice represents a small part of the log. A slice of data is called a sample (or frame).

<p>15th frame). For this reason, the digital record consists of a series of samples. Additionally, the frequency band representation of an audio signal is called the “amplitude spectrum” or simply the “frame”.</p>

spectrum” of the signal. Each sine wave line of the spectrum is called the component of the total signal in the sample.

When recording audio signals, there are always noise values ​​within the recorded signals. Signals recorded during transmission and recording can also be deconstructed by non-uniform attenuation of the frequency components

<p>20 high. The signal in the sound wave image will lose its accuracy due to the accompanying background noise and attenuated high frequency components during transmission and recording. Noise filtering can directly improve signal quality. In general, filtration is an essential step to recover attenuated high frequency components. A number of noise removal methods are known. Conventional methods typically first convert audio signals from a time-domain format to a frequency-band format, sample by sample, and attempt to filter or reduce noise,</p>

٦٥٧٥

-١٧-

And then you try to restore the components that were attenuated. For noise filtering, conventional methods typically first distinguish/estimate the noise, and then reduce the noise using the specified noise, either by subtraction, filtering, or suppression. Various methods include using a fixed amplitude limit for a selected register, a fixed frequency limit for a selected register, or in special cases, pure noise data such as, for example, pauses between talking during a mobile phone conversation.

10

15

20

25

As shown in Figure 1D two audio signals are shown, one problem could be that audio signals may not contain pure noise frames and may not be filtered background noise using a fixed amplitude or frequency limit. Another problem is that noise estimation is often inaccurate, especially when the noise has a time variance. As a result, maintaining a constant threshold value of 25 according to conventional methods results in either excessive signal removal (see Figure 1b) or some residual noise after noise removal, which can be over-amplified during recovery (see Figure 1c).

As shown in Figs 1a, 1e-1f, various embodiments of the invention can provide both signal filtering and recovery of high frequency components (shown as covered solid lines vs. unattenuated accelerometer signal shown as dashed line). Over embodiments of the invention, background noise is filtered by a “dynamic threshold value” problem, specific, or otherwise specified by a process according to one or more embodiments of the invention. Using this process, a specific noise amplitude limit is evaluated for each individual sample for a given record and then applied to the sample to filter out the sample's background noise. The attenuated spectral components of the samples are then recovered from the filtered or purified samples.

According to various embodiments of the invention, all extreme values ​​on the amplitude spectrum can be treated as portions of the signal and large, featureless portions on the amplitude spectrum as background noise. For example, the portion surrounded by dashed rectangles on Figures 3a and 3c is processed as background noise. As noted above, background noise is typically time-varying, ie changing from frame to frame on the FFT spectra. Based on the foregoing, various models of the invention treat background noise

٦٥٧٥

-١٨-

As temporal variation, that is, background noise changing from frame to frame is processed. Within one frame, that is, within the FFT spectrum, however, the background noise is treated as a constant value. That is, the background noise of all data points (in the whole frequency range) in the FFT spectrum is taken as a constant. Various embodiments of the invention provide an assessment of

5 of the limit value of the dynamic amplitude (cutoff) for each frame signal, that is, for each sample FFT, based on

Its own background noise attributes, for a particular record. The record is then filtered frame by frame

Using the resident boundary of the frame. Characteristically, this can provide an assessment of the capacity limit of a frame and beyond

It is applied to the same framework.

10

15

20

Figure 2 shows an example of a system 30 to provide dynamic noise filtering and restore an attenuated spectral component. The system could include 30 computers for dynamic noise filtering and signal retrieval 31 comprising one or more processors 33, memory 35 associated with the processors 33 for storing software and/or database records in it, and optionally a user interface 37 that could include a graphic display 39 for displaying graphic images 41, and a means of user input, as is known to those experienced in the field, to provide access for the user to manipulate the software and database records. Please note, that the computer 31 can be in the form of a separate unit, built with a tool with a well, or a personal computer, or in the form of a server or multiple servers serving many user interfaces located remotely 37. Based on the above, an interface can be User 37 either directly connected to the computer 31 or through a network 38 as is known to those who are skilled in the field. The system may also include 30 one or more databases 43 stored in memory (internal or external) that are operationally computer-linked for dynamic noise filtering and signal retrieval31, as perceived by those in the field. One or more may include From the databases 43 sets of sound wave files

Recorded, for example, during drilling operations to provide real-time characterization of rocks during drilling.

٦٥٧٥

-١٩-

System 30 may also include a Signal Recovery and Dynamic Noise Filtering computer program 51 that is provided stand-alone or stored in memory 35 in a Signal Recovery and Dynamic Noise Filtering computer program 31. The Signal Recovery and Dynamic Noise Filtering computer program may include instructions that lead When executed by a processor or computer such as, for example

<p>5 For example, the Signal Retrieval Computer and Dynamic Noise Filtering 31, the computer performs operations to perform dynamic noise filtering and retrieve attenuated spectral components in each of the multiple samples of multiple audio signal records or files . Note that the computer program for signal retrieval and dynamic noise filtering 51 can be in the form of microcode, programs, routines, and languages ​​using symbols that provide specific groups or sets of</p>

<p>10 The organized processes that control and direct the operation of devices, as is known and understood by those skilled in the art. Note also that the computer program for signal recovery and dynamic noise filtering 51, according to one or more embodiments of the present invention, does not have to be entirely in a temporary memory, but can be selectively loaded, as necessary, according to various methods as</p>

Known and understandable to those skilled in the field.

<p>15th The system may also have signal interfaces 53 connected through a cable 54 to a data acquisition unit (DAU) 55 data acquisition unit (DAU), which are connected to a computer 31. According to this modular design, the signal interface 53 includes amplifiers or some other form A device for capturing or recording audio signals, including accelerometers and a floor speaker, through which an audio signal (acoustic wave) can be recorded. The data acquisition unit 20 55 receives the analog audio signal from the signal interface 53, samples/converts to digital image and stores the converted audio signal to digital image in the database 43.</p>

25

Figures 3a-17b provide graphs generated from an actual example that are used to better understand representative models of the invention. To provide typical graphs for discussion, a machine-generated sound (not shown) was recorded by a measurement loudspeaker and accelerometer (not shown) for a period of more than 71 hours to obtain an atypical loudspeaker record and an atypical accelerometer record.

