Figure 3 | Scientific Reports

Figure 3

From: Machine learning and statistical classification of birdsong link vocal acoustic features with phylogeny

Figure 3

Principal component analysis (PCA) of all 21 acoustic features measured from each syllable. Syllables cluster within species (distinct clusters of the same color) and across species (muti-color clusters). (a,b) Syllables of African and Australian species clustered separately along PC1, with African species having higher values than Australian species, and the Bengalese distribution spanning the majority of the range between them. (a,c) The syllables of species with broadband species (ZF, DF, BF) and those with narrowband, tonal songs (LF, RF, CB, GW) clustered along PC2. (b,c) There are relatively uniform distributions along PC3, except for in the BF, whose distribution includes many low values in PC3. Spectrograms on the axis anchors are synthetic sounds that represent what each respective PC axis captures. These spectrograms show the distribution of energy across frequency (y-axis) and time (x-axis). Spectrograms to the right of each plot show real syllable samples from the respective plot. Symbols within the syllable spectrogram are color-coded according to species identity. Syllable locations within the plots are outlined in their respective shape (black outline used here for contrast). Overall, sample syllables confirm the distribution of energy expected for different positions on each graph (i.e., by combination of the axis coordinates).

Back to article page