The use of low-frequency ultrasonics in speech processing

One phonetically significant and growing application of ultrasound is ultrasonic speech, as a contribution of ultrasound to speech analysis and processing. The technology acts to augment the human natural speech production system, using data extracted from ultrasonic analysis to synthesize elements of audible speech. In this application, an ultrasonic signal in the kHz range is injected into the vocal tract (VT), which propagates and resonates through the vocal cavities and is emitted from the mouth as ultrasonic speech. This signal could potentially be converted to audible speech by appropriate down-conversion. This chapter introduces ultrasonics, particularly low frequency (LF) ultrasonic waves, and analyses their interaction with the human vocal tract. Ultrasonic frequencies used in this application are relatively low (near-audible), extending upwards from the upper threshold of human hearing to around 100 kHz in frequency. These frequencies are easily generated, captured and recorded by much existing audio equipment – and can be readily processed in real-time with available hardware. In this chapter, the basic principles of ultrasonic speech technology and the most important issues concerning its implementation and signal processing are addressed. Most importantly, a modelling scheme describing ultrasonic behaviour within the VT is proposed and proven as the theoretical basis and the main framework of implementation of ultrasonic speech. This chapter concludes with open areas of research concerning the technology.