Animal communication and specifically acoustic communication is in the focus of ecological and biological research. With the advancement of monitoring technology, a vast amount of acoustic recordings of birds is continuously accumulated. As manual segmentation and annotation of this data is impractical, development of efficient algorithms for accurate detection, classification and segmentation of birdsong is therefore a prerequisite for further analysis. In this study we present an algorithm for automatic segmentation and parameters estimation of one type of bird vocalization, namely, the trill song. The algorithm is based on computing the short-Time variance of the fundamental frequency derivative of bird acoustic signal for initial detection of syllables. The boundaries of each syllable are consequently obtained using a Gaussian smoothed short-Time energy function and an adaptive threshold based on the energy envelope. The performance of the algorithm was evaluated using a comparison to human expert segmentation, as well as to ground-Truth values of synthetic trills produced by the Harmonic + Noise model. A correct detection rate of more than 95% was yielded for SNR levels of-5 dB or higher for signals with additive colored noise, and for signals with additive white Gaussian noise more than 92% was obtained for SNR>-5dB. In addition, a high correlation between the automatic segmentation and that of a human expert was exemplified.