摘要 |
A speech frequency bandwidth is split into a series of consectuive narrow sub-bands and each band is processed to derive from it phase samples and magnitude samples representative of the signal contents, and expressed in polar coordinates. The contents are slowed down or speeded up respectively by repeating phase and magnitude samples, or deleting them from the contents at a rate depending upon the desired slowing-down or speeding-up rate. Each sub-band of phase/magnitude data is then recombined to form a new sub-band signal, and the sub-band signals are recombined to form speech, slowed-down or speeded up relative to the input. Pref. the processing includes deriving an analytical signal comprising in-phase and quadrative components using complex quadrative mirror filters (COMF) and sampling-down by dropping every other sample. |