Skip to main content
Fig. 5 | BMC Biology

Fig. 5

From: TrackUSF, a novel tool for automated ultrasonic vocalization analysis, reveals modified calls in a rat model of autism

Fig. 5

Analyzing bat USVs recorded in a natural scene using TrackUSF. A The t-SNE analysis of all USFs extracted from audio recordings of bat calls, made from three distinct heights (50, 100, and 150 m) above ground level at the Hula Valley. These recordings were preselected to include vocalizations of either R. microphyllum or P. pipistrellus. Each USF is represented by a dot, color-coded according to species and height. Note the gray line manually defining the noise USFs. B Automatic clustering of the USFs based on the t-SNE analysis. Note that the cloud of noise USFs defined by the gray line in A is now broken into multiple clusters, while only clusters 1–3 represent real vocalizations. C Examples of spectrograms of vocalizations represented by USFs of the three clusters. Insets are showing one magnified vocalization per spectrogram. D Mean PSD profiles of USFs from the noise clusters (gray), cluster 1 (blue), and clusters 2–3 (red and orange). Inset showing the peaks of clusters 2–3 in higher magnification. E Mean number of P. pipistrellus USFs detected in each height, separately plotted for the three distinct clusters of USFs. Each dot represents one audio clip. F Mean number of R. microphyllum USFs detected in each height, separately plotted for the three distinct clusters of USFs. Each dot represents one audio clip. **p<0.01, ***p<0.001, post hoc Dunn test following main effect. All error bars represent SEM

Back to article page