Calatroni, Alberto, Dr.

Alberto Calatroni

ETH Zürich
Dr. Alberto Calatroni
Institut f. Elektronik
ETZ H 63
Gloriastrasse 35
8092 Zuerich

Phone: +41 44 632 33 91

Dr. Alberto Calatroni received his degree in Electronic Engineering from Politecnico di Milano, Italy. He worked in the Image and Sound Processing Group on sound steering through loudspeaker arrays and joined the group as a Jr. Researcher after finishing his Master thesis. There he worked for two years on algorithmic as well as software/hardware aspects. He pursued his PhD in the Wearable Computing group within the Electronics Laboratory at ETH, from which he graduated in 2013. He worked on the FET-Open project Opportunity, specifically in the direction of dynamic adaptation and autonomous evolution in context recognition systems. He is co-founder of Bonsai Systems, a technology startup which provides an ultra-low-power sensing platform with applications ranging from smart tags using iBeacon microlocation technology for retail markets to smart musical instruments.


In the European Project (FP7) Opportunity, a consortium of four groups (Wearlab at ETH, Passau, EPFL (Lausanne) and JKU (Linz) developed a set of algorithms and a framework which pushed the boundaries of human activity recognition by solving many challenges with respect to training sensor systems with milder assumptions compared to the state of the art. The motivation behind the project was that in reality, sensors just happen to be present whenever a user wears them or finds him or herself in proximity of them. Thus, they should be used "opportunistically".

The following list shows a subset of the methods developed:
- New sensors can be integrated into an existing system by training them in-place, through machine learning, without the need of any labeling process. The labels are provided by the system.
- Mappings between modalities provided by different sensors (e.g. acceleration and position) are established automatically, so that the framework can substitute a sensor with another, without a sharp degradation of system performance. For example, the system learns that acceleration can be derived from position, thus an accelerometer can be used to substitute a position sensor, after proper conversion of the training data and the features.
- Anomalies in a set of sensor systems can be detected and automatically corrected for, in order to avoid that a degradation in the performance of a sensor affects heavily the performance of the activity recognition system.
- Long-term drifts in sensor readings can be tracked and the system can adjust for them by a self-calibration process.
- The Opportunity framework integrates these and many other methods, combined with an ontology, so that the system can decide how to use the best resources available at a given place and time.

Research interests


2016 2015 2014 2013 2012 2011 2010 2009 2005 2003

Wichtiger Hinweis:
Diese Website wird in älteren Versionen von Netscape ohne graphische Elemente dargestellt. Die Funktionalität der Website ist aber trotzdem gewährleistet. Wenn Sie diese Website regelmässig benutzen, empfehlen wir Ihnen, auf Ihrem Computer einen aktuellen Browser zu installieren. Weitere Informationen finden Sie auf
folgender Seite.

Important Note:
The content in this site is accessible to any browser or Internet device, however, some graphics will display correctly only in the newer versions of Netscape. To get the most out of our site we suggest you upgrade to a newer browser.
More information

© 2017 ETH Zurich | Imprint | 4 August 2015