Computer Speech and Language (2001) 15, 335 doi:10.1006/csla.2000.0170 Available online at http://www.idealibrary.com on
Obituary M. W. Macon 1969–2001 On 15 March 2001, at the age of 32, Mike Macon died during a heroic battle with lymphoma due to secondary complications of his treatment. Mike worked in the area of speech synthesis for less than a decade. Nevertheless, in those very few years, he had a phenomenal impact on the field, both by his scientific contributions, his activities in national and international organizations, and his work as a faculty member at OGI. Few individuals achieve his breadth of expertise at so young an age. This impact was due to an exceptional combination of creativity, dedication, and, most importantly, personality. Michael Macon first entered Georgia Tech in 1991, a young graduate student with much promise. He had a great propensity for digital signal processing and mathematical analysis and quickly distinguished himself as a top signal processing graduate student. Mike was bright, creative, and adventurous. While still a graduate student, he would travel to laboratories in France, Florida, Texas, New Hampshire, and Great Britain for research collaborations. As a graduate student, he became a mentor to many others, often delaying his own progress as a result. After receiving his Ph.D. from Georgia Tech in 1996, Mike joined the Oregon Graduate Institute as a faculty member in electrical engineering and a member of OGI’s Center for Spoken Language Understanding. While at OGI, his creative work was rewarded by the prestigious NSF Career Development Award in 1999, as well as by several research grants and honors. He was also an important contributor to public domain speech processing toolkits. Mike focused on signal processing algorithms for speech modification, which many believe to be a critical area for the next generation of speech synthesis systems. Mike’s work has significantly impacted the trend in speech synthesis towards “concatenative” systems, in which speech was created by splicing stored natural speech fragments instead of using equations and models as in earlier work. A major challenge in this approach is that, for speech output to have the appropriate timing and intonation, the stored speech fragments must be modified using signal processing techniques, which can cause audible distortions.As storage costs continue to drop, some researchers have been tempted to design systems to minimize the need for such modifications, reducing synthesis to a search and data annotation problem. Macon instead attacked the problem of signal modification directly, developing methods to reduce distortion associated with concatenation and prosodic mismatch between target and stored speech units, as well as techniques for voice adaptation. He provided a framework for introducing model-based techniques into concatenative systems, which is good news for those who would like to see a close link between speech science and technology but also for those who understand that the combinatorics of language exceeds anything one could conceivably store. Mike Macon was an active member of the research community, and he has had a great influence on both the technology and the people with whom he worked. All of us in the speech community will miss him greatly. This issue of Computer Speech and Language is dedicated to Michael W. Macon. Mari Ostendorf, Jan van Santen, Mark Clements 0885–2308/01/030335 + 01
$35.00/0
c 2001 Academic Press