Highly efficient online stochastic gradient and sliding window stochastic gradient signal modelling methods for multi-frequency signals

This paper designs the signal parameter identification methodology for the signal which is composed of the sine components and cosine components. With the help of the gradient search, a stochastic gradient modelling method is presented to estimate all of the trait parameters of the multiple sine-cosine components. Further, some improvement schemes are designed to be aimed at enhancing the precision and convergence speed. Moreover, a rolling optimisation loss function based on the cumulated dynamic measurements is proposed to present a highly efficient and high precision signal modelling methodology. Finally, the algorithm emulation is introduced to confirm the feature of the proposed signal modelling methodologies in improving the accuracy of parameter estimation.