٦٥٧٥

-٢٠-

10

15

20

25

Built-in amplifier, amplifier, and accelerometer. They are attached to a metal adapter that is attached to the machine. The recorded audio signals are first amplified by the built-in amplifier and then sent to 55 DAU, where they are sampled and converted into a digital image. Signals from both sensors were sampled at the same time series. The converted digital image data is sent to computer 31 and saved in database 43 for analysis. Data samples were in time scale format. Each of them was converted to frequency band format, i.e. amplitude spectrum format by applying Fast Fourier Transformation (FFT). Because both records were sampled from the two recorders at the same time sequence, each sound had two corresponding samples stored in the corresponding records. For clarity, the letter A, for the accelerometer, and M, for the loudspeaker, are added as suffixes on the sample label. For example, sample 1A and sample 1M represent the recorded pair of the same sound recorded by the accelerometer and loudspeaker, respectively. For convenience, the letters "A" and "M" are added as suffixes to any of the labels corresponding to the accelerometer and loudspeaker register, respectively. Note that the example used in the present disclosure is only intended to better explain the principle. In implementation, one or more of the models of the invention can be applied to other situations. Similarly, different embodiments of the invention are not limited to the types of sensors (i.e. loudspeaker and accelerometer) used in this example, but other types of acoustic sensors can also be used.

Figures 3a-3c are graphs showing the amplitude spectrum of an audio signal sample recorded by an accelerometer (Figure 3a) and a sample of an audio signal (sound) recorded by a loudspeaker (Figures 3b-3c). By the accelerometer the name "sample A1", the audio sample recorded by the loudspeaker is referred to as "sample 1M". A loudspeaker produces an audio signal by measuring the pressure charge in the air, hence, the unit of amplitude is pascals (Pa), while an accelerometer records the audio signal by measuring the acceleration of vibration, and thus, the unit of amplitude is the acceleration of gravity (g).

There is background noise in the recorded audio. A portion of the background noise is shown framed at 1003 in the A110 amplitude spectrum of sample 1A and is framed in 1007 in the 111M amplitude spectrum

٦٥٧٥

-٢١-

For the 1m sample. Background noise is generated implicitly by the audio signal recording system (eg loudspeaker, cable, etc.) and from the surrounding environment. In fact, there is always background noise in the recorded audio.

By comparing the 110A and 110M amplitude spectrum (see, for example, Figure 1A for a trajectory comparison), 5 it can be seen that the spectrum patterns recorded by the accelerometer and amplifier are identical for frequencies below 1200 Hz. However, the 110M amplitude spectrum frequency components greater than 1200 Hertz recorded by an amplifier greatly diminishes Amplitude attenuation increases with increasing frequency As such, the quality of the 110M audio signal recorded by an amplifier is not only degraded by background noise, but also deteriorates significantly by attenuating its higher frequency components.

10 To increase the quality of the audio signal, the signal should be filtered to remove background noise, and the attenuated high-frequency components should be recovered as much as possible. The background noise should be removed first and then the attenuated high frequency components restored using the amplitude spectrum filtered or otherwise cleaned. Otherwise, the high-frequency components are recovered without removing the background noise, and the background noise is generally amplified in the recovered portion of the signal.

15th For the purpose of illustration, as shown in Figures 3a-3c, audio samples from the accelerometer and amplifier records are provided to represent a raw noise signal that can be filtered to remove noise. The 1M audio sample from the loudspeaker register is used as an example of the attenuated, noisy raw signal sample and its high-frequency components to be retrieved; and the acoustic sample from the accelerometer, whose high-frequency components have not been damped, as a reference sample 1A to check the result

20 Retrieval of the 1M loudspeaker sample.

According to a representative model, there are two main steps to the solution for noise filtering and recovery of the attenuated high-frequency components of the audio signal sample. First, log samples are filtered using a 'Dynamic Threshold'. The "Dynamic Threshold Value" is the "Dynamic Amplitude Noise Cutoff" and is evaluated from a sample then

٦٥٧٥

-٢٢-

Apply it to the same sample. Second, the attenuated high-frequency components of the samples being cleaned or filtered are recovered.

Referring to the 1M amplifier sample in Figure 3b, it appears that there are no signals above 2200 Hz on the 110M amplitude spectrum. Referring to the accelerometer sample 1A in Fig. 3a, the . spectrum is shown

5 Amplitude 110A however there are four clear peaks: peak 1001, peak 1002 and two peaks before peak 1001. When the amplitude scale of 110M is changed to logarithmic scale 111M (see Figure 3C), the corresponding four peaks are more clearly depicted on the amplitude spectrum of 111M provided by the amplifier.Of these peaks, 1005 and 1006 correspond to the 1001 and 1002 of the 110A spectrum, respectively.Comparing the 110A and 111M spectrum, it is clear

10 Also, the peaks of the 111M capacitance spectrum correspond almost exactly to the peaks of the 110A capacitance with respect to

its frequencies, and that the amplitude of the 110M's high-frequency components has been greatly diminished, and that the attenuation has increased with frequency. To avoid noise being amplified during recovery, the recorded raw data should be filtered to remove background noise. After filtration, the process continues to recover the recoverable attenuated spectral components. After restoration, the amplitude spectrum of the amplifier should be 110M

15th Similar to nice accelerometer capacitance 110A.

Various models of the invention are designed to deal with situations where there are no preset patterns for clean signals or noise. In these situations, the signal cannot easily be distinguished from noise by applying clean signal patterns or noise according to traditional signal configuration systems.

