One of my many vices is reading the weekly column in the Centre Daily Times called “News of the Weird.” The column features preposterous stories, ranging from incompetent criminals, e.g., masterminds who rob a bank while wearing a name tag, to strange practices such as the man in New York City who frequents bars and wants people to walk on him, and many other absurd anecdotes.
Occasionally, there will be a note about new technology or hard-to-believe science studies. A recent column (July 5) mentioned the technology company CrazyLobo in Fukuoka, Japan, which markets two personal-hygiene robots that emulate having a sense of smell. One robot resembles a dog that checks a person’s feet. If the feet “smell good” (i.e., have no odor), the robot cuddles up to the subject. Conversely, if the subject’s feet “smell bad,” the “dog” appears to pass out. I won’t discuss the robot shaped as a female who “kisses” a person and comments on their breath.
Reading this got me to thinking about the emerging area of “computer smell.” Most of us are at least somewhat familiar with the concept of computer vision, in which image processing and machine-based learning (pattern recognition) techniques are applied to images so that a computer can emulate human vision; recognizing shapes, faces, and even the elements of a scene.
In the College of Information Sciences and Technology, Professor James Wang is a noted expert in this area and has developed numerous techniques for semantic indexing and labeling of images. His techniques involve image processing, extraction of feature vectors to represent the image, and machine-learning (pattern recognition algorithms) to match images with associated semantic labels, similar to how humans would label an image. One of his IST Ph.D. students developed methods to emulate how art experts can spot a forgery.
It is a natural extension of this area of research to consider the sense of smell. What would be the value of such an artificial nose? Extensive studies have confirmed that a dog’s sense of smell is 10,000 to 100,000 times that of a human. A trained dog can use their sense of smell in a wide variety of applications such as detection of diabetes, early detection of ovarian and lung cancer, detection of early warning signs of weather changes, and detection of field mines. This augments well-known applications such as forensic dogs used to find victims of crime or casualties in natural disasters. Dogs have even been trained to call 911 in the event of an impending medical emergency in their owners.
There are a number of challenges in creating an artificial dog nose, including:
- Understanding the structure of the dog’s sensing “system”
- Understanding how air and odors flow inside a dog’s nose
- Understanding the sensing mechanism (e.g., the molecular exchange between air and the sensing membrane)
- Designing and building an artificial sensor
- Developing the signal processing feature extraction, and machine-learning algorithms for recognizing and identifying “smells.”
Penn State researchers are working on all of these areas. Researchers such as Brent Craven, from Penn State’s Applied Research Laboratory, and Gary Settles,Distinguished Professor of Mechanical Engineering, and colleagues, have focused on the fluid dynamics of air intake and molecular exchange, the development of micro- and nano-scale sensors, and signal processing algorithms for “smell detection” and characterization. The techniques developed by James Wang and our IST students could be applied to develop semantic labels to characterize the sensor output (e.g., “smells bad”). Thus, this research involves collaboration among diverse academic disciplines.
The creation of artificial dog noses would have broad applicability for improved medical diagnosis, threat warning systems, and weather prediction. The artificial dog nose would not require walking, feeding, or removal of dog hair from furniture. The down side is that the artificial dog nose alone would not provide the comfort and pleasure of a loyal dog.
I encourage dog owners to take notice of your dog’s behavior the next time you go for a walk. Try to imagine the olfactory “view” of the world that they perceive compared to your limited view. Conversely, for cat owners, when you observe the behavior of your cat – try to determine what planet they think they’re on!