Recording and reproducing speech airflow outside the mouth

Abstract

Creating a technique for artificially producing accurate time and intensity varying airflow can help us better understand the influence of airflow on speech perception.  Such a technique would allow differentiation between airflow from stop release bursts, affricates, and fricatives.  While customized high-speed short-burst airflow systems for testing somatosensory responses have been available since 1987 (Hashimoto, 1987; Mizobuchi, et al., 2000), the systems were difficult to obtain outside of Japan, so researchers studying the effects of airflow on speech perception began with solenoid operated on/off airflow devices (Gick & Derrick, 2009, Gick, et al., 2010).  These systems were adequate to producing artificial airflow for aspirated stops such that correctly time-aligned airflow enhanced their perception in noisy environments.  Recently, however, we created a custom airflow production system, called Aerotak.  Aerotak itself relies on Murata's Microblower for airflow. The Microblower is a commercially available 20x20x1.85mm piezoelectric air pump with a maximum 0.8 l/m flow, 19.38 cm/h2o pressure, and an ability to change air-flow from 10% to 90% max cm/h20 in approximately 90 milliseconds.   Aerotak allows researchers to produce artificial approximations of continuously varying airflow in speech (Derrick, et al. 2014a). We used Aerotak to demonstrate that airflow can enhance perception of English stops and fricatives (Derrick, et al. 2014b).  We have also designed a custom mask-less airflow estimation system that allows accurate recordings of changes in turbulent speech airflow from the lips.  In this presentation and paper, we will describe how to record airflow from speech, process the data for use with Aerotak, produce artificial speech airflow, and confirm the quality of the artificial production using our mask-less airflow system.  Examples will include aspirated stops, fricatives and affricates from English speech.