A new application of digital spectrum estimation algorithms is presented. The frequency content of ultrasonic echolocation signals emitted by various bat species are investigated. Using several different analysis algorithms, an empirical determination of the most suitable technique for this class of waveform is reached. In addition, interspecies and intraspecies classification of bats utilizing the most appropriate estimators is investigated for feasibility.
The motivation for this research arose from a need in the zoological community for faster, more efficient, and more accurate analysis tools. Using computers and digital spectral estimation techniques, in conjunction with a high-speed analog-to-digital converter and broadband microphone, spectral information could be obtained and analyzed faster, and more efficiently due to the elimination of unnecessary or inappropriate processing.
Results showed that the most appropriate analysis algorithm for these waveforms was dependent on the spectral properties of the particular signal. The FFT with a Kaiser window applied was shown to be very useful, and the Autoregressive parametric estimator with an approximate system order of 8 to 10 also proved to be advantageous in certain situations. However, the conclusion is reached that accurate spectral interpretation will require analysis over several time bins in the signal, in order that frequency information with respect to time can be obtained.
The spectra produced by the estimators indicated surprisingly distinguishable characteristics with peaks that were easily discernable. Past research findings regarding spectral shape of echolocation bursts were supported, and the implication is that classification of species or even individuals is possible. With some modification and enhancement, the system could be contained in a portable computer, allowing use in remote locations where classification is the primary intent.