SonoBat automated classification

SonoBat bat echolocation call analysis software applies the most exacting analysis method available, enhanced with signal processing innovations not available elsewhere. SonoBat's classification algorithms use the most extensive library of reference recordings expertly collected and verified by bat biologists working in deliberation over the past 20 years to fill out the call repertoires of each species. In contrast, other classifiers use more limited source libraries, some based on user submitted reference calls. Variability in recording conditions, quality of recordings, and especially variations in the calls that the bats themselves make (with overlap in characteristics among many species) impart substantial challenges to acoustic bat classification. SonoBat has logic to recognize many of these confounding influences and performs a number of stringent quality control procedures and redundant checks to only output a classification when confident using many logic steps based on two decades of expert experience working with these bat species' calls. For some background information, review the Eastern or Western Classification notes to recognize and interpret some of the challenges inherent in this work refer to further information regarding classification, and specific classification of Indiana bats (Myotis sodalis) and little brown bats (M. lucifugus).

Although some other classifiers such as Kaleidoscope accept full-spectrum signals, they extract data from calls using zero-crossing analysis. Although fast, it lacks fine scale resolution and cannot track call data as robustly through noise or when signals become weaker from bats at distance. Using the full information content of full-spectrum data enables tracking call signal trends when weaker than ambient sound and discriminating call content from noise and echoes. SonoBat takes advantage of this full information content for more precise analysis and data integrity (Fig.1).

 
Lano comparison plot
 
     
  Figure 1. The same silver-haired bat call processed in Kaleidoscope Pro (left) and SonoBat (right) showing the trend lines used to acquire the data used in classification of the calls. SonoBat uses the complete information content inherent in full-spectrum data to track call signal content using an intelligent tracking algorithm. This provides more robustness and accuracy on which to base species classifications. For more examples see the SonoBat difference and view this presentation.  

The SonoBat approach of robust and precise data extraction from unknown field recordings and expert system classifiers based on data extracted from extensive reliable reference libraries yields the most reliable classification performance available (Tables 1 and 2).

 
Eastern spp comparison
 
     
  Table 1. Comparative classification performance of Kaleidoscope and SonoBat on 7115 good quality species-known recordings collected and processed by Janet Tyburec in Missouri (2014). "% classified" indicates the percentage of known recordings given a decision, and "% correct" indicates the rate of correct classification of those outputted with a decision.
** The SonoBat classifier used for this study did not include the gray bat (MYOGRI). In other tests, SonoBat has classified 90.3% of MYOGRI recordings with 100% correct.
 

 

 
US west comparative results
 
     
  Table 2. Comparative classification performance of Kaleidoscope and SonoBat on 5224 good quality species-known recordings collected and processed by T. Malloy of Stanford Jasper Ridge Biological Preserve in California (2014). "% classified" indicates the percentage of known recordings given a decision, and "% correct" indicates the rate of correct classification of those outputted with a decision.