Electronic Nose Sniffs out Mold
It turns out, that mold is everywhere. The problem is when it becomes too much, as mold infestations can have serious health effects on both humans and animals. Remediation is extremely expensive, too. So there are plenty of benefits to finding mold early. Now, German researchers are proposing an electronic “nose” that uses UV-activated tin oxide nanowires that change resistance in the presence of certain chemicals, and they say it can detect two common indoor mold species.
The nanowire sensors can detect Staachybotrys chartarum and Chaetominum globosum. The real work, though, is in the math used to determine positive versus negative results.
Traditional methods take some sort of physical sample that is sent to a lab and require days to process. However, trained dogs can also smell mold, but as you might expect, there aren’t many dogs trained to find mold. Besides, the training is expensive, you have to maintain the dog all the time, and if the dog knows what kind of mold it is, it can’t say. So an electronic nose that can give fast, specific results is quite attractive.
Even if you don’t care about mold, the data crunching to classify the sensor data has application to many types of sensors. They used training to build multiple models, then they combine the outputs using a regression algorithm to predict the true output. Finally, they use a majority voting technique to combine the results of the model and the regression output.
Could you make a sensor like this? Reading section 4.2 of the paper, it looks like you need a pretty stout set of lab gear to play. But the math ideas are certainly something you could replicate or use as a starting point for your own sensor fusion projects.
Want a deep dive into sensor fusion? You should have been at the Hackaday Superconference a few years ago. Luckily, you can still watch [Christal’s] talk about fusing multiple streams of sensor data.