dB difference Z score technique
| Loading and viewing > Calibration > Background noise-removal > Detection and filtering > Classification > Characterization |
The Z-score technique described by De Robertis, McKelvey, and Ressler (2010)1 is a dB-difference classification method.
Method overview
De Robertis, McKelvey, and Ressler (2010)1 describe a multifrequency classification method based on the normal deviate (Z-score).
Abstract: We evaluated the feasibility of identifying major acoustic scatters in North Pacific ecosystems based on empirical measurements of relative frequency response. Acoustic measurements in areas where trawl catches were dominated by single taxa indicated that it might be possible to discern among key groups of scatterers such as fish with gas-filled swimbladders, euphausiids, myctophids, and jellyfish. To establish if walleye pollock (Theragra chalcogramma), a key species in the ecosystem, can be separated reliably from other groups under prevailing conditions, we developed a method based on the normal deviate (Z-score) to identify backscatter consistent with the pollock relative frequency response. We evaluated the performance of the method by comparing it with the traditional method of species identification (i.e., directed trawl catches and subjective interpretation of echograms) during five large-scale acoustic surveys of the eastern Bering Sea. Pollock abundance estimates employing the multifrequency method were highly correlated with those using the traditional method, which indicates that the multifrequency method performs well in this situation. In this environment, multifrequency methods will allow more inferences to be drawn when direct sampling of organisms is limited and will also complement existing abundance surveys by improving species classification and providing information about key non-target species.
This page shows how the Z-score classification method relates to the example dataflow demonstrated in Echoview’s Introduction to the Formula operator tutorial.
Normal deviate equations
The normal deviate equations for the Z-score technique are:
The Formula operator is used to calculate a normal deviate analysis for four frequencies (n = 6 pairwise differences), where i is the frequency pair and j is the backscatter (species) class. Estimated values for μ and σ are used.
The Formula operator can be used to implement these equations for Pollock and krill frequency responses.
Using the Zi,j values for Pollock, tabulated values from Table 2. Summary of ΔSv statistics measured by taxon in De Robertis et al. (2010)1 are applied.
The Pollock Z-score expression is:
Where the input variables are:
- V1 is the 18 kHz resampled variable.
- V2 is the 38 kHz resampled variable.
- V3 is the 120 kHz resampled variable.
- V4 is the 200 kHz resampled variable.
Species frequency response
Dataflows
Stages of computation
 |
(Left) The numbered dataflow stages indicate grouped operators. The following table provides more information about the stages related to Pollock analysis.
*The Background noise removal operator estimates and removes background noise in one step.
|
Z-score computation
|
(Below) The complete data flow for four frequencies. Dataflow objects are grouped (as labeled) in slightly different conceptual stages. The highlighted operator expresses the Pollock Zi,j equation. This dataflow is available in the Introduction to the Formula operator tutorial.
|
References
For the full citation, see the relevant entry on the References page:
- 1 De Robertis, McKelvey, and Ressler (2010). See Multiple frequency analysis.