Digit Concatenation from Simple Rules

A previous paper by Olive and Nakatani [“Quality Judgments of Spliced Synthetic Telephone Numbers,” J. Acoust. Soc. Amer. 50, 145(A) (1971)] described an experiment to evaluate the relative importance of pitch, timing, and formant coarticulation rules in digit concatenation from isolated digits. That paper concluded that pitch was the most important aspect of concatenation while the other two parameters did not contribute much to the quality of the concatenated speech. This paper will describe the simple concatenation rules that were based on the recommendations of the previous paper, and evaluate the results. A digit recall test was administered to measure comprehension and its results prove that the rule‐concatenated numbers are at least as recallable as the same numbers spoken naturally. A subjective preference test revealed that the rule‐concatenated speech was judged to be close to speech resynthesized from natural speech. Half the subjects preferred the rule‐concatenated speech to speech spliced from isolated digits even when the isolated digits were naturally spoken; the other half disliked any form of synthetic speech.