Speech enhancement in crypto-bridge communication systems
Daniel Wei Lin, Yacov Yacobi · The Journal of the Acoustical Society of America · 1987
In secure audio teleconferencing, a crypto-bridge system synchronously adds simultaneous encrypted signals, modulo a known number P, and broadcasts the result to the participants. Each terminal, using a secret key, then decrypts the modular sum of encrypted signals to obtain the desired ordinary sum of plain text signals. Such secrecy systems are provably secure, assuming the existence of one-way functions (e.g., DES) for the modular sum cipher [E. F. Brickell et al., CRYPTO-87]. In this paper, the implementation of additive crypro systems in a narrow-band (2B + D) ISDN environment are considered. The problem of noise reduction for 6-bit linear PCM speech coding (an essential constraint for bridging the encrypted signals in a single B-voice channel) is discussed. In particular, an enhancement procedure using a prefilter, “dither” noise and an adaptive postfilter is proposed. The system is based on an all-pole model for the degraded speech. The coefficients of the postfilter are easily derivable from the all-pole filter parameters. Informal listening tests have shown that the enhanced speech is near “toll” quality.