Page 58 - Summer2020

SPEECH ACOUSTICS
handful of languages (e.g., English, French, German, Dutch, Japanese, and Chinese) came to dominate most publications investigating the acoustic characteristics of speech. A notable exception for JASA is a paper in a special volume on communication from 1950 in which the author, John Lotz, points out the need for studying speech acoustics from a cross-linguistic perspective (available at asa.scitation.org/toc/jas/22/6).
“Every speech event belongs to a definite language. Any speech analysis that disregards this fact...will lack adequate principles for the classification and description of the com- plexities of speech” Lotz, 1950, p. 712.
Although most of the papers in this special volume were theoretical in nature and therefore contain very little actual acoustic information and no acoustic investigations about specific languages, Lotz’s point is well founded. Despite the call for more diversity in the acoustic studies of languages in 1950 and despite a renewed interest in languages of the world among linguists, the bias toward relying on a hand- ful of languages persists even today.
This general lack of acoustic description and research on underdocumented languages in JASA and in other journals inspired us to host two special sessions on the phonetics of underdocumented languages at Acoustical Society of America meetings (in Salt Lake City, UT, May 2016, and Victoria, BC, Canada, November 2018) and to organize a special issue on the phonetics of underdocu- mented languages (Tucker and Wright, 2020). One goal of the special issue is to increase the number of under- documented languages described in JASA. In the special issue, there are descriptions of aspects of 25 different underdocumented languages from 5 different continents.
Acoustics of the World’s Languages
Many sounds have been described phonologically in lin- guistic grammars of languages, descriptions that explore and explain the patterns of a given language, although the phonological section in these grammars typically makes up a very small portion of the grammar. Most linguistic grammars use impressionistic methods where essentially the researcher writes down what they think they hear. In these grammars only a fraction of the described sounds have been examined in phonetic, and particularly acous- tic, detail. PHOIBLE is a database of speech sounds that lists 3,183 speech sounds in 2,186 languages (Moran and
McCloy, 2019). Many, maybe most, of these sounds and sound combinations have not been described acoustically. First, we follow the International Phonetic Alphabet (2018; see at acousticstoday.org/ipa-chart) conventions for tran- scribing speech sounds and indicate that these are speech sounds using square brackets on either side of the sound.
Then, we briefly describe some of the unique acoustic characteristics of speech sounds from several different languages that are acoustically underdocumented.
Source-Filter Model of Speech Production
A simplified way of modeling speech sound produc- tion is using a source-filter model (Chiba and Kajiyama 1945; Fant, 1960) with a source (e.g., vocal-fold vibra- tion or aperiodic turbulence) that is filtered by the shape of the vocal tract (Figure 4A). A way to realize this model is by using a tube as the filter with a source at one end of the tube (e.g., Figure 4B). There are a variety of possible sources at different points along the vocal tract. The easiest vocal tract configuration to start with is a vowel, where vocal-fold vibration creates a complex
Figure 4. A: midsagittal view of the human vocal tract with important places of articulation labeled. B: neutral closed- open tube with labels indicating the approximated places of articulation. The closed end is the dark oval on the right at the vocal folds and the open end is located at the labial end of the tube. See text for further explanation.
58 Acoustics Today • Summer 2020