As some of you may know (but I did not until recently), there is a software package out there that helps you automate your home by Automated Living. Their product is called HAL and I found a length web page describing an implementation of it. The system is voice operated, though the simplest way is to use any telephone in your house. On the page linked above the person managed to get it to work with microphones with decent success. The main issues that came up with the microphones are
1) The acoustics of the room had an affect on the quality of the recognition. For instance recognition was much tougher in bathrooms with tile floors.
2) If other people are speaking in the room or some other noise occurs, recognition is very difficult.
Based on these results and from my conversations with others, it is clear that background noise and acoustics are going to be my biggest hurdles in getting this system to work. Of course I can hook up Speech Server to a telephone like HAL since Speech Server is designed for telephony but I have decided to go the hard route and see if I can have success in eliminating background noise.
My first recourse will be to see what support we have for this in the engine, but I will probably not be allowed to report those results here. More likely it seems that I will have to try some signal processing to isolate my wife and I's voice.
This is quite a tall task given that I do not know much about signal processing. So I found a used copy of the book used at MIT at Half Price Books and I have started reading it. Once I finish that, I will need to study up on speech identification. Upcoming blogs will focus on what I am learning, as that will help me learn and perhaps I will receive a few pointers by individuals more knowledgeable about the subject.