This study shows how deep learning can identify heart activity from chest vibrations, even in real-world conditions, by using data from multiple sensors and personalizing the model for each user.