Talk about it...
This article covers speech synthesis - your computer talking to you - and speech recognition - you talking to yourself, sorry your computer.
Speech Synthesis
Most of the mainstream PC controllers include the option of speech synthesis and speech recognition. I think it's fair to say that these are built on one of three prevailing technologies: Microsoft Speech Agent, the Festival Speech Engine or AT&T; Natural Voice.
All three are capable of performing text-to-speech conversion - taking a typewritten sentence and speaking it through the computers built in sound card. The quality of the conversion, intonation and inflection varies significantly from product to product. Speech synthesis itself is also a relatively CPU intensive task, and benefits from a powerful CPU. Of the three engines, AT&T; natural voice is vastly superior in terms of quality, and is able to speak with a British accent (male or female), and well worth the USD $50 investment required. The Festival Engine, whilst free, sounds very mechanical, and is less polished in it's pronunciation in my opinion. Microsoft Speech Agent falls somewhere between the two stools.
It is important to remember that speech synthesis on its own is only half the story. The main application of speech synthesis in the HA environment to date is for background announcements initiated by the control system.
Speech Recognition
Speech recognition is notoriously difficult, and the technology isn't really ready for prime-time use yet. Microsoft speech recognition engines require training, but once trained can be reasonably reliable if used in a consistent environment. The problem is that the environment, even in the home, rarely is consistent -- various background noises, whether casual conversation or the TV, tend to screw things up. The situation can be marginally improved through the use of a wireless microphone, but this isn't very practical. The equipment involved in mixing and compounding free air speech, only to achieve marginal reliability, doesn't make the technology worthwhile in my opinion.
|
|
|
