Marginals It can also be interesting to consider the marginal distributions of a time-frequency representation. These marginals are defined as: $$\begin{eqnarray*} m_f(t)=\int_{-\infty}^{+\infty} \mbox{tfr}(t,f)\ df && \mbox{\... ...^{+\infty} \mbox{tfr}(t,f)\ dt && \mbox{\it frequency marginal} \end{eqnarray*}$$ and express, by integrating the representation along one variable, the repartition of the energy along the other variable. A natural constraint for a time-frequency distribution is that the time marginal corresponds to the instantaneous power of the signal, and that the frequency marginal corresponds to the energy spectral density: $$\begin{eqnarray*} m_f(t)=\vert x(t)\vert^2\ \ \mbox{ and }\ \ m_t(f)=\vert X(f)\vert^2. \end{eqnarray*}$$ The M-file margtfr.m computes the marginal distributions of a given time-frequency representation.
© GdR ISIS - Contact