TIME-FREQUENCY DISTRIBUTIONS BASED ON GENERALIZED CONE-SHAPED KERNELSFOR THE REPRESENTATION OF NONSTATIONARY SIGNALS

A generalized category of cone-shaped kernels is proposed. Analysis of the kernel in the 2-D time, 2-D frequency, and ambiguity domains is p erformed. The shape of this kernel in the 2-0 time plane is bow-tie, w hich effectively suppresses cross-terms especially in the frequency di rection. In the 2-0 frequency plane, the shape of the kernel is of a l ateral inhibition form, which enhances spectral peaks when convolved w ith the instantaneous spectral correlation. By investigating the new k ernel in the ambiguity domain, it is shown that the resulting distribu tion has many desirable properties that encourage its use as a time-fr equency signal analysis tool. By proper selection of kernel parameters , a family of kernels is obtained, hence, providing many kernels which may be used in the analysis of different types of signals, the most i mportant of which are the Born-Jordan and ZAM (Zhao, Atlas and Marks) kernels. Experimental results on a simulated test signal that represen ts two short-duration Gaussian pulses show the advantage of the propos ed distribution in comparison with ZAM and Born-Jordan distributions. (C) 1998 The Franklin Institute. Published by Elsevier Science Ltd.