Introduction to Spectral Analysis

When listening to birdsong, it is immediately apparent that each song has a distinct rhythmic and sometimes even melodic structure. Songs of individual birds in a flock sound similar to each other and differ from those of other flocks. As early as the 18th century, Barrington noted that the songs of cross-fostered birds differed from the species-typical song, suggesting a role for vocal learning. However, until the late 1950s, there had been no objective way of confirming these observations by physical measurements of the songs themselves. The invention of the sound spectrograph (sonogram) at Bell Laboratories was a significant breakthrough for quantitative investigation of animal vocal behavior.  The sonogram transforms a transient

stream of sound into a simple static visual image revealing the time-frequency structure of each song syllable. Sonogram images can be measured, analyzed, and compared with one another. This allows the researcher to quantify the degree of similarity between different songs by inspecting (or cross-correlating) sonograms and categorizing song syllables into distinct types. Each song is then treated as a string of symbols, corresponding to syllable types, e.g., a, b, c, d..., and song similarity is estimated by the proportion of shared syllable types across the sonograms of the two songs. The procedure is equally useful in comparing the songs of different birds and that of the same bird at different ages or after control and experimental treatments.