Template-based automatic recognition of birdsong syllables from continuous recordings

Abstract

The application of dynamic time warping (DTW) to the automated analysis of continuous recordings of animal vocalizations is evaluated. The DTW algorithm compares an input signal with a set of predefined templates representative of categories chosen by the investigator. It directly compares signal spectrograms, and identifies constituents and constituent boundaries, thus permitting the identification of a broad range of signals and signal components. When applied to vocalizations of an indigo bunting (Passerina cyanea) and a zebra finch (Taeniopygia guttata) collected from a low-clutter, low-noise environment, the recognizer identifies syllables in stereotyped songs and calls with greater than 97% accuracy. Syllables of the more variable and lower amplitude indigo bunting plastic song are identified with approximately 84% accuracy. Under restricted recordings conditions, this technique apparently has general applicability to analysis of a variety of animal vocalizations and can dramatically decrease the amount of time spent on manual identification of vocalizations.

Species-specific audio detection: a comparison of three template-based detection algorithms using random forests

“…Algorithms for species identification have been developed using spectrogram matched filtering (Clark, Marler & Beeman, 1987;Chabot, 1988), statistical feature extraction (Taylor, 1995;Grigg et al, 1996), k-Nearest neighbor algorithm (Han, Muniandy & Dayou, 2011;Gunasekaran & Revathy, 2010), Support Vector Machine (Fagerlund, 2007;Acevedo et al, 2009), treebased classifiers (Adams, Law & Gibson, 2010;Henderson, Hildebrand & Smith 2011) and template based detection (Anderson, Dave & Margoliash, 1996;Mellinger & Clark, 2000), but most of these algorithms are built for a specific species and there was no infrastructure provided for the user to create models for other species.…”

Temporal Structure in Zebra Finch Song: Implications for Motor Coding

“…Syllable onsets and offsets were identified by mapping individual syllable waveforms to amplitude derivative templates using a modified dynamic time-warping algorithm (Rabiner and Juang, 1993;Anderson et al, 1996). Our implementation was developed to match waveform peaks to corresponding templates by finding a warping of time that maximizes the average product of the template and candidate waveforms.…”

Parameter-Free Audio Motif Discovery in Large Data Archives

“…We naturally expect repeated structure in most poetry [13], and although this short poem only has 24 lines in two stanzas, we do find two obvious repetitions as the audio motif (the last line of both verses). 2 This (carefully annotated) code, along with all code and data used in this work is archived at [19]. [19] to hear the original sound file and the discovered motifs.…”

“…Birds, though still a common sight even in cities, are facing threats from habitat reduction. While bird songs have been explored in several research efforts [2][5], like human sound processing, the algorithms tend to be very specialized and parameter-laden. How well can we do with no parameters?…”