20

According to a representative model, all clear peaks on the amplitude spectrum are treated as parts of the signal and the large uncharacterized portion on the amplitude spectrum is treated as background noise. For example, also referring to Figures 3a-3c, the 1001 and 1002 peaks of the spectrum are treated.

110A and 1005 and 1006 peaks of the 111M spectrum as part of the signals; While the portion surrounded by rectangles is dealt with in 1003 of the 110A spectrum and 1007 of

Spectrum 111M as background noise.

٦٥٧٥

-٢٣-

Also, under each signal data point in the corresponding 1M, 1A full sample bandwidth, there are background noise contribution arrows in the capacitance. The amount of contribution is treated similarly, ie, for the maximum spectrum amplitude level which is in the featureless portion at 1003 on the 110A spectrum, and 1007 on the 111M spectrum.

5 To remove background noise, appropriate noise separation is required, eg 1004 on

The amplitude spectrum 110A (Fig. 3a) and 1008 on the amplitude spectrum 111M (Fig. 3c), to separate the signal from the background noise. Once the noise has been properly separated, the background noise can then be filtered by subtracting the amplitude separation from the raw amplitude spectrum, as defined by equation ( 1( :

𝐴_𝑖 =𝐴𝑟_𝑖-𝑁, 𝑖𝑓 𝐴𝑟_𝑖 &gt;𝑁

)1( 𝐴_𝑖 =0, 𝑖𝑓 𝐴𝑟_𝑖 ≤𝑁 10

where Afi is the amplitude of a data point, i , the amplitude spectrum of a sample after filtering;

where Ari is the amplitude of the data point, i , on a raw amplitude spectrum before filtering; And

where Nc is the noise amplitude separation.

15th When filtering the raw data, equation (1) is applied to the entire relevant frequency band of the sample.

For example, for 1M sample data recorded by the amplifier, the spectral components are attenuated to the same level as the background noise after approximately at least 4000 Hz. So the frequency range in question is between zero and 4000 hertz. . From this discussion, someone who is normally skilled in the art should realize that adequate noise separation is important in the application of

20 the scheme shown above, and that adequate noise separation should result in maximum noise removal as well as maximum signal preservation.

Figures 4a-4d provide amplitude spectrum plots of the two samples, sample 1M and sample 2M, recorded at two different times to illustrate that the background noise level varies with time. be the amplitude spectrum

٦٥٧٥

-٢٤-

111M in Fig. 4b is the same as the 110M amplitude spectrum in Fig. 4a, but the axis of the amplitude is on a logarithmic scale. The 1M sample in this diagram is identical to the sample in Figures 3b-3c. However, this set of figures shows comparatively that background noise is not constant, but can instead be variable over time. The background noise level is 1007 (Fig

5 4b for sample 1M significantly different from that for background noise 2003 (Fig. 4d) for sample

.2M

It can be seen from this comparative illustration that applying a constant noise separation to these two samples should lead to erroneous results. For example, if a constant 2000 noise separation is applied (running across Figs 4b and 4d), the two signal peaks, 1005 and 1006, are removed from the 111M amplitude spectrum

10 For the 1M sample because the amplitude of each is less than the constant class 2000, the background noise is not removed

2003 for the 221M spectrum for sample 2M because the 2003 background noise amplitudes are above the 2000 constant class.

The illustration shows that applying constant amplitude noise separation to filtering can remove some signal components and eliminate some background noise. Ideally, a specific sample 15 noise separation should be chosen, such as the 2001 sample 1M chapter (Fig. 4b) and the 2002 sample 2M chapter (Fig. 4d), to optimally separate the signal from the background. In summary, a fixed noise separation should not be applied to situations in which noise is variable with time. As such, in accordance with the representative design, a more ideal method is provided to assess the separation of a specific sample and to apply the separation to the sample in question.

20 The good noise limit is the limit that is derived from a sample and applied to the same sample. A representative model of the invention provides such a methodology. Referring to Figures 5a-5c, sample 3a, recorded by an accelerometer, provides one example to explain this principle. The spectrum diagram 311a (Fig. 5b) shows the approximated amplitude of the spectrum 310a (Fig. 5a). Spectrum 312a (Fig. 5c) represents the candidate result for spectrum 310a after highlighting the methodology described in this representative body.

25 310a and 311a, each point represents a data point.

٦٥٧٥

-٢٥-

As shown in Figure 5b, in a frequency range of, say 3000 - 5000 Hz (in 3001) of the amplitude spectrum graph 311a, we can be confident that, for a given register, there is a percentage Kth number that can be at a value lower than Definitely treat data points, or components of all samples within the log, as background noise. For example, % . represents

5 The fiftieth 50th, at 3002 for the 311a spectrum, is such a percentage of the amplitude. This percentage in the current application is called "Base Noise Percentile"

However, defining the percentage of noise will not guarantee that all data points are above signals. For example, the 50th percentile, at 3002, for the frequency range 3000–5000 Hz for graph 311a in Figure 5b represents the percentage of noise for the register. For sample 3a, the data points are represented between the 3002 noise percentage and the 3003 line of the amplitude spectrum

311a also has background noise, although it is higher than the 3002 noise percent.

15

20

For a given record, there is only one percentage of noise by definition. When the percentage of noise is specified for a record, any percentage below the specified percentage of basic noise is the percentage of basic noise. For example, because the 50th percentile 3002 of Chart 311a of Figure 5b represents the percentage of noise, the 40th percentile also represents the percentage of noise, simply because all of the data points

The lowest of them will be less than the 50th percentile.

The Noise Percentage cannot be used directly as the noise threshold for a particular record because it is very possible that there may be some noise data points higher than that which cannot be removed after the record is filtered. Since it is below the noise percentage, all data points are treated as noise and there are still noise data points higher than the noise percentage, the correct noise amplitude limit must be higher than the noise percentage.

The representative model of the invention presents the aforementioned correct amplitude limit, which is called "dynamic threshold value", or "dynamic amplitude noise limit".

