To improve the signal-to-noise ratio (SNR) of the dynamic spectrum (DS) data and to increase the stability of the model and the prediction accuracy, the harmonic waves of DS data were introduced into DS method. Sixty samples were determined as the research objects according to the quality of the pulse wave and the distribution of the harmonic waves after further analysis of 110 volunteers’ data acquired in vivo. This paper took whether adding the energy of harmonic waves into the DS data as the division standard to generate two groups. BP artificial neural network was used to establish the calibration model of subjects’ hemoglobin values against DS. The correlation coefficients of the predicted values and the true values in experimental group, containing the energy of harmonic waves, was 0.91, much higher than 0.80 in the control group. Other indexes were all improved too. The results showed that the modified method can enhance the SNR of DS method and accelerate the development of noninvasive blood components measurement based on DS method.