, any cell that contains fewer than one ping's valid samples is ignored. For example, given a cell height of 5 samples and width of 10 pings, any cell with fewer than 5 valid samples (hence more than 45 no_data samples) is ignored.The Background noise removal operator removes background noise for single beam Sv variables. The algorithm estimates the noise for each ping and subtracts it.
You can specify six settings on the Background Noise removal page of the Variable Properties dialog box. Four settings specify the size and nature of the averaging cell used to determine the noise estimate. Two Threshold settings specify the Maximum Noise and Minimum SNR for the algorithm.
The algorithm is based on concepts discussed in "A post-processing technique to estimate the signal-to-noise ratio and remove echosounder background noise." by A. De Robertis and I. Higginbottom (2007).
"A simple and effective post-processing technique to estimate echosounder background-noise levels and signal-to-noise ratios (SNRs) during active pinging is developed. Similar to other methods of noise estimation during active pinging, this method assumes that some portion of the sampled acoustic signal is dominated by background noise, with a negligible contribution from the backscattered transmit signal. If this assumption is met, the method will provide robust and accurate estimates of background noise equivalent to that measured by the receiver if the transmitter were disabled. It provides repeated noise estimates over short intervals of time without user intervention, which is beneficial in cases where background noise changes over time. In situations where background noise is dominant in a portion of the recorded signal, it is straightforward to make first-order corrections for the effects of noise and to estimate the SNR to evaluate the effects of background noise on acoustic measurements. Noise correction and signal-to-noise-based thresholds have the potential to improve inferences from acoustic measurements in lower signal-to-noise situations, such as when surveying from noisy vessels, using multifrequency techniques, surveying at longer ranges, and when working with weak acoustic targets such as invertebrates and fish lacking swimbladders."
See also: Echoview website: Background noise removal overview and templates
The Background Noise Removal algorithm assumes that some portion of the sampled acoustic signal is dominated by background noise. It estimates the background noise value for each ping and subtracts it from the ping's samples. The algorithm analyzes the echogram by averaging the sample values, with TVG removed, within averaging cells around each ping. The averaging cells have a fixed horizontal extent in pings and a fixed vertical extent in samples. The noise estimate for a ping is the minimum of its cell averages. The estimate is thresholded against Maximum Noise and the result, with appropriate TVG added, is subtracted from each sample. If the result is less than the Minimum SNR it is set to -999.
The averaging cell is specified by four settings:
The Horizontal extent defines width of the averaging cell. This value also defines the averaging ping interval around the central ping. The interval moves across the echogram one ping at a time, so that ping to ping noise removal is smooth.
The Vertical extent defines the height of the averaging cell. Averaging cells vertically divide the samples in an averaging ping interval. A smaller height is more affected by spuriously low values. A larger height is more likely to contain traces of signal.
The Vertical overlap defines the amount that successive cells overlap vertically. The overlap helps to search out regions of pure noise between samples containing signal. The larger the overlap, the larger the number of cells processed, and hence longer the processing time. If your data has few areas of pure background noise, the algorithm may have difficulty finding them, and the calculated value may exceed the maximum noise in some places. In that case you can specify a vertical overlap between cells. This increases processing time but makes the noise estimate search more thorough.
, any cell that contains fewer than one ping's valid samples is ignored. For example, given a cell height of 5 samples and width of 10 pings, any cell with fewer than 5 valid samples (hence more than 45 no_data samples) is ignored.The implementation is slightly different to that used by De Robertis and Higginbottom (2007). The points of difference are:
The background noise removal operator works very well on averaged data. Under averaged data, the mean background noise does not change but its variance decreases. De Robertis and Higginbottom (2007) argue and demonstrate this in their discussion directed to Figure 2 in their paper.
The minimum SNR threshold value required to suppress noise depends on the degree to which input data are averaged.
From left to right:
In this case for low Minimum SNR, it is clear that smoothing is preferable beacuse it removes the ‘speckly’ noise that is visible above background in the smoothed data (i.e. the method computes the noise level relative to the mean noise level, and some samples will be substantially above if unsmoothed). The smoothed and unsmoothed data become more alike as the Minimum SNR is raised.
SNR criteria are best applied at an resolution larger than a pixel. A useful approach is to smooth/average above-bottom and below-bottom data separately. Then combine the results with a Select operator after averaging so that the bottom is not combined with near-bottom data above the SNR.
A dataflow to achieve this is:
 |
|
Where:
M = The height of the averaging cell.
M is specified by Vertical extent (samples) or Vertical extent (meters) on the Background Noise Removal page of the Variable Properties dialog box.
De Robertis and Higginbottom (2007) define their averaging cell in terms of M samples and N pings.
N = The width of the averaging cell.
N is specified by Horizontal extent (pings) on the Background Noise Removal page of the Variable Properties dialog box.
