Performance comparison of multi-component signals using WVD and Cohen´s class variants

Time Frequency distributions (TFD), are representations that shows in which way the energy of the signal is distributed over the time and frequency dimensions. Since there is no such TFD which is suitable for all applications, many TFDs have been formulated, where each corresponds to a different, fixed kernel distribution. The greatest demerit of all fixed kernel TFDs is that, none of them perform adequately for signals that differ in how they are aligned in time-frequency region. Hence fixed kernel TFDs provide good performance only for a small class of signals. To overcome this demerit we have used a signal dependent TFD which utilizes a radially Gaussian kernel. The kernel that is produced by these block oriented techniques doesn´t vary as the input signal varies with time and so they are unsuitable for on-line implementation. Hence for such signals whose properties change with respect to time, or for lengthy signals, an adaptive signal dependent TFD is more desirable. This method changes the kernel for every time intervals to achieve optimal performance, so as to follow the changes in a signal. In this paper, we compare the different fixed kernel TFDs of Cohen´s class-Wigner-Ville, Choi-Williams, Zhao-Atlas-Marks and Born-Jordan distributions with the signal dependent TFD and adaptive optimal kernel TFD by applying them to multi-component signals.