The Artificial Larynx: A Review of Current Technology and a Proposal for Future Development

Abstract

The loss of speech following a laryngectomy presents substantial challenges, and a number of devices have been developed to assist these patients, ranging from the electrolarynx to the unidirectional valve used in Tracheoesophageal speech. However, all of these devices have concentrated on producing the sound from the patient’s vocal tract.

A brief review of the current state of the artificial larynx will be presented, followed by an introduction and system-level overview of a new type of artificial larynx currently in the early stages of research and development by our group. This new device will utilize dynamic measurement of tongue position to infer intended speech, and will transmit these signals to an electronic unit for near-real-time speech synthesis. The dynamic tongue measurement is achieved with the use of an existing palatometer and pseudopalate. This device, consisting of over 100 contact sensors spread over the palate, detects the exact tongue-palate contact pattern in speech. The tongue contact positions as a function of time are presented as a 2-D time-space plot. A brief summary of the planned future work will be given, which will include proposals for traditional image processing techniques amongst others to be used for extraction of information. The extracted features will be fed into a suitably designed Artificial Neural Network. By using a combination of data from the palatometer and from other biological signals it should be possible to infer what the patient is saying. Other methods for data presentation, feature extraction and signal processing will also be considered.

It is anticipated that the intended speech will be reconstructed using Artificial Intelligence techniques, and that this information will be streamed to a speech-synthesizer. Algorithms will ultimately be developed to achieve a synthesized voice pattern as close to the patient’s original sound as possible.