Locally Periodic Signal Estimation and Pitch Detection, and Acoustic Communication in American and Northwestern Crows

Abstract

Crows and other members of the genus \emph{Corvus} employ a complex system of acoustic communication, based largely on the calls known as "caws." These are locally periodic signals with rapidly varying pitch, and are difficult to analyze with conventional methods of pitch detection and periodic signal estimation. In this dissertation, I present novel methods of reconstructing locally periodic signals from noisy data, improve the performance of autocorrelation and the Pseudo-Maximum Likelihood Estimator as period indicator functions, and introduce averaged autocorrelation as a new period indicator which combines the advantages of autocorrelation and the PMLE. Using these signal analysis tools, I identify acoustic correlates of sex, behavioral context and individual identity in the caws of the American crow. Finally, I demonstrate that mobbing behavior in the Northwestern crow can be explained by the adaptive functions of antipredator surveillance and communication, and that the "funeral" behavior exhibited by crows toward dead conspecifics may simply be another facet of mobbing behavior.