Ridge extraction based on adaptive variable-bandwidth cost functions by edge detection of time frequency images

Abstract

Ridge extraction of a time frequency image (TFI) can serve to estimate an instantaneous frequency of a signal. However, there is a considerable difficulty in ridge extraction of a multi-ridge TFI due to mutual interference between ridges. Some traditional methods for ridge extraction fail to resolve this difficulty. The one-step cost function (OSCF), a typical method for processing a multi-ridge TFI, lacks self-adaptation in determining searching regions of a targeted ridge. To overcome this question, this paper proposes an adaptive variable-bandwidth cost function (AVBCF) for ridge extraction. Firstly, a TFI is acquired by time frequency analysis of a signal. Afterwards, the TFI is filtered into a binary image by a Canny detector. Also, some dispersive ridges in the binary image serve to form a synthetic dispersive ridge. Next, searching regions of a targeted ridge are estimated by projecting statistics of edges of the synthetic dispersive ridge to the targeted ridge. Next, the AVBCF is compared with the OSCF and some widespread methods numerally and experimentally. The results demonstrate that the AVBCF performs better than the other methods.