GENERATE BILINEAR TIME-FREQUENCY TRANSFORMATIONS BILINEAR TRANSFORMATIONS transform the time domain signal (the variable INPUT SIGNAL) to an output quadratic time-frequency distribution (the variable OUTPUT TIME-FREQUENCY ARRAY) There are many different types of quadratic time-frequency distributions that can be specified: (i) Wigner-Ville Distribution (WVD) (ii) Spectrogram (SPECX) (iii) S_Method (S_Method) (iv) Smoothed Wigner-Ville (SMOOTHED) (v) Rihaczek-Margenau-Hill (RM) (vi) Choi-Williams (CW) (vii) Born-Jordon (BJ) (viii) Zhao-Atles-Marks (ZAM) (ix) Cross-Wigner-Ville Distribution (XWVD) (x) B-distribution (B) (xi) Modified B-distribution (MB) (xii) Extended Modified B-distribution (EMB) (xiii) Compact kernel Distribution(CKD) (xiv) Multidirectional Kenel Distribution (MDD) and also an Ambiguity Function with transforms the signal into the dopple-lag domain. The various parameters associated with the distributions are as follows: TIME-FREQUENCY ARRAY (tfd) The computed time-frequency distribution. size(tfd) will return [a, b], where a is the next largest power of two above FFT length, and b is floor(length(signal)/time_res) - 1. INPUT SIGNAL Input one dimensional signal to be analysed. An analytic signal is required for this function, however, if signal is real, a default analytic transformer routine will be called from this function before computing tfd. TIME RESOLUTION The number of time samples to skip between successive slices. LAG WINDOW LENGTH This is the lag window length and controls the size of the signal kernel (or instantaneous autocorrelation function) used for analysis (lag_window_length must be odd). The kernel used will be defined from -(lag_window_length+1)/2 to +(lag_window_length+1)/2 in both time and lag dimensions. FFT Length: Zero-padding at the FFT stage of the analysis may be specified by giving an FFT length larger than lag window length. If FFT length is not specified, or is smaller than the lag window length, then the next highest power of two above lag window length is used. If FFT length is not a power of two, the next highest power of two is used. KERNEL OPTIONS Each time-frequency distribution has various options that can be adjusted to produce the desired tfd. See Also: quadtfd, analyt