“…The different observation vectors are contained as columns of V, allowing to represent a single column element as v = Wh, where h is the corresponding individual column of H which describes the amount that each basis vector contained in W is required for such representation. The total number of basis vectors corresponds to K, which is the common dimension of H and W. In speech and audio applications, it is common to choose F as the number of frequency points representing the spectrum [171], and V will contain in each of its columns, the periodogram or power spectrum of an individual noisy segment. The dimension R usually corresponds to the number of processed segments.…”