A speech scrambler based on time-domain sample scrambling is described. Synchronization between transmitter and receiver is unnecessary, because special windowing and scrambling combinations make the process equivalent to a frequency-band scrambler with a limited number of possible scrambling permutations. The required computation is typically only 4-8 multiplications per sample. Although processing is digital, the output is intended for analogue speech transmission over standard telephone channels. In addition to the message being scrambled, speaker identity is also masked in the scrambled speech, but it is readily recognized in the descrambled output. The algorithm can be modified to form a cheap but low-quality bandwidth compressor, and to correct the pitch of recorded speech played back at a different rate. A limited number of simulation results are given, and the system is compared with two previous time-domain scramblers.
展开▼
