Abstract: A novel multi-component instantaneous frequency (IF) extracting method using the corresponding relation between the position information of time-frequency ridges and IF of signal is introduced. After the redial Gaussian kernel time-frequency distribution (RGKD) is studied, the improved RGKD is proposed. By segmenting the one dimensional time signal in time domain by way of adding windows, the whole time series is cut into many sub-segments and the whole RGKD can be obtained by concatenating the sub-segment RGKD. Then, the whole RGKD is regarded as an image, and the ridges of the image are extracted by using digital image processing method including smoothing processing, two dimensional Laplacian operator and refining treatment. Hough transform for detecting lines automatically is used, especially, the process of detecting beeline and sine curve are discussed. The validity of the whole method has been verified by computer simulation and fault diagnosis in gear box.

Key words: Curve detection, Hough transform, Instantaneous frequency, Ridges extraction, Time-frequency distribution