Speech Recognition: SoundMax Cadenza


During the past ten years, the computer industry has seen speech recognition technologies arrive with a flash and quietly fold without a whimper, as most of these technologies have lacked decent translation capability and just didn't work very well. They boasted the ability to control Windows by voice and to dictate to Microsoft Word, but even after many hours of "training" the software, in the end it was usually easier to just type the text with the keyboard. In this article, THG will examine speech recognition technology of today, how it is integrated into Windows XP, and some promising, new speech recognition technologies.

Speech recognition research has its origins in the late 1940s when the U.S. Department of Defense saw a need for and supported the first initiatives into speech recognition equipment. The purpose of the technology was based in espionage and national defense, far from the commercial types of uses that are being explored today. Over the next thirty years, companies such as Bell Laboratories developed commercial applications for voice recognition software programs. Basic programs were developed that could recognize single words and simple spoken numbers, zero through nine. However, adequate processing power did not yet exist to support the speeds required to support speech recognition software.

As processing power in computers has dramatically increased, the number of companies participating in this arena has, also. IBM and Dragon Speaking are two of the largest companies involved in the development of speech recognition technology. Both companies have invested vast amounts of time and money in the development of this technology for various products; particularly in each company's own speech-to-text dictation software.

Speech recognition has finally found its way into commercial systems, from the telephone number information database to over-the-telephone broker systems. These "people-free" systems are useful and cost effective for the companies that employ them. However, many customers find these systems hard to use, often annoying and time consuming, and sometimes error prone.

Some of the information above originated at History of Speech Recognition .