٦٥٧٥

-٢٦-

“Amplitude Noise Cutoff” This noise amplitude limit is dynamic because it is evaluated for each single sample in the range of a record and applied to the same single sample. As a result, it can ideally separate noise from signals; that is, it removes noise to the maximum and preserves The signals are at maximum during filtering.

<p>5 Since for a given record, the dynamic amplitude noise limit is higher than the noise percentage,</p>

The following equation (Equation (2)) is formed to define such a boundary value:

(2) 𝐴𝑡ℎ =𝐶𝑡ℎ∙𝑃^

where Ath is the dynamic amplitude noise limit, such that the unit is the same as the amplitude of the amplitude spectrum. Line 3003 on the amplitude spectrum 311a of Fig. 5b represents the mentioned term.

<p>10 where Pb is the percentage of noise for a given record, such that the unit is the same as the amplitude of the amplitude spectrum. Line 3002 on the amplitude spectrum 311a of Fig. 5b represents the percentage of noise for sample 3a. He will easily understand the definition of percentage and the percentage evaluation of those who are experienced in the field.</p>

where Cth is a constant coefficient, called the boundary value factor. It is a unitless constant for a given record.

<p>15th The frequency range from which the percentage of noise is derived is called the “specific frequency range</p>

Specific Frequency Range for a given register, the specified frequency range is the same for all samples within the register. For example, the frequency range 3000 - 4000 Hz is chosen as the specified frequency range of the microphone register, and the frequency range 3000 - 5000 Hz is chosen as the specified frequency range of the accelerometer register in This example.

<p>20 The percentage of noise Pb is also the same for all samples within a given record in this embodiment of the invention. For example, the 50th percentile is chosen as the base percentage of noise for both the microphone register and the accelerometer register for this example. The 50th percentile for both records was chosen because it provides a sufficient reference percentage for both records. and with</p>

٦٥٧٥

-٢٧-

However, a different percentage can be used, as a percentage of noise for the two records. It should be noted, that although the percentage of noise is the same for all samples in a given log, the actual amplitude value of each sample is evaluated to be the same as the percentage of the sample and, therefore, will usually be different from that of every other sample in the log.

<p>5 The threshold value factor, Cth, is constant for a given log and, therefore, is the same for all samples in the given log.</p>

Rooted in its definition in equation (2), the dynamic amplitude noise threshold, Ath, has the following property: It uses the noise information for an entire log, namely the threshold value factor, Cth, and the same “Specific Frequency Range” for the entire log, and the same noise percentage 10 baseline for the entire record, and initialized according to each sample using the capacitance value specified for the percentage

for noise, Pb, for the sample, at the corresponding percentage of noise.

When the background noise varies, the noise percentage value follows the variation in the background noise. The threshold value factor, Cth, makes the dynamic amplitude noise threshold higher than the background noise and lower than the signal.

<p>15th As a result, the dynamic amplitude noise limit follows the variation in background noise and at least to a large extent, if not completely, maximally separates the background noise from the signals.</p>

It turns out that the following alternative definition of the dynamic amplitude noise limit is as effective as that defined in equation (2) for separating the background noise from the signals:

(3) 𝐴𝑡ℎ = 𝑃^ + 𝐶^

<p>20 where, Ce is a constant coefficient, called the highest bound value, such that the unit is the same as the amplitude of the amplitude spectrum. It is fixed for a specific record. Its function, which is the same as the limit value factor, Cth, makes the dynamic amplitude noise limit higher than the background noise and lower than the signals.</p>

٦٥٧٥

-٢٨-

Thus, it separates at least to a large extent, if not completely to a maximum of background noise from the signals.

Using the dynamic amplitude noise limit, a maximum of background noise can be removed and the signals can be saved at a maximum using equation (1). When using equation (1) the noise limit 5, Nc is replaced by the dynamic amplitude noise limit, Ath, to form equation (4) :

𝐴_𝑖 = 𝐴_𝑟𝑖 - 𝐴_𝑡ℎ, 𝑖𝑓 𝐴_𝑟𝑖 &gt; 𝐴_𝑡ℎ

)4( 𝐴_𝑖 =0, 𝑖𝑓 𝐴𝑟_𝑖 ≤𝐴𝑡ℎ

The procedure for filtering and restoring a record. Figure 6 represents a high-level flow diagram showing the steps for background noise filtering and recovery of the attenuated high-frequency components of audio signals using the Dynamic Amplitude Noise Cutoff filtering technology, according to a representative model.

When automatic 410 audio signals are received, they are converted into frequency band data (420 FFT data) by a 500 pre-processor. The FFT data, when plotted, is called the amplitude spectrum. The values ​​for the M110 amplitude spectrum are presented in Figure 4a, M220 in Fig. 4c 15 and 310a in Fig. 5a are examples of plotted FFT data.

The FFT data is passed through the dynamic filter 640 to filter out the background noise, thus, the FFT data is filtered 430.

The purified FFT data 430 is processed with a recovery processor 930 to recover the attenuated high-frequency components from the register, and in this way, the Cleaned &amp; Restored 20 440 FFT data is produced.

The purified and restored FFT 440 data can be used directly in 470 user applications. The purified and restored FFT data 440, which is in frequency band format, can also be inverted by applying an Inverse Fast Fourier Transformation 450 to transform the data

٦٥٧٥

-٢٩-

Purified and Restored FFT 440 to Purified and Restored FFT 460, which can be used directly in user applications 471, ie is retrieved by an audio device.

5 The described filtering and recovery procedure can be applied to audio data of both recorded records and live records of audio signals in real time as understood by those with ordinary skill in the field.

As shown in Figure 7, in a typical configuration, the 500 preprocessor in Figure 6, used to produce FFT data from audio signals, includes two main steps. First, 410 audio signals, in digital format, are sampled and converted to digital using a 55 data acquisition unit (DAU) data acquisition unit to 520 digital data according to this analog model. Second, 10 520 digital data, which is In time scale format, by fast Fourier transform

530 Fast Fourier Transformation to 420 FFT data, which is in the frequency band. The above procedure for producing FFT data from audio signals is well understood by those experienced in the field. The data monitor is also known to those skilled in the field as a digitized data converter.

