SonoBat 3 intelligent processing of bat call data

SonoBat 3 applies intelligent call recognition and feature extraction routines for automated processing of individual calls and sequences of batches of files or directories. It will also batch process files to extract time-frequency and time-amplitude parameters from calls within sequences. View this presentation to learn the details about what SonoBat 3.1 does and watch this movie clip of SonoBat 3 in action.

Confident automated data parameterization from field recorded bat calls depends upon faithful recognition of bat call trends and recognition and rejection of poor quality call samples. SonoBat implements a number of sophisticated routines to minimize the influence of noise and obscuring echoes to optimize faithful call trending. SonoBat also implements multiple assessments of signal reliability and secondary checks call trends to ensure to minimize the inclusion of unreliable signals and call fragments that lack the full information content of calls.

Because bats vary the amplitude through their calls, the farther a bat flies from the detector the more the lower amplitude portions fall below the microphone's sensitivity for detection. This truncates calls to just their strongest portions along with the loss of call content. In some cases these fragments of fully formed calls can mimic other species, e.g., the body fragment of a Myotis lucifugus may render as a simple curved call that mimics a Lasiurus borealis. SonoBat performs a number of signal quality checks to reject poorly formed calls, overloaded calls, or those with distorted signals or too much noise to optimize faithful call parameter extraction and classification.


This image from a poster presented at NASBR 2012 by S.E. Romeling et al. compares SonoBat automated call data extraction with other systems. SonoBat output fewer pulses (upper left) from the same data set as the other systems because it rejected lower quality call fragments. SonoBat only processed and reported calls having the most robust call content. The data set from SonoBat had the longest reported duration (upper right) and greatest bandwidth as evidenced by the lowest minimum frequency (lower left) and highest maximum frequency (lower right). The SonoBat data would most faithfully represent the call content vocalized by the bats.


call trending sho
  Example call with substantial echo obscuring the end of the call. SonoBat's automated intelligent call trending analysis recognized the call trend did not extend into the echo content. This would prevent erroneous parameter extraction.  


Intelligent call trending of SonoBat using full-spectrun data compared with zero-crossing call rendering of same signal.
  Example Myotis lucifugus call with mild trailing echo. SonoBat intelligent call trending followed the downward ending trend of the call despite its diminishing amplitude that fell below the trailing echo from the stronger body of the call. The zero-crossing rendering of the same signal (right) followed the echo rather than the myotis-discriminating downward trend because zero-crossing analysis can only detect the strongest frequency component at any time interval. The zero-crossing generated call trend appears to turn up at the end of the call and that would likely render a misclassification as a Lasiurus borealis rather than a myotis species.  

call trending sho

Example call with background signal noise overlapping the call. SonoBat's automated intelligent call trending analysis recognized the call trend and its end despite the confounding noise. This would prevent erroneous parameter extraction.


call trending sho
  Another example call with more challenging echo content obscuring the end of the call. SonoBat's automated intelligent call trending analysis recognized the call trend did not extend into the echo content. This would provide faithful parameter extraction of the call content despite the obscuring echo.  

For additional examples of SonoBat's intelligent call trending performing under different conditions follow this link and view this slide presentation.

For a more complete overview of the benefits of full-spectrum analysis of bat echolocation calls, and the differences in data interpretation between full-spectrum and zero-crossing, view this presentation.


return to front page