No More Foul Fowl

Electronic noses were first conceived of decades ago, and since that time the technology has steadily improved. Today, these devices are successfully being utilized to monitor diseases, detect food spoilage, recognize dangerous gasses, and more. As technology continues to advance, electronic noses hold the promise of even greater applications in fields such as healthcare, environmental monitoring, and quality control in various industries, making them a crucial tool for enhancing safety and improving our overall quality of life.

To implement their advanced functionalities, these artificial sniffers most typically contain arrays of sensors, with the data they capture being forwarded to external compute resources consisting of analog-to-digital converters, microcontrollers, memories, and other processors. For the most accurate results, the sensor data is commonly processed by machine learning algorithms that help to make sense of it.

This separation of sensing, processing, and memory — the traditional von Neumann computing architecture — introduces certain limitations into the resulting artificial noses. They tend to exhibit, for example, high levels of energy consumption, excessive latency, and even a loss of data. For many applications, these factors are roadblocks that prevent the adoption of the technology.

In order for the field to continue moving forward, a new architecture will be needed that addresses the present technological limitations. Recently, a team of researchers at Seoul National University in Korea have put forward a potential solution. They have developed an electronic nose chip with near-sensor computing that exhibits low power consumption, minimal latency, and no loss of data. This device was tested and shown to be highly accurate in detecting food spoilage.

The researchers’ device was designed to mimic biological olfactory systems. It consists of zinc oxide sensing films that can detect even very low concentrations (0.01 parts per million) of hydrogen sulfide and ammonia gasses. These gasses are important markers of spoilage in high-protein foods like meat.

When the zinc oxide films interact with the target molecules, it triggers a charge transfer that is directed into an amplifier array that enhances the signal. The amplified signal is then fed into an AND-type nonvolatile memory (NVM) array on the same chip that can be programmed to perform in-memory computations. In the NVM array, olfactory sensing data are linearly combined and multiplication and accumulation operations are performed for processing and interpretation. Rather than raw sensor data, meaningful information about the presence of specific scents is the output from the chip.

The artificial nose chip can be modified, by changing the sensing material, to detect the presence of many types of scents. The initial prototype, however, was tailored to recognize gasses that indicate the freshness level of meats, like chicken. By recognizing the target gasses that are present, the chip uses an algorithm that linearly combines the two gas concentrations to classify meat as fresh, edible, spoiled, or completely spoiled.

To test the system, the sensor chip was placed near a chicken tenderloin that was left at room temperature for an extended period of time. As the meat went through various stages of spoilage, it released the target gasses in different concentrations and proportions. Over the course of a twelve hour period, the sensor showed the chicken transition from fresh to completely spoiled. And unlike visual inspection, the device was shown to be capable of detecting when the chicken was anything less than fresh — visual appearances are only sufficient to detect completely spoiled chicken.

The team believes that the accuracy and utility of their electronic nose chip can be further enhanced by incorporating the detection of a variety of volatile organic compounds. Hopefully we will see further advances in the future that improve the performance of this proof of concept, but the techniques already demonstrated are expected to move the field forward.