Sensor signal processing
Fraunhofer Institute for Photonic Microsystems
- Dies of an UHF transponder ASIC with IPMS_16CXX core.
- Recognition and classification of various motion patterns.
© Fraunhofer IPMS
Sensor data are in many cases ascertained peripherally and subsequently transferred to more powerful processors for further interpretation. Due to reasons of energy consumption, wireless data transmission for portable systems can oftentimes only be implemented in systems with low data transmission rates, making transmission of large amounts of data extremely time-intensive. Fraunhofer IPMS's aim is therefore to process data as close to the sensor as possible, thereby substantially reducing the amount of transmitted data. Another advantage of near-sensor signal processing is the ability to link together data from multiple sensors to improve the resulting interpretation.
Online ECG interpretation poses an example: Through the integration of motion analysis the influence of body strain on an ECG test is taken into account. The measured data are processed through various digital filters and transformational procedures, and are subsequently classified. Self-learning nets, such as neural nets or gases, are used alongside statistical procedures. The collected, compressed results can then be transferred to a higher-level system for further processing.