Philips NV is calling its low-cost speech recognition package – FreeSpeech 2000, launched at PCExpo in New York this week – the next step in desktop consumer speech software. The Dutch electronics company claims that the $99 application outperforms software that costs two or three times more.
Ralph Preclik, director of corporate communications for Philips Speech Processing, explained that, in common with all speech recognition packages, the software’s speech engine is based on Hidden Markov Models (HMM), statistical probability models that look for the closest match between the spoken information and phonemes, the fragments of spoken sound that make up language. FreeSpeech 2000 uses new high speed and accuracy algorithms (HSA) that delve deeper into the HMM tree in a shorter amount of term. Preclik says that the program is 30% faster and more accurate than the previous version of FreeSpeech.
Philips has also developed a combined microphone, loudspeaker and trackball controller for use with the software. Preclik said that this had been developed to allow users to have one control unit for their desktop. FreeSpeech 2000 will be available with US and UK English, French, German, Italian and Spanish versions on one CD-ROM.
