Deep learning has recently revolutionized the field of machine hearing and vision, by allowing computers to perform human-like activities including seeing, listening, and speaking. Such systems are constructed from biomimetic, ‘deep,’ artificial neural networks. Now a team of researchers at the University of Washington School of Medicine has created deep learning-based software that can understand high-pitch calls of rodents.
Called DeepSqueak, the software program takes an audio signal and transforms it into an image, or sonogram.
By reframing an audio problem as a visual one, the researchers could take advantage of state-of-the-art machine vision algorithms developed for self-driving cars.
DeepSqueak represents the first use of deep artificial neural networks in squeak detection.
“DeepSqueak uses biomimetic algorithms that learn to isolate vocalizations by being given labeled examples of vocalizations and noise,” said co-author Dr. Russell Marx, a researcher in the Psychiatry Behavioral Sciences at the University of Washington.
“Our goal is to develop treatments for withdrawal from alcohol or opioids,” said senior author Professor John Neumaier, head of the Division of Psychiatric Neurosciences and associate director of the Alcohol and Drug Abuse Institute at the University of Washington.
“If scientists can understand better how drugs change brain activity to cause pleasure or unpleasant feelings, we could devise better treatments for addiction.”
“The animals have a rich repertoire of calls, around 20 kinds,” said co-author Dr. Kevin Coffey, a postdoctoral fellow at the University of Washington.
“With drugs of abuse, you see both positive and negative calls.”
“The rodents seem the happiest when they are anticipating reward, such as sugar, or are playing with their peers. Interestingly, when two male mice are together, they make the same calls over and over.”
“However, when they sense a female mouse nearby, their vocalizations are more complex, as if they are singing a courtship song. This effect is even more dramatic when the male mouse can smell but not see the female mouse.”
“This observation suggests that male mice have distinct songs for different stages of courtship.”
DeepSqueak is highlighted in a paper published in the journal Neuropsychopharmacology.
Kevin R. Coffey et al. DeepSqueak: a deep learning-based system for detection and analysis of ultrasonic vocalizations. Neuropsychopharmacology, published online January 4, 2019; doi: 10.1038/s41386-018-0303-6