Technique for frequency division/multiplication of speech signals
James L. Flanagan, S. W. Christensen · The Journal of the Acoustical Society of America · 1980
As a follow-up to two previous studies, we examine the feasibility of dividing the bandwidth of a speech signal while keeping the time dimension intact. We utilize the real and imaginary parts of the short-time Fourier transform to describe contiguous bandpass segments of the signal spectrum. We implement a frequency division that maintains unchanged the Hilbert envelope of the waveform for each bandpass segment. We show that the resulting signal can be usefully confined to one-half the original bandwidth. Transmission can be effected either by an analog channel of one-half bandwidth, or by digitally encoding at one-half the conventional Nyquist rate. Reanalysis at the receiver and complementary multiplication of the spectrum parameters reconstruct the signal to the original bandwidth. Analysis, transmission, reanalysis, and reconstruction processes are not transparent. They provide an output of less than typical toll quality, but of reasonably good communications quality. A potential use for the method might be the simultaneous transmission of voice and data over a single voice circuit. We also show that the division and multiplication technique can be used to modify the time dimension of the speech signal (i.e., compress or expand the time scale) while keeping the frequency dimension intact.