Health monitoring by tensor-based signal processing

As people's health cannot be overrated and 'prevention is better than a cure', monitoring one's health is a key process. For increased comfort, such monitoring should preferably be using contactless technologies that preserve the privacy. Radar systems have proven useful in this context but are currently being used mainly in controlled environments. On the other hand, in sub-optimal conditions, e.g., in the presence of additional people or high levels of noise, further study of the problem is required. In this PhD trajectory the signals acquired by multiple-input multiple-output MMWave radars in an environment with multiple moving persons will be analyzed with tensor techniques. These have ample perspective to discover fine-grained semantics describing both the subtle movements related to vital signs as larger ones such as raising an arm for each individual. Different tensor decompositions will be used and compared. Furthermore, links between tensor decompositions and techniques from machine learning, e.g., neural networks will be studied. Attention will be paid to the theoretical as well as to the practical aspects of the proposed research.