In-Sensor Computing: Opportunities and Challenges
Many sensor systems measure quantities that cannot be described by simple physical models, and thus analytical techniques can be useful to infer underlying information out of noisy signals. However, both data computation techniques and radio transmission of raw recorded data can be prohibitively energy-expensive for wearable, implantable, or small IoT-like devices. The purpose of this e-workshop is to explore techniques that do preliminary computations such as feature extraction closer to the sensor, either in the analog or digital domains, with and without precise and/or reliable computational elements, to ease radio communication power, computation power, or both. This workshop explores where to place processing steps by exploring multiple layers of design hierarchy: devices, circuits, and algorithms; all of these layers which must work together to yield significant device-level power reductions.
|In-Sensor Computing: Opportunities and Challenges|
Tuesday, April 25, 2017, 10:30 a.m.–noon CT
Urbana, IL, United States