I did some work during my PhD on recognizing scenes from electronic-nose data. The problem is more on the hardware side of things than the algorithm side of things. Today's electronic noses are still pretty large devices in which a sample needs to be placed within the device. Unless the state-of-the-art has advanced significantly in the past 3 years, none of them are like wielding an artificial nose on a stick that you can just wave around sniffing things.