Figure 5
From: Driving singing behaviour in songbirds using a multi-modal, multi-agent virtual environment
Analysis pipeline for real-time motion detection. (A) The pipeline for analysis uses the methods described in Fig. 4A–E for segmentation and feature extraction. The resulting feature vector is post-embedded in the 2D space and classified based on which region it falls within. Syllables are collected in sequences and screened for motifs. (B) The behaviour map annotated manually based on sample sound segments from each region. (C) Song spectrogram with vertical lines indicating onset (green) and offset (red) events of automatically detected syllables. The letters indicate the class of the syllable.
