Modern wearable devices allow monitoring vital parameters such as heart or respiratory rates, electrocardiogram, photo-plethysmographic or even video signals, and are being massively commercialized in the consumer electronics market. However, a common issue of wearable technology is that signal processing and transmission are power demanding and, as such, require frequent battery charges. To ameliorate this, in this talk we consider biometric signal compression as a means to boost the battery life of wearables, while still allowing for fine-grained and long-term monitoring applications. Toward this end, we propose two original dictionary based algorithms, where the dictionary is learned and maintained at runtime utilizing motif extraction, pattern identification and neural network techniques. Other recent approaches are also explored, including autoencoder architectures, compressive sensing and standard techniques involving Discrete Cosine, Wavelet transforms, principal component analysis and vector quantization. A thorough numerical analysis is carried out to assess pros and cons of the considered schemes. Dictionary and autoencoder based techniques are found to be the best options at very high compression efficiencies. As we quantify in our performance evaluation, these schemes allow reductions in the signal size of up to 70 (dictionary-based) or 100 (autoencoders) times, which entail similar reductions in the energy demand, by still keeping the reconstruction error within 4% of the peak-to-peak signal amplitude. Remarkably, our novel algorithm based on Self-Organizing Maps (SOM) is shown to provide excellent approximations, while adapting the dictionary in an online fashion, without requiring any prior information on the signal itself. Its performance is remarkable, as it provides high compression ratios (up to 80-fold) and excellent reconstruction performance (RMSE around 2%), while being able to dynamically adapt to the signal statistics of a new subject through the exploration of just half an hour of data.

After the talk, Michele and his Ph.D. student will show a demo of getting the biometric footprint of a walk from smartphone sensors and using this to identify who is carrying the phone.

About Michele Rossi

Michele Rossi is an Assistant Professor in the Department of Information Engineering (DEI) at the University of Padova, Italy. He is also a member of the SIGNET research group at DEI. Michele´s great passion is telecommunication networks, and his latest studies are centered on wireless sensor and cellular networks with energy harvesting capabilities.

Personal site

Este evento se impartirá en inglés