15th The center of the dynamic filter 640 (Fig. 6) is represented by equations (2), (3) and (4). By applying equations (2) and (4) or (3) and (4) to each sample one by one in a register, the Remove background noise from the record from the entire record.

For a given record, before the 420 FFT data can be filtered by the dynamic filter 640, the dynamic filter should be tuned to optionally separate the background noise from the signals.

20 For “tuning” the Dynamic Filter means setting an integer percentage as the base noise percentage Pb, adjusting the Threshold Cth factor, Factor, or raising the Ce Threshold Elevator of Equation (2) or (3). Because only one is used. From equations (2) and (3) in the filter, and the procedure for adjusting the value factor

٦٥٧٥

-٣٠-

The boundary, Cth, is the highest boundary value Ce symmetrical. Based on this, for the sake of brevity, only one variable, which is the limit value factor, Cth, was chosen to clarify the synthesis procedure.

Figure 8 presents a high-level flow diagram describing a representative process of dynamic filter synthesis 640. At the beginning of the process, representative FFT data 421 are used in dynamic filter synthesis.

5 There are two main event sequences in the selection of representative FFT data 421. First, if the log is recorded, the FFT data recorded at different times with different background noise levels is used as the representative FFT data 421. Second, if an online record is to be processed and cannot be selected Its FFT data At different times, some FFT data at the beginning of the record is used as the representative FFT data 421. In both event sequences, 10 representative FFT data is picked from the record to be processed / being processed.

In the next step 610 the “Specific Frequency Range” is determined As described above, the specified frequency range is a frequency range within which the baseline noise percentage of all samples can be easily determined using the given register. For example, in the frequency range 3000 – 5000 Hz, (at 3001) the amplitude spectrum 311a of Fig. 5b, we can ensure that, below the 50th percentile (at 3002), all data points are background noise.

20

As the example from Figure 5b shows, for a given record, it will be easier to determine the percentage of noise within a frequency range dominated by background noise. For this reason, if there is a frequency range that is dominated by background noise for a given record, it should be chosen as the specified frequency range. Otherwise, the frequency range with the highest portion of the background noise data points is chosen as the specified frequency range. The selected frequency range should be wide enough, to ensure that the noise percentage value is constant.

Analogously, representative FFT data samples 421 are verified to find a wide frequency range dominated by background noise as a specific frequency range. If there is no such frequency range mentioned, then a wide frequency range with the highest portion of the background noise data points is chosen as the selected frequency range.

٦٥٧٥

-٣١-

In step 620, the Noise Percentage is selected. As defined above, Base Noise Percentile is the percentage at the bottom of which data points in the specified frequency range on the amplitude spectrum are processed, with confidence, as noise for all samples within the log. To optimally separate background noise from the signals, the “noise percentage” should be high. However, choosing a higher “percentage noise,” than necessary, can increase the likelihood of signals with lower amplitude values ​​being treated as background noise. That is, a value that is too high can result in excessive filtration.

As introduced before, the threshold value factor is used to increase the percentage of noise to a threshold value for noise of a higher level (see, for example, Figure 5b). Based on the above, it turns out that it is indistinguishable to take the risk of choosing a high noise percentage over However, it is also indistinguishable to choose an excessively low noise percentage as this will increase the potential for lower filtering.

Briefly, in step 620, the determination of “percent noise” involves selecting an apparent dividing line under which all data points can easily be considered within the “Specific 15 Frequency Range” set to be background noise for all samples within a representative 421 FFT data. For example , the 50th percentile 3002 in Figure 5b can easily and ostensibly be considered a good candidate for "percentage noise".

20

When “Percent Noise” is selected, eg 50%, the “Basic Noise Percentage Percentage” value is evaluated within the “Specific Frequency Range” set for each sample in the range of representative FFT data 421. The method for evaluating the percentage value is well understood and known Good for those experienced in the field. Then, for each sample in a representative FFT data frame 421, the data points in the given “specific frequency range” frame are compared against the evaluated value of the sample’s “percent noise” to see if all data points are below the value of “percent noise” representing Processing noise data, and if most of the noise data points are below the value of

٦٥٧٥

-٣٢-

Noise percentage. If it is, then the selected "percent noise" is accepted as the correct data.

If, for some samples, the data points are below the “percent noise” value, unprocessed noise data, but signal data, and when the “percent noise” 5 is too high; It should be reduced, for example, from 50% to 45%. Or if, for some samples, most of the background noise data considered is not below the “percent noise” value, the “basic noise percentage” is too low a percentage and should be increased. Note also, it is allowed if some noise data points are higher than the “percent noise” value when “basic noise percentage” is selected 10 percent is correct, because the signal data points will be separated from the noise data points by the “dynamic amplitude noise limit” Dynamic Amplitude Noise Cutoff", which is higher than the "Percentage Noise" value.

In step 630, the Cth Threshold Factor is set. If equation (3) is used, then the Ce Threshold Elevator is set or distinguished

15th In another way. Since the procedure for differentiating the two variables is identical, only one variable, Cth, is chosen to clarify the procedure.

The initial value of the boundary value factor, Cth, is chosen. With certainty, the dynamic amplitude noise limit corresponding to a given sample can be evaluated from its noise percentage and initial limit value factor. This particular sample can be filtered using equation (4).

20 The performance of this dynamic raw filter 640, defined by a combination of equations (2) and (4) or equations (3) and (4), is then checked. The dynamic raw filter 640 can be checked or otherwise checked directly, in step 655, By testing the dynamic filter 640 with each sample in Yiyi

Representative FFT data range 421 using a graph such as, for example, histogram

٦٥٧٥

-٣٣-

For the audio spectrum 311a from Figure 5b, for visual examination to see if the dynamic amplitude noise limit 3003 is set to optimally separate the background noise from the signals.

