As I am sitting here, writing this, we are experiencing record-breaking high temperatures.
Yesterday, on my way back from client, I caught my dashboard thermometer displaying 37°C, which is equivalent to 98.6°F. I was extremely grateful for the air-conditioning in my car!
However, neither my office, nor my home office are fitted with proper climate control. Usually this is not an issue at all, but this summer is somewhat extreme. My home office is the coolest of the two. And I am currently measuring the temperature of 30.1°C (86°F) here.
The only reason I am still able to do some productive work, is the fact that I have a fan blowing constantly and I have a wet towel in my neck.
It may surprise you to hear that right next to the fan is a portable airco unit. It is not even broken, if I switch it on it will work just fine. So you may wonder why I’m not using it.
Unfortunately, the amount of noise that the unit produces totally negates the usefulness of the drop in temperature. After all, I need to be able to dictate with my speech recognition software. The airco unit is… problematic in that regard.
So today I did a little bit of research concerning the use of fans and air-conditioning while dictating with Dragon. I’m not the only one who needs to work in a hot environment.
How much noise do fans and air-conditioning units make ?
Earlier today I installed an app on my iPhone which is capable of measuring decibels.
The results cannot be considered scientific, as I have not used a calibrated microphone, but still it gives an impression of the difference in sound levels between the fan and the air-conditioning unit.
My iPhone measured 42 dB from my fan blowing at its lowest level.
The air-conditioning unit produced 63 dB. At first glance it seems that the air-conditioning is only about 50% louder than the fan, but that turns out to be a wrong conclusion. My physics knowledge is not very impressive, but I know that sound isn’t measured on a linear scale.
Fortunately my husband is a physicist and he came to the rescue: if the sound level is 20 dB higher, the sound pressure is 10 times more.
What are the consequences for speech recognition?
Here my experiment turned entirely practical: I’ve been dictating for hours today with the fan quietly blowing. My impression is that recognition rates are not influenced at all by the background noise of the fan in its lowest speed.
The results are very different when I switch on the air-conditioning unit: the volume indicator of Dragon’s microphone immediately shows that Dragon has started processing sound. I can dictate and Dragon will get most of my words, but it is absolutely obvious that recognition rates are below par. Dragon also seems to need extra time to process the calculations, for speed performance is worse as well.
Dragon’s most recent version, version 15, continuously learns from the dictation. To me it seems that any background noise is also included in Dragon’s processing. Earlier versions of Dragon allowed you to decide not to save the speech profile after session, but this is not the case with Dragon 15. The speech profile is automatically saved sooner you close Dragon.
I suspect that dictating with continuous background noise will negatively impact recognition rates in the long run. Admittedly, I have no proof of this and I am not prepared to experiment long-term to find out if this is true. These types of experiments are pretty hard to set up in such a way that you can come to a clear conclusion. Still, I myself will not feel easy about using Dragon with significant background noise.
What to do??
What if you can’t escape the background noise?
Create a backup of your current Dragon speech profile. If it later turns out that your recognition rates are deteriorating due to continuous exposure to background noise, you can always get back to your current situation. And lets be candid, if you don’t regularly make a backup of your profile, you’re in trouble if the PC or Dragon has a tech issue…
You can download a cheat sheet about how to create a backup of your user profile by filling out your email address here and joining the free content of the LearnSpeechRecognition Academy.
You’ll be signing up for the newsletter then as well, and you will have immediate access to other the free resources.