Master Assignment
automatic identification of bat species by their echolocation calls
Student project on bat sound analysis
Type: Master EE/CS
Period: (TBD)
Student: (Unassigned)
If you are interested please contact :
Description:
Student project on bat sound analysis
Very few organisms use vision for mid-air catching of small flying prey in low-light conditions. Bats emit loud, brief clicks and use the echoes to localize prey and objects in their surroundings. In the Netherlands the frequency of these bat echolocation calls varies between 20 and approximately120 kHz; call duration varies between 1 – 15 ms and call repetition rate between 2 – 20 calls/s. Bat echolocation sound can easily be recorded with a sensitive microphone and digitized. A common challenge is to use these calls to identify bat species. It is thereby important to know that the properties of echolocation sounds may depend on the direct environment the bat is flying in. For example, most species wait with the emission of the next call until the echo of the previous call has been received. Hence, the distance to the nearest object determines the pulse repetition rate and pulse duration. Likewise, the frequency of bat calls depends on how cluttered the direct environment is: in open environments, bats tend to lower frequency in order to produce wide-range calls. Conversely, bats increase frequency in closed spaces to save energy and get more precise information on their surroundings. However, as many bats occupy specific ecological niches, about half of the (~19) Dutch bat species can be identified based on their echolocation sound.
Here, the challenge is to use advanced machine learning techniques to link specific bat call parameters (or combinations of these) to species, and to use these to identify species in large volumes of automatically collected bat recordings. For this a reference set of manually identified calls (currently made at NIOO) will be used. Hurdles include f.e. harmonics and different species flying simultaneously at the same location. Ample possibilities for the application of such an identification system are present. Because of their high protection status, generally there is a high demand for automation of bat call analysis.