Additional or additionally, the dynamic raw filter 640 can be checked by filtering each sample in the representative FFT data 421 using the dynamic raw filter 640 to produce the FFT data

5 Purified 650. The purified FFT 650 data is then examined in step 700.

Figure 9 presents a high-level flowchart describing screening step 700 according to a representative model. The purified FFT data is 710 in Fig. 9 and the purified FFT data 650 is in Fig. 8. The purified FFT data 650 is either examined directly in step 720 by comparing each of its samples against the corresponding data from the representative FFT raw data 421, and/or is converted to data

10 Time scale 732 by the Inverse Fast Fourier 731 Transformation. The 732 time range data can be played back by a 733 bug, and 734 is then checked in step 735.

Referring again to Figure 8, after examination 700 and/or examination 655, a decision is made in step 660 to infer whether or not the dynamic pre-filter is acceptable. If 15,660 is not acceptable then the dynamic filter needs to be further tuned by setting the value of the threshold value factor inversely in

Step 630. Then the steps are repeated until step 660.

If 660 is acceptable, the initial dynamic filter is tested in step 670 with a new small set of representative FFT data. The procedure for “DF test on new FFT data” 670 is identical to the procedure for checking using representative FFT data 421. It is achieved by following the steps

20 from 645 to 660, but on a new set of representative FFT data.

If the test is not acceptable 680, then we need to tune dynamic filter 640 further by repeating the procedure from step 620. If 680 is acceptable, then dynamic filter 640 is tuned and can be easily applied to filter the log.

٦٥٧٥

-٣٤-

For the exemplary test described in the present application, there are over 51,400 samples in the exemplary microphone record and in the exemplary accelerometer log. Among the 51,400 samples, thirty samples recorded at different time were selected as the represented FFT data. Among the representative FFT data, it was determined that 3000–4000 Hz

5 The correct specified frequency range for the microphone register was as shown, for example, by the two samples in the figures. 4a-4d, and a value of 3000-5000 Hz for the accelerometer record shown, for example, by sample 3a in Figures 5a-5c. We can easily observe from the corresponding figures that for all 30 samples, all data points in the given frequency range were below the 50th percentile of background noise for both the microphone and accelerometer data. That's why

10 Reason: The 50th percentile is set as the percentage of base noise for both the microphone and accelerometer records. Following steps 630 onwards in Figure 8, it turns out that 1.4 and 1.3 represent the best value of the threshold value factor, Cth of the microphone and accelerometer registers, respectively. Now equation (2) is fixed for the sample records, that is, the dynamic filter is tuned to each one of the sample records.

15th Lines 2001 and 2002 in Figures 4b and 4d, respectively, characterize the dynamic amplitude noise limit computed using Equation (2) that is fixed for sample M1 and M2, respectively; and line 3003 in Figure 5b characterize the dynamic amplitude noise limit for sample 3a.

Figures 10a-10b show the dynamic amplitude noise limit derived using equation (2) synthesized for a portion of the typical microphone and accelerometer registers, M800 for the microphone and A800 20 for the accelerometer, respectively. The figures also show the dynamic amplitude noise limit that varies with time. If a fixed noise limit is used, such as the 8001 vertical line for the microphone record, and the 8002 vertical line for the accelerometer record, any sample to the left of the fixed limit will be over-filtered, that is, either the signals are removed or quenched; Any sample to the right of the constant bound is without filtering, ie, the noise will not be filtered to a maximum or otherwise optimally.

٦٥٧٥

-٣٥-

Accordingly, these representative schematics show that the use of a fixed capacity limit generally results in poor filtration quality. Based on this, a fixed value should not be used as a noise limit. If used, it will be equivalent to assuming that the background noise amplitude is the same for all samples within a given register. However, this hypothesis, although made in most cases, is not a hypothesis.

5 mast.

Additionally, the percentage, for example, the 50th percentile, should not be used, on its own, as well as a noise boundary to separate noise from the data. If used, it can be equivalent to assuming that within the specified frequency range the ratio of noise data points is the same for all samples within a given register. That is, it can be equivalent to assuming that all samples are within

<p>10 A particular record has the same percentage of error as the error data points. This hypothesis is also not considered one of the valid hypotheses.</p>

According to the representative model, we can safely and easily find the percentage of the “basic noise percentage” in which all the lowest data points are noise. Then the best interval between the noise and the signal data points is above the “Noise Percentage”. Then the 'value factor'

<p>15th Threshold which is tuned by making the Dynamic Amplitude Noise Threshold the best interval between the noise and the signal data points. Considering that each sample is evaluated within a given register, the value of the “percent noise” from the sample data, i.e. evaluated for the sample, is applied to Same sample through Dynamic Amplitude Noise Limit, this and other models optimally separate background noise from the signals.</p>

<p>20 As described above, during transmission and recording, high frequency components can be attenuated</p>

The audio signals have more of the lower frequency components. That is, the attenuation is a function of frequency. The flow diagram from Figure 11 includes a 930 recovery processor used to restore attenuated signals. The Recovery Wizard 930 has the following two equations:

)5( 𝐴_{𝑟_𝑖} = _𝑖 ∙ 𝐴_{_𝑖}

٦٥٧٥

-٣٦-

)6( 𝑖=𝑓(^𝑖), 𝑖≥ 1

where Afr_i is the capacity of the data point i after filtering and recovery;

where Af_i is the amplitude of the data point, i, of a sample, after filtering;

where G_i, without units, is the gain applied to the data point i; And

5 where F_i is the frequency at data point i.

10

Equation (6) represents a comprehensive picture of the relationship between gain and frequency, and it is called the “Gain Function” to recover the attenuated amplitude, and the attenuated amplitude is amplified using Equation (5) to restore the attenuated amplitude using an integer increment. Considering that the attenuation is Dependent on frequency, as shown in Equation (6), the gain is dependent on frequency. Because the attenuation depends on many factors, such as the medium in which the sound wave is transmitted, the recording medium, the recording device, among other factors, there will be different forms of Equation (6) for different sequences of events. For this reason, a general, rather than specific, picture is presented. From Equation (6) in this example. However, when running, a specific image suitable for the specific situation should be selected or selected, such as, for example, the example shown in Figure 12.

