GANspire is a deep learning tool that generates expressive breathing waveforms for art-health applications using an interpretable latent space.
Our design intention for GANspire is to enable both clinicians and artists to create and experiment with breath-based human-machine interactions. We thus proposed to explore deep learning to capture fine-grained, expressive features of human breathing, while simultaneously supporting intuitive, real-time generation of breathing waveforms.
GANspire is currently being developed in tight collaboration with pulmonologists, using a participatory design approach. We first crafted a deep learning prototype by creating a small dataset of human breathing signals sampled in our health lab, then training a WaveGAN over it. We are currently using an algorithmic technique, called GANSpace, to extract interpretable controls from the breathing latent space. We will soon evaluate the quality of the generated breathings in a workshop with expert pulmonologists.
GANspire is being practiced within the RÉESPIRATION art-health research project.
The project is being developed with Baptiste Caramiaux, Thomas Similowski, and Samuel Bianchini in collaboration with the HCI Sorbonne group of ISIR, the R3S department of AP-HP, and the Reflective Interaction group of EnsadLab, in the context of a postdoctoral fellowship program at Faculty of Medicine, Sorbonne Université.