Voice replacement for the severely speech impaired through sub-ultrasonic excitation of the vocal tract

• Main Author: Ahmadi, Farzaneh
• Other Authors: Ian Vince McLoughlin
• Format: Theses and Dissertations
• Language: English
• Published: 2013
• Subjects: DRNTU::Engineering::Computer science and engineering::Mathematics of computing::Numerical analysis; DRNTU::Engineering::Bioengineering
• Online Access: http://hdl.handle.net/10356/52661
• Institution: Nanyang Technological University

Post-laryngectomized and speech impaired patients lose the ability to create natural pitch, and therefore seek methods to provide replacement voicing for their speech. Non-invasive pitch generation can be accomplished by coupling an artificial pitch into the vocal tract (as in the electrolarynx). However this audible pitch leaks out, causing a monotonous and continuous background noise which significantly degrades the speech quality. This research aims to investigate the idea of substituting the missing pitch of such patients with a low frequency ultrasonic signal, propagating inside the tract to map the vocal tract (VT) shape. The emitted “ultrasonic speech” signal can then be down-converted to generate elements of audible speech. This technology can use the VT physiology in a similar way to the well-established electrolarynx, but the ultrasonic excitation conveys advantages of being non-audible (i.e. no pitch leakage), analysable by computer, and adjustable without directly affecting the speech quality. Other applications of the technology include: speech communications in high noise environments such as factories and battle fields, as well as secure/quiet mobile telephony modes to prevent others from overhearing sensitive communications. Another application is supplementary data provision in the reconstruction of voiced speech from whispers. Ultrasonic speech has received little research attention before the course of this research and the current state of the art of the technology is still far from achieving a complete system. Accordingly, this research aims to take a fundamental step towards ultrasonic speech deployment, using physical analysis and modelling of ultrasonic propagation inside the human vocal tract. Major questions including the differences between attributes of audible and ultrasonic output of the human vocal tract and the possibility of extending the source-filter theory of audible speech to the ultrasonic domain are answered in this thesis. This work proves that the linear source-filter theory can still be valid for low frequency and sub-ultrasonic (14 to 100 kHz) propagation inside the VT. In addition, linear predictive analysis is extended from the one-dimensional sound propagation inside the VT in audible speech to the three-dimensional propagation of ultrasonic waves. The next important aim of this thesis is to gain a model that describes propagation of ultrasonic waves in the vocal tract. This has been accomplished through finite element analysis of ultrasonic propagation in the geometry of human vocal tract. The models are later constructed using three-dimensional printing and the results of the simulations are verified using acoustic measurements. In addition, since implementation of ultrasonic speech involves timely exposure of human body to low frequency ultrasonic waves, this research reviews the safety requirements of implementing ultrasonic speech which is a necessary requirement in testing and implementing the technology.