<p>15th For a given register, the value of the dynamic amplitude noise threshold varies from sample to sample, but is constant for a given sample, that is, according to equations (2) or (3), it does not vary with frequency for the given sample. However, the value of the gain varies, with frequency, However, it does not depend on samples i.e. for part of or for the whole record, and the gain function is constant. When using the gain function, equation (6) is considered constant for a given record, and the 930 recovery processor can be used to recover the attenuated signals.</p>

<p>20 Record, equation (5) is applied to each sample one by one in succession until all samples are recovered in the record.</p>

For a given register, before the 930 Recovery Wizard can be applied, the gain function (Equation (6)) is set or selected and optimized. Figure 11 is a high-level flow graph

٦٥٧٥

-٣٧-

Describes a step-by-step procedure for identifying/distinguishing or selecting, and then tuning the amplification function. Some representative 431 purified FFT data is selected and used in the synthesis procedure. There are two sequences of events in picking representative FFT-purified 431. First, if a record is recorded, the FFT-purified data recorded at different times with different levels of background noise is used as the representative FFT-purified data 431. Second, in the case where an online record is being processed by far in

Real time, therefore, its FFT data cannot be picked easily at quite different times, some of the FFT-purified data recorded at the beginning of the record is used as the representative FFT-purified data 431. Note that, we should be aware that the representative FFT-purified data is being picked from the same 10 The record to be processed regardless of whether the record was previously stored or is currently being received and processed.

The next step 910 is to build an integer "amplification function" or pick out a preconfigured function from the 990 database. For example, the 100M graph in Figure 12 represents a typical gain function that was found to be satisfactory when applied to the microphone register for this example. In the frequency range, 1200 - 2800 Hz (at 1010), for this particular function, the gain is a power function of frequency.

<p>15th Like most functions, there can be some variables in the gain function. Step 920 prompts the tuning of the gain function variables. When these variables are initially tuned, a raw recovery processor 930 consists of equation (5) and a gain function (Equation 6). Then, each sample is processed into the representative 431 purified FFT data using a raw recovery processor 930 to produce the recovered FFT data. 940.</p>

<p>20 The 940-restored FFT data in step 700 is then scanned. The scan 700 is detailed in Figure 9, but by replacing the purified FFT data 710 in Figure 9 with the 940-restored FFT data in Figure 11. Either the restored 940 FFT data is examined directly in Step 720 by comparing each One of its samples versus the corresponding from representative 431 purified FFT data, and/or converted to 732 time-domain data by an Inverse Fast Fourier Transform</p>

٦٥٧٥

-٣٨-

731 Transformation. The time range data of 732 seconds can be displayed by the 733 bug, and the 734 sound can be checked in step 735.

After the scan, an adjustment is made in step 950 to conclude whether the 930 Recovery Wizard is acceptable or not. If it is not acceptable, the procedure from step 910 is repeated. If otherwise it is considered acceptable 5, the Recovery Wizard 930 is tested, in step 960, with some new purified FFT data.

Note that, in step 960, the procedure for step “Test recovery processing on data” is

corresponding to "est Restoring Processor on New Cleaned FFT Data

To a large extent to perform the scan using representative purified FFT data 431; realized in the tenancy steps 930 to 700.

<p>10 After the check, an adjustment is made in step 970 to conclude whether the 930 Recovery Wizard is still considered acceptable. If it is not acceptable, the procedure is repeated starting in step 910. If it is acceptable, then the 930 Recovery Wizard is reported or otherwise specified and tuned, and can be applied to process the entire registry.</p>

After the 930 Recovery Wizard has been tested and accepted, if the gain function is newly created (step 980), it is stored (step 985) in the gain function 990 database for future use.

Test using real data for typical records. A representative model of the invention has been applied to the exemplary records, previously defined, to test the principles and methods described in the present application. Given that the signals recorded by the accelerometer are not attenuated, and the signals recorded by the microphone are attenuated, the signals from the microphone were compared against the corresponding signals from the

<p>20 Acceleration Defines the amount of actual attenuation. Only the accelerometer record was filtered and the microphone record was filtered out first and then the attenuation was restored.</p>

For testing purposes, 30 of the 51,400 samples in the microphone register were used to synthesize the dynamic filter and to create the amplification function (Equation 6). An example describing the tuning of the dynamic filter was discussed earlier. The amplification function was successfully created using the procedure defined in

٦٥٧٥

-٣٩-

Figure 11. The built-in amplification function is shown in Figure 12. Using the dynamic filter tuner 640 and the recovery processor 930, the methodology was applied according to one of the representative models of the invention for processing records. The results are shown in the figures. 13a-17b for some typical samples.

Display shapes. 13A-13E FFT Unprocessed Data and Filtered Results

<p>5 and retrieved for samples 1M and 1A recorded by the microphone and accelerometer, respectively, during the corresponding audio time frame. Untreated sample data FFT by spectrometers 110M and 110A are indicated in Figs. 13a and 13b, respectively. The 111M spectrum (Fig. 13c) is the same spectrum as the 110M, but with a logarithmic vertical axis.</p>

axis. Lines 2001 (Fig. 13c) and 1102 (Fig. 13b) are a noise boundary

<p>10 The rated dynamic capacity of the spectrometers is 110M and 110A, respectively. The spectrum is 112M (Fig. .).</p>

13D) is the processed result of 110M after filtration and recovery. Spectrum 111A (Fig. 13E) is the processed result of 110M after filtration. The comparison shows 110M (Fig. 13A) and 112M (Fig. 13D), 110A (Fig. 13B) and 111A (Fig. 13e) Efficient and optimal removal of background noise after filtering. The amplitude spectrum of 112M is approximately similar to that of

<p>15th 111A, meaning that the background noise of only the 110M is not removed efficiently and optimally,</p>

But the attenuated high frequency components were also properly restored.

