- „Smart Sensors for Future Wearable Devices and Implants”
- „Reincarnation of Ballistocardiography – A Sensor System for Holistic Health Care Applica-tions“
- „Engineering AI: from sensors to embodied systems”
09:00 „Smart Sensors for Future Wearable Devices and Implants”
In recent years, smart sensors and intelligent sensor systems enabled a tremendous suc-cess of biomedical wearable devices in the health, fitness and consumer market. These de-vices, that for example measure the heart rate activity, monitor a personal body condition or a patient’s rehabilitation process, are ubiquitous in todays life. The success of these sys-tems is based on smart sensors, integrating classical sensing techniques with advanced and intelligent smart functionality into highly miniaturized and specialized ultra-low-power sen-sor systems.
In future, techniques like edge computing, artificial intelligence based signal processing and ultra-low power circuit and system design will lead to an even higher seamless integration of the smart sensors with the body and to the development of electrically active implants. The functionality will go far beyond the well-known pace maker and hearing aid devices; towards wearable brain-computer-interfaces, epileptic seizure tracking and medication de-vices and devices for active prosthesis control.
In this talk, I will exemplify with results from my research, which challenges have to be addressed in this field of sensor application and how smart sensors contribute to the de-velopment of future electrically active biomedical wearable devices and implants. An im-plantable device for the electrical monitoring of brain activity is presented, a promising alternative approach for the monitoring of brain activity based on biomagnetic sensors and an edge-computing device for epileptic seizure detection using a convolutional neural net-work.
Meeting-ID: 865 6847 9609 / Kenncode: 318620
11:00 „Reincarnation of Ballistocardiography – A Sensor System for Holistic Health Care Applica-tions“
Abstract: Ballistocardiography (BCG) is the measurement of heart and cardiovascular related acceleration of the body to enable broad health diagnostics. Actually, BCG research has its roots in the 19th century while modern research interests origin from space flights in the late 60’s. However, due to the evolving echocardiography techniques, BCG fell into oblivion as a possible tool for heart diagnostics. Research activities increased again since 2010 as, mainly pushed by smart phone industry, precise and inexpensive COTS MEMS-Sensors (e.g. accelerometers) have become available. This talk will provide a brief overview on BCG and its opportunities for health care applications, while technically the challenges and solutions for the measurement of tiny accelerations are discussed in detail. The SoC/FPGA based im-plementation enables concepts of e.g. differential signaling on sensor level while signal pro-cessing tasks can be shifted towards the sensor system in an efficient manner for further integration. Beside a custom prototype implementation, which defines the current gold standard for BCG measurement, current research activities and upcoming projects are pre-sented.
Zugangscode: 843579 – Bitte den Browser Chrome nutzen!
10:00 „Engineering AI: from sensors to embodied systems”
Abstract: To steer AI in the direction that is particularly profitable for the development of our infrastructure, environment, and medicine, machine learning methods have to be evolved to deal with the challenges such as multiple sensor modalities, multi-task learning, or high environmental variability Models should be sample-efficient and adapt quickly to new con-ditions. To this end, we focus on directions, such as meta-learning neural network architec-tures, multimodal Bayesian modeling of time-series and change detection, scaffolding for learning with reinforcement. Furthermore, we will discuss research opportunities and syn-ergies that are created by bringing together novel robot platforms, sensor technology, and machine learning to enable us to engineer AI.
Zugangscode: 063263 – Bitte den Browser Chrome nutzen!
Eine Anmeldung ist nicht erforderlich.
Mit freundlichen Grüßen,
Prof. Dr. Volker Turau