Background

Analysis of physiologic signals consists of several steps and is, in many cases, both complex and computationally expensive. Different physiologic signals also require different processing steps, i.e., an electrocardiogram (ECG) used as the basis for studying heart rate variability is not analyzed in the same fashion as an electroencephalogram (EEG) used to study, e.g., vigilance. Although the exact analysis steps required for different signals vary, there the general goal of analysing physiologic signals is to perform feature extraction, i.e., to obtain metrics describing the signal, such as average heart rate or power in a certain EEG frequency band. The continuous physiologic signal is hence transformed into metrics that are usable as features in, e.g., machine learning tasks.

In many cases, physiologic signals are analyzed individually, and conclusions are drawn separately from different signals. However, more value can be obtained from the physiologic data if metrics from different signals are fused, since different physiologic signals might provide complementary information concerning the underlying physiologic state.

Recent research promotes the idea of using indices representing cognitive state for adaptive control of systems in human-computer interaction. To achieve this, the physiology must be analysed online. Also, the context can determine which physiologic signals are relevant for constructing indices representing some cognitive state, such as, e.g., mental workload.

The goal of the MIDAS system is to provide a modular framework to facilitate

• adding new physiologic signals to the analysis system
• construction of online analysis modules
• integration of processing of physiologic signals into machine learning systems.
• make MIDAS compatible with the internet of things (IoT)