Figures 4A-14E show the untreated sample FFT data and the filtered and extracted results for samples 2M and 2A recorded by microphone and accelerometer, respectively, in the corresponding audio time frame. The 220M and 220A spectrometers show the unprocessed 20 FFT sample data. The 221M spectrum is similar to the 220M spectrum, but with a logarithmic R axis. indicate

Lines 2002 and 1202 to the rated dynamic amplitude noise limit of 220M and 220A, at

straight. Spectrum 222M is the processed result of 220M after filtering and recovery. Spectrum 221A is the processed result of 220A after filtering. The comparison between 220A and 221A shows the efficient and optimal removal of background noise after filtering. As previously discussed, record

25 The microphone used in the example, the amplitude spectrum components recorded by the microphone are attenuated by frequency

٦٥٧٥

5

-٤٠-

Greater than 3500 Hz decreases to the same level of background noise. For this reason, only components with a frequency of less than 3500 Hz can be recovered. A comparison of the spectrometers 222M (Fig. 14d) and 221A (Fig. 14e) shows that the amplitude spectrum of the 222M is almost the same as that of 221A before 3500 Hz. This means that, not only has the background noise of the 220M been eliminated efficiently and optimally, but it has been Also, properly recoverable attenuated high frequency components are recovered.

Figures 15a-15d illustrate the comparison between the results processed using the typical dynamic noise amplitude cutoff process described in the current application, and the traditional fixed noise amplitude cutoff value methodology for the two samples. The 112M amplitude spectrum (Fig. 15a) is the filtered and recovered spectrum 10 for sample 1M in Fig. 13a, processed using a typical dynamic amplitude noise reduction process.

The 113M spectrum (Fig. 15c) is the filtered and retrieved spectrum for the 1M sample using the constant noise cut-off value process. The amplitude spectrum 222M (Fig. 15b) is the filtered and retrieved spectrum for the 2M sample in Fig. 14a, and processed using the model process. The 223M spectrum (Fig. 15d) is the spectrum Filter and recover for the 2M sample using the constant noise cut-off process The value of 15 constant noise cut-offs for the two samples is the cut-off value of 2000 in Figs 4b and 4d.

As shown in 1301 by the resulting amplitude spectrum 113M in Fig. 15c, since the max values ​​1005 and 1006 (Fig. 4b) on the graph of the amplitude spectrum 111M for sample 1M are less than the constant 2000 (Figs. 4b, 4d), these two are eliminated The two chi values ​​when the static limit is applied Another peak is also suppressed before the two chimes offset within the dashed rectangle 20 in 1301 seriously when compared to the 112M spectrum (Fig. 15a) processed using a typical dynamic amplitude noise limit process.

In the 221M spectrum (Fig. 4D), the background noise 2003 is higher than the constant noise cut-off value of 2000 (Fig. 4D). For this reason, the background noise will not be filtered efficiently, that is, it is not filtered. As a result, the unfiltered noise was amplified during the Restore opinion, according to what was done

٦٥٧٥

-٤١-

It is illustrated by the maximum values ​​of hyper-recoverable and unfiltered and amplified noise 1302 in the 223M spectrum (Fig. 15d).

Figures 16a-16b display unprocessed data 140a (several samples) and candidate result 141a, respectively, for a portion of the accelerometer log. This part of the log covers the time duration of log 5 over three hours and consists of 2,300 samples. The vertical axis is the sample time. , the horizontal axis is the sample frequencies The amplitude value of each frequency is represented by the chromatogram The amplitude spectrum of each sample, represented by 310a for the unprocessed sample data and 312a for the filtered result, is plotted in a narrow horizontal frequency band.

The graphs of the amplitude spectrum 140a and 141a are the result of all samples plotted together 10 respectively along the time axis. That is, graphs 140a and 141a are the amplitude spectrum for a group of samples, horizontal lines 310a and 312a are the amplitude spectrum for individual samples. A comparison of the unprocessed amplitude spectrum 140a and that of filter 141a shows that background noise is removed from the unprocessed data 140a efficiently and optimally and that the graph of the filtered amplitude spectrum 141a is much purer.

15th Figures 17a and 17b display the unprocessed data 150M and the filtered and recovered result 151M, respectively, for a portion of the microphone register using a representative embodiment of the invention. The log duration is the same as the log duration in Figures 16a and 16b. That is, both Figures 16A and 17A are records of the same sound samples, but they are recorded by different devices. The high frequency components of the microphone register are greatly attenuated. Show comparison of spectrum chart

20 140A in Figure 16a and the 150M spectrum diagram in Figure 17a that the higher components

The mostly (&lt;1500 Hz) frequency of the microphone register is attenuated too low to be recognized. After filtering and recovery, these highly attenuated high frequency components were, however, well restored, as shown in the 151M spectrum diagram (Fig. 17b). The filtered and recovered spectrum for the 151M microphone register (Fig. 17b) is roughly the same as that for the 151M microphone register.

٦٥٧٥

-٤٢-

Accelerometer 141A (Fig. 16B). This proves that the disclosed invention works very efficiently and satisfactorily.

In summary, the examples shown in Figures 13a-17b show that foundations, processes and procedures, according to one or more representative embodiments of the invention, are capable of efficiently and optimally filtering out background noise, and restoring the attenuated high-frequency components to nearly their true values.

This application is a non-provisional application based on precedence and benefiting from the provisional US application, serial number 877,117/61, filed on September 12, 2013, titled "Dynamic Threshold".

Methods, Systems, and Program Code for Filtering Noise and Restoring, included in Attenuated High-Frequency Components of Acoustic Signals

10 The current application for reference in its entirety.

In the figures and the specification, a preferred typical form of the invention is revealed, and although specific expressions are used, the expressions are used in a descriptive meaning only and not for restrictive purposes. The invention has been described in great detail with specific reference to these illustrative models. 15 It will be clear, however, that various modifications and changes can be made within the spirit and scope of the invention as described in the previous specification.

٦٥٧٥

-٤٣-