De Robertis and Higginbottom (2007) define their averaging cell in terms of M samples and N pings.
k, l
= k is the index for an averaged ping interval. The noise estimate for an averaged ping interval's central ping is a result of the vertical consideration of samples within the ping interval.
l is the index for the averaged vertical intervals. The vertical consideration of samples within an averaged ping interval is achieved by examining results from averaged vertical intervals. The number of averaged vertical intervals is determined by the averaging cell height M and the specified Vertical overlap. Both types of intervals are illustrated in the Figure: Intervals and further implementation detail is in Algorithm summary: Notes and Algorithm summary: Points of difference.
De Robertis and Higginbottom (2007) define k as the index for the averaged time interval. They define l as the index for the averaged vertical intervals. Their averaging cells do not overlap (vertically) and their noise estimate for the averaged time interval was applied to all pings in the interval.
Vertical overlap = The percentage overlap of the vertical averaging cells. This is illustrated in the Figure: Intervals. Further detail about Vertical overlap is in the Algorithm summary.
This is Vertical overlap (%) on the Background Noise Removal page of the Variable Properties dialog box.
ri, , j = The uncorrected range (m) for the center of sample j in ping i.
rtvg(i, j) = The TVG corrected range (m) for the center of sample j in ping i.
The TVG correction to the range is a function of the AbsoprtionCoefficent, SoundSpeed, and TransmittedPulseLength. These quantities and the TVG range correction algorithm are taken from the ping's calibration settings.
Note: The corrected range can also be written as R(i, j).
The Background Noise Removal algorithm assumes that some portion of the sampled acoustic signal is dominated by background noise, with a negligible contribution from the backscattered transmit signal. This is expressed in the equation for Sv, meas.
By examining Sv, meas, an Sv, noise value can be estimated and a background-noise-corrected Sv can be calculated.
Sv, meas(i, j) = Mesured mean volume-backscatter strength (dB re 1m-1) of sample j in ping i.
where:
Sv, signal(i, j) is the contribution from the backscattered transmit pulse (dB re m-1) of sample j in ping i.
Sv, noise(i, j) is the contribution from the noise (dB re m-1) of sample j in ping i.
De Robertis and Higginbottom (2007)
"The primary assumptions of the method are that background noise is independent of elapsed time during one transmit-and-receive cycle, and that at some point in the measured cycle, the measurement is dominated by contributions from background noise (i.e. Sv,noise >> Sv,signal). This assumption means that noise “spikes” such as short-duration interference from the transmit signal of other echosounders are not present, or have been excluded from the data. If these assumptions are met, a portion of the return observed from an active ping (i.e. transmitter enabled) will give similar readings to those of an echosounder in passive mode, which is a measurement of background noise (Figure 1). If this assumption is violated, this and other methods (e.g. Kloser, 1996; Watkins and Brierley, 1996; Higginbottom and Pauly, 1997; Korneliussen, 2000) based on active pinging will overestimate noise because the portion of the measurement used to estimate noise levels will include appreciable backscattered signal as well as background noise."Sv, corr(i, j) = Corrected Sv (dB re 1m-1) of sample j in ping i.
The algorithm averages the samples within each cell of an averaging ping interval. The minimum result is the noise estimate for the ping at the center of the ping interval.
= Resample Powercal(i, j) by averaging the samples in the vertical intervals of an averaging ping interval.
where:
Powercal is Sv,meas with the TVG removed.
l is the index for the vertical intervals in an averaging ping interval k.
The number of vertical intervals is specified by Vertical extent and Vertical Overlap(%) and the arrangement of the intervals over the total number of samples in a ping. See also: Algorithm summary.
value of the vertical intervals l.
Maximum Noise = The maximum acceptable value for the calculated noise (dB) for the operand data. If the the calculated noise exceeds Maximum Noise, Maximum Noise is used instead.
This is Maximum Noise on the Background Noise Removal page of the Variable Properties dialog box.
Maximum Noise is equivalent to Noisemax used by De Robertis and Higginbottom (2007).
If Noise(k) > Maximum Noise then set Noise(k) = Maximum Noise.
Sv, corr(i, j) = Noise-corrected mean volume-backscatter strength (dB re 1m-1) of sample j in ping i.
where:
Sv, meas(i, j) = Measured mean volume-backscatter strength (dB re 1m-1) of sample j in ping i.
Noise (i) is the noise estimate for ping interval k and is equal to Noise(k).
Minimum SNR = The acceptable signal to noise ratio (dB) for samples in the operand data. If the signal to noise ratio of a noise-corrected sample is less than or equal to Minimum SNR, the Sv is set to -999.
This is Minimum SNR on the Background Noise Removal page of the Variable Properties dialog box.
Minimum SNR is equivalent to thresholdSNR used by De Robertis and Higginbottom (2007).
The signal to noise ratio can be expressed as SNR(i, j) = Sv, corr(i, j) - Sv, noise(i, j)
If SNR(i, j) is less than or equal to Minimum SNR then set Sv, corr(i, j)= -999.
About virtual variables
Background noise in Echoview
Background noise calculation on the Analysis page
Echoview website: Background noise removal overview and templates
