Abstract
Abstract We apply discrete Fourier transforms to musical tempo. Then we perform pitch saliency-based melody extraction for each frame. Secondly, we use the finite element differential mathematical model for speech tracking with non-fixed parameters. In this paper, the masking threshold of each edge of the speech signal is dynamically determined by introducing a Bayesian tangent shape model. The simulation results show that the finite element differential mathematical model can remove the noise in the speech signal to a certain extent. In the environment of a low signal-to-noise ratio, the advantages of this algorithm over other spectral subtraction methods are more significant.
Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
