Yu Jiang

A Low-Cost Surface Acoustic Wave Sensor for Classifying Biomaterials and Other Ubicomp Applications
Advisor(s)
Sai Ganesh Swaminathan and Prof. Scott Hudson,
CMU Human-Computer Interaction Institute Dev Lab, Pittsburgh
Status
Paper submitted to IMWUT'21 August round (first author)
Duration
September 2019 - present
Role
Team lead

Prototyping and characterizing the sensor
Data collection and signal processing
Sensor-embedded utensils
Surface acoustic wave (SAW) sensors are a promising class of microelectromechanical devices that leverage changes in the transmission of an acoustic wave across a sensing surface to detect physical phenomena. These devices can be small, versatile, and are extremely sensitive to even low-intensity events like a liquid drop. It may open whole new types of sensing opportunities in biomaterial monitoring and interaction. Unfortunately, despite the widespread adoption of mass-produced SAW devices by the industry, the cost and difficulty of fabricating custom SAW devices for research and prototyping have remained high.
We introduce a low-cost, simpler, and accessible method for fabricating custom SAW sensors. This method relies on commercially only available commercial printed circuit boards (PCBs) as the primary fabrication process and does not require cleanroom conditions, nor specialized equipment. This makes the fabrication of these devices much simpler and more accessible. The devices are then characterized for material interaction and classification. Finally, we show two feasible sensing applications that detect hard (such as solids and granules) and soft food types (such as liquids, and slurries). Our aim is to empower the research community to develop custom SAW devices more easily to explore novel sensing scenarios that will ultimately unlock new interactive applications.


Using a sandwiched PCB approach to reproduce the parameterized transducer patterns to prototype embeddable, cheaper, and more durable SAW sensors.


Prototyping interdigitated transducer patterns by depositing conductive ink on Lithium Niobate substrates with direct-ink write technique.
Data collection on biomaterials (liquids and solids) using a NanoVNA for signal transmission. We investigated influencing factors including hydrostatic pressure, weight, contact area, and biomaterial type.



Signal analysis was performed based on logmag and phase offsets at 90-110 Hz freqeucies. We proposed a machine learning pipeline with a SVM classifier for biomaterial classification.

We prototyped SAW sensor embedded "Augmented Mug" and "Smart Plate" everyday utensils capable of detecting and classifying biomateral in real-time.
Future work to be updated!