Idle Channel Noise Suppression by Relaxation of Binary ADM-Encoded Speech

The source of idle channel noise in most ADM (Adaptive Delta Modulation) speech encoders is the nonuniformity of the data stream generated by the encoder during silence periods. From a relaxation* study * In the context of this paper, relaxation implies changing the binary data to a state such that any further change will not yield an improvement in the quality of the speech or of silence. 1699 on the computer it becomes obvious that if the decoder is forced into a repetitive input bit pattern synchronized with the main ADM clock, the decoder idle channel noise can be suppressed. The decoder, responding only to the incoming data stream, cannot generate any noise on its own, and a perfect encoder, to assure silence, would provide a repetitive pattern to the decoder. In the absence of such an encoder, it is possible to determine the silence periods during speech by a computed decision and force the bit pattern to be repetitive during such periods. The silence periods are particularly conspicuous in exposing the imperfection of the encoder. Being devoid of any meaningful information, they can become annoying if the encoder does not produce a bit pattern which forces a complete silence for the decoder. For instance, the most commonly used bit pattern is 0 1 0 1 . . e v e n though other patterns such as 001100..., 01001101..., etc., all yield a type of semisilence or humming silence with intertwined frequencies. The nature of decoder silence is also influenced by the companding algorithm and the output filter characteristics.