Contribute Media
A thank you to everyone who makes this possible: Read More

Neural Networks for Segmentation of Vocalizations


Neural networks for speech-to-text avoid dividing speech into segments, such as syllables, but segmenting has important applications. We compare different neural networks for segmentation of vocalizations using the song of songbirds, which we study as neuroscientists. Initial results suggest a bidirectional LSTM-CNN architecture outperforms others in both segmentation and classification.


Improve this page