ResNet Based Classification of Female Menstrual Circle from Pulse Signal

The regularity of the menstrual cycle can reflect the physiological health of women. Nowadays, more and more women are paying attention to the problems of their own physiological period. However, there are few researches on this subject in the academic field, and most of them are explored from the perspective of ECG. The collection equipment based on ECG has the high cost. In this paper, we attempt to identify the characteristics of menstrual cycle in healthy female college students from the perspective of pulse. We extract the effective time frequency characteristics through wavelet transform, and make them as input of the residual neural network then, so as to recognizing the characteristics of different physiological stages. We have collected 120 different female pulse once and tracked the pulse of a female volunteer for 3 months in order to imitate the traditional Chinese medicine and the family-based “personal doctor”. From the two aspects, we train and predict the next round of women’s physiological period. The accuracy reached 66.7% and 81.8%, respectively.