Electronic Nose

Definition

Electronic noses (or e-noses for short) are scientific instruments that detect, recognize, and quantify volatile chemicals in a manner that is loosely reminiscent of the sense of smell. The concept of an e-nose dates back to a 1982 article by Krishna Persaud and George Dodd, which showed that fine discrimination between odorants is possible without highly selective sensors. Instead, and much like the human nose, e-noses rely on cross-selective sensors; that is, sensors that respond to many different odorants, though the strength of the response varies with each odorant. To discriminate odorants, e-noses (and human noses as well) use an array of sensors with different cross-selectivities. In this way, the response pattern across sensors becomes a fingerprint for each particular odorant. In addition to chemical sensors, e-noses require a pattern-recognition system and an odor-delivery system.

Chemical sensors consist of two elements: a sensing layer and a transducer. The sensing layer is made of a material whose properties change when odorant molecules are adsorbed on its surface; as an example, the sensing material may change its electrical resistance or its mass. In turn, the transducer converts these changes into an output signal; as an example, a transducer may convert mass changes in the sensing layer into frequency changes of an output signal, which then can be easily recorded with a computer. Sensing materials for e-noses can be inorganic (metal oxides are typical), organic (e.g., conducting polymers), and biological (e.g., proteins). Typical transducers for e-noses are based on conductivity or mass changes, though other transduction principles can be used as well (e.g., capacitive, calorimetric, optical, field effect), sometimes simultaneously.

Most e-noses operate under two basic forms of odor delivery: static headspace and dynamic head-space. In static headspace analysis, a sample is placed inside a closed container along with the sensor array, and the headspace above the sample is allowed to equilibrate before recording the sensor array response. In the dynamic headspace method, an effluent (e.g., filtered room air) flows through the sample continuously, carrying odorant molecules to the sensor array located in a separate chamber. Several techniques (e.g., preconcentration, separation) can also be used to condition odorant samples before they are delivered to the sensors.

Once an odorant sample has been delivered to the sensor array and the corresponding response has been recorded, the signals are analyzed with a pattern-recognition system. Depending on the application, the goal of the analysis may be to classify the sample into one of several potential categories (e.g., [Page 384]fresh vs. spoiled milk), predict a set of properties for the sample (e.g., concentration, quality), or find samples that are similar (e.g., clustering).

...