The invention discloses a physiological
signal quality evaluation method and
system for magnetocardiogram. The physiological
signal quality evaluation method comprises the following steps of 1, obtaining a plurality of
cardiac cycle segments; 2, constructing a multi-view sample set; step 3, inputting the multi-view sample set into the improved BYOL network, including sharing an
encoder and a multi-scale path, the multi-scale path corresponding to the input of different
time windows, configuring differentiated projection heads and predictors, and outputting an encoding representation vector and a prediction vector; step 4, generating a cross-scale alignment
error signal; 5, forming joint comparison loss; step 6, based on a DENCLUE clustering
algorithm, dynamically adjusting a density weight through combined comparison loss, and generating a clustering feature center set; and 7, carrying out distance matching analysis on the prediction vector and the clustering feature center set to generate a quality
evaluation result. According to the method, the improved BYOL network and the DENCLUE clustering
algorithm are fused, and unsupervised magnetocardiogram quality evaluation is achieved.