Turning Detection in Sandbar Sharks Through Accelerometer Data
Detecting turning using accelerometer data alone is typically impossible, but by putting the indirect “tells” associated with turning through a supervised classifier, periods of frequent turning can be observed.
Marine wildlife is difficult to track or observe through cameras. Instead, attachable tags, typically incorporating accelerometers or gyroscopes, are often used instead. The raw data from those tags is processed and fed into supervised classifiers, machine learning algorithms that compare the data to reference data to make predictions. Many researchers (Whitney et al., 2010; Noda et al., 2013; Brownscombe et al., 2014) have successfully used accelerometers to classify long, periodic behaviors in fish, such as swimming, coasting, and mating. However, shorter and more complex behaviors, such as turning, are not as frequently covered.
Existing methods for detecting turning require more...
- Jacobstein N, Porter B (eds) (2011) 11th IEEE international conference on data miningGoogle Scholar