The first phase of a project whose goal is to allow people with ALS and similar conditions to "speak" through a computer in their own voices has come to a successful completion, says investigator Tim Bunnell of the Speech Research Laboratory of the Alfred I. duPont Hospital for Children and the University of Delaware in Wilmington.
The study is moving into a second phase, and researchers are seeking people familiar with augmentative alternative communication (AAC) programs.
|Tim Bunnell works on speech recording in his lab's sound booth. He’s trying to develop software that would generate speech in the user’s own voice.
Bunnell, who has a doctorate in experimental psychology and specializes in speech perception and production, is working with an academic rehabilitation engineering group. They hope to develop software that would allow people who expect to lose speaking ability to record and store a variety of sounds from which a computer will later generate speech.
The software now under development is far more sophisticated than the AAC programs now available, some of which use prerecorded phrases the user has banked. The new product would let the user to say anything in his or her own voice after selecting letters or words from a computer screen.
Grateful to early bug finders
During the first phase of the software development (see "Study to Evaluate Speech Synthesis," June 2002), the researchers recruited people with ALS and other disorders to try out the system and help discover its pitfalls and strengths.
Bunnell said, "We had a great deal of help from quite a few people, and we can’t thank them enough. We’re incredibly grateful to the people who helped us with the development phase. They did it in many cases with no guarantees that they would be helped, but only that they might be helpful to others."
The initial volunteer group, which included about 20 people with ALS, found that the software worked better for those able to produce a relaxed, natural speech without wide fluctuations in tone or emphasis, Bunnell said. Speech that was "breathy," as it can get in ALS, was all right, he said, but slurred speech was a problem.
The investigators also found that the software couldn’t handle certain accents, such as those used in Australia.
Phase 2 digs deeper
The goal of the second phase, Bunnell said, is to make the new speech software work with existing AAC products, such as EZ Keys software, a communication program available through Words+, and software available through Enkidu Research.
These programs use word prediction, a function designed to save keystrokes by allowing the user to enter an abbreviation or the first few letters of a desired word. The program then tries to "guess" the rest of the word or phrase and offers a list of choices to complete the entry.
Bunnell says the developers are seeking participants to install the new system on their home computers and experiment with it. He cautions that the endeavor at this point is "not trivial" and "not for the faint of heart."
The investigators are looking for people who are very familiar with and comfortable using computers, and prefer that users have an engineering background. Details and downloadable software can be found at www.asel.udel.edu/speech/DownloadInstr.html.
Volunteers will be asked to register but won’t be placed on any mailing lists or subjected to any advertising or fees.
Questions can be addressed to the text-to-speech project investigators at email@example.com, who will respond as time permits, Bunnell notes.
Flexible product is long-range goal
Bunnell says the speech-recording part of the system will probably be available in the future only through speech-language professionals, because of its likely high price tag and its complexity.
However, once a professional has assisted the user to put his voice into the system, the resulting program should be useable by anyone, anywhere.
With funding from the National Institutes of Health, the developers are now working with a Delaware software company to move their product from the academic to the commercial realm.
For continuing information about the project, visit the lab’s Web site at www.asel.udel.edu/speech.