We created the Institute for Signal and Information Processing (ISIP) in 1994 with a simple vision to develop public domain speech recognition software. We used to think of our organization as "GNU for DSP" since we were strongly influenced by the success of tools like emacs and gcc. It was a long uphill battle to get this vision funded, so progress was slow. Fortunately, in 1997, things fell into place, and ISIP embarked upon a serious effort to develop a speech recognition system.

It has been a pleasant surprise that our modest vision has now migrated into the mainstream via a movement called Open Source. Powerful forces behind this movement include the Free Software Foundation, the Linux operating system, and even Sun Microsystem's recent announcement that it is making the source for its Solaris operating system freely available. Within the speech research community, we are seeing similar trends as many sites are now following the open source paradigm.

The term open source, however, means different things to different people. At ISIP, our interpretation is unmistakingly clear: our code is placed on our web site in an unrestricted fashion - no licenses, no copyrights, etc. We prefer the term public domain since this is the term originally used to describe such software. Too often, open source means the software is restricted to be used for research purposes only, and cannot be used for commercial purposes without negotiating licensing fees.

The lack of freely available state-of-the-art speech recognition tools seriously hinders the development of audio information processing technology. The lack of associated infrastructure to run large-scale experiments continues to impede progress. It is the goal of our program to facilitate research and technology development by individuals, small groups, and researchers new to the field. We want to make it as easy as possible for you to get started.

Sun Microsystem's former CTO, Bill Joy, once explained Sun's interest in the Internet and open systems as a way to increase the intellectual pool of scientists working on a problem of interest to Sun, thereby giving them tremendous cost-savings over monolithic corporations. The Internet's impact on education has been unprecedented and unparalleled. Its use to accelerate technology development in areas such as high performance computing has been impressive. Therefore, the primary goal of our program is to to increase the pool of researchers working on fundamental speech problems. This can only happen if we streamline the start-up costs of such research.

It is important to understand that ISIP's mission is to create a lasting infrastructure. Support, training, education, dissemination of information are all very important to us. For example, if you download our software and have trouble building it, we want to hear about it (send email to ies_help@cavs.msstate.edu). If you can't get an experiment to run properly for some reason, let us know. More often than not, we will give you immediate feedback and work closely with you until the problem is resolved. If there are new features you want added to our software, let us know. Better yet, attend one of our public design reviews to be held in an annual basis starting in Spring'00. The feedback loop is a very important tool for improving our technology.

Finally, we want to welcome you to the ISIP speech recognition web site, and hope you find it to be useful. We always encourage feedback if you have the time.

Best regards,

Joe "The Terminal Man/" Picone
Director, ISIP