Pattern of Air Flow out of the Mouth during Speech

Since the 19th century, kymographic recording of total air flow out of the mouth has been used to diagnose the varying durations and degrees of constrictions of the vocal tract during speech. The present project attempts to introduce a second dimension to recordings of air flow out of the mouth—namely, cross‐sectional area of flow—on the hypothesis that this will reflect changes in the location and cross‐sectional shape of constrictions of the vocal tract. Consequently, the findings are pertinent to automatic speech recognition and allied objectives. The procedures included the development of a matrix of 64 hot‐wire anemometers and associated circuitry, where output is sampled and digitized for computer processing. Computer programs permit storage, averaging, transposition, normalization, and matching of flow patterns, as well as sustained CRT display and numerical readout. The patterns associated with various sounds and then within‐ and between‐speaker variance are reported.