Contactless Seismocardiography via Deep Learning Radars

A contactless sensor includes a cardiac beamformer, a wireless-to-seismocardiogram translator, and an automatic labeler. The cardiac beamformer determines at least one beam for receiving wireless signals generated based on movement of a heart. The at least one beam is generated based on phase information and a heart signal extracted from a time-domain signal generated from one or more receiver elements. The wireless-to-seismocardiogram translator implements a convolutional neural network to transform time-series data detected from the at least one beam to a seismocardiogram. The automatic labeler identifies and labels one or more micro-cardiac events in the time-series data. The cardiac beamformer may be considered an optional feature in one or more implementations.