Enhancement of electrolaryngeal speech by adaptive filtering

Artificial larynges provide a means of verbal communication for people who have either lost or are otherwise unable to use their larynges. Although they enable adequate communication, the resulting speech has an unnatural quality and is significantly less intelligible than normal speech. One of the major problems with the widely used Transcutaneous Artificial Larynx (TAL) is the presence of a steady background noise caused by the leakage of acoustic energy from the TAL, its interface with the neck, and the surrounding neck tissue. The severity of the problem varies from speaker to speaker, partly depending upon the characteristics of the individual's neck tissue. The present study tests the hypothesis that TAL speech is enhanced in quality (as assessed through listener preference judgments) and intelligibility by removal of the inherent, directly radiated background signal. In particular, the focus is on the improvement of speech over the telephone or through some other electronic communication medium. A novel adaptive filtering architecture was designed and implemented to remove the background noise. Perceptual tests were conducted to assess speech, from two individuals with a laryngectomy and two normal speakers using the Servox TAL, before and after processing by the adaptive filter. A spectral analysis of the adaptively filtered TAL speech revealed a significant reduction in the amount of background source radiation yet preserved the acoustic characteristics of the vocal output. Results from the perceptual tests indicate a clear preference for the processed speech. In general, there was no significant improvement or degradation in intelligibility. However, the processing did improve the intelligibility of word-initial non-nasal consonants.