Speech as an Error-Resistant Digital Code
Cecil H. Coker · The Journal of the Acoustical Society of America · 1974
Distinctive features, dominant perceptual cues—indeed many concepts of linguistics equate redundancy with waste. In doing so, they ignore an important principle of communication—that noise resistance derives from duplication. The remarkable indestructability of natural speech is well known. Doubtlessly that error resistance results from worthwhile redundancy. This paper presents a treatment of speech as the transmission of discrete units (words) coded as varying-length sequences of smaller elements (phonemes) over a noisy acoustic channel. This is, in every respect, a problem in coding theory, and the literature on error recovery in Huffman-like codes is extremely pertinent. In such codes, the dominant error problem is to maintain synchronism between sender and receiver. With poor codes, a single error can propagate to the end of the message, with the receiver never again correctly determining where one word ends and another begins. In practical codes, the mean time to recover from error is shortened by frequent occurrence of sequences that clearly signal the location of a boundary. In this presentation we review some of the coding literature. We argue that intonation and durational gestures, and consistent allophone variation at boundaries all serve an indispensable role: synchronization cues for error resistance.