How can I keep a percentage r of the low frequencies of a Nx1 signal x using the Discrete Cosine Transform of it? Is it simply zeroing all but the first rN elements of DCT(x)?
What I've understood is that k-th DCT-II component is expressed in a basis of cosine functions with frequency $k\pi/(2N)$ rads, where $N$ is the number of samples of the signal. Thus k-th coefficient of such a signal has frequency of $k/(2N)$ in Hz. It seems to me that my understanding is wrong because this constains highest frequency to $(N-1)/(2N)$.