Speech Band Compression by Analytic-Signal Rooting
Manfred R. Schroeder, James L. Flanagan, E. A. Lundry · The Journal of the Acoustical Society of America · 1966
Speech is filtered into contiguous passbands. The signal s(t) in each passband is represented as Re[σ(t)], where σ(t) is the analytic signal corresponding to s(t). By definition, σ(t)=(t)+jŝ(t), where ṡ(t) is the Hilbert transform of s(t). The nth root of the analytic signal is formed, and the real part of the rooted signal is confined to a fraction of its original frequency range. After transmission, the rooted signals are restored to their original frequency positions by raising to the nth power. The restored signals from each channel are then combined to synthesize an approximation to the original speech signal. A complete transmission system has been simulated on a digital computer, and the results of restricted bandwidth transmission are demonstrated.