This invention discloses a non-contact electrocardiogram (ECG) measurement method using
millimeter-
wave radar. First, the original chest wall displacement
signal acquired by the
millimeter-
wave radar is preprocessed to obtain a displacement feature
branch input. Then, the original displacement
signal undergoes cardiac sound band enhancement,
wavelet denoising, and envelope extraction to obtain
radar cardiac sound features. A hidden semi-Markov model is constructed to model the states of the first heart sound,
systole, second heart sound, and
diastole in the
cardiac cycle, obtaining the
posterior probability of the state at each time point. Next, the displacement features,
radar cardiac sound features, and state probability features are input into a multi-
branch convolutional fusion network, combined with a bidirectional long short-
term memory network to achieve
temporal correlation modeling, finally outputting the reconstructed human ECG result. This invention achieves non-contact
ECG measurement without
skin electrode contact by fusing the original mechanical displacement of the chest wall, the
mechanical vibration features of the
heart sounds, and the explicit state
prior information of the
